3Dfx or Power VR? Why?
TormatoJES
I have a Lightspeed 128 (ET6000) and am going to buy a 3D card. Should I
get a 3Dfx-based card or one based on the Power VR chip? Why?
Thanks,
E-mail replies are always welcome
Jim Husband
On 1 Jun 1997 02:29:03 GMT, torma...@aol.com (TormatoJES) wrote:
>I have a Lightspeed 128 (ET6000) and am going to buy a 3D card. Should I
>get a 3Dfx-based card or one based on the Power VR chip?
3dfx - no question.
> Why?
The best way to determine this would be to scan DejaNews on
this newsgroup over the past 3 months. However - I'll list a few
points.
o Code support
3D cards have to be supported in code to have any value. There
are two methods of support - native code and generic (D3D, OpenGL).
Native code is almost always faster than a generic port - but will
become less popular with developers as the number of good 3D cards
increases.
There are a number of 3dfx native games (Glide/WinGlide) - and
many of these are major releases. The PowerVR has 1 (or 2) and they
aren't Descent II, MDK, or Tomb Raider type hits. Ultim@te Race or
somesuch is the only one I can think of right now.
o Speed
Despite some of the benchmarks floating around - the PowerVR
is significantly slower than the 3dfx in D3D. I don't really feel like
going through the "benchmark fiasco" story again - you'll find it
easily on DejaNews or Dimension 3D. You'll either think "No big deal"
or you'll vow to never buy a NEC/VL card. Remember - it happened
*twice!*
o Features
Lack of filtering and a weird (or revolutionary) method of
z-Buffering hurts the PVR, IMO. Filtering is one of the most important
improvements by 3D cards (second only to frame rate). It was added in
the PVR2 - so apparently NEC/VL realized it was a significant feature.
I'm not sure whether z-Buffering can even be done on a PVR without
writing in native code.
The 3dfx has all the standard 3D features - and they are
implemented so either native or generic ports can utilize them.
Anti-aliasing is the 3dfx's biggest gap - although it can be handled
by the driver (with alpha-blending by the card).
o Future support
This is always critical - and can never be more than a
prediction. The 3dfx is currently, IMO, ahead of the Rendition in
support by developers (albeit not by much). After Rendition would be
the Virge chips (just because of the sheer number of S3D cards).
PowerVR is somewhere behind these. Again - this could all change in
one fell swoop... but PowerVR isn't even in the top bracket right now.
I doubt they will be in the immediate future.
You didn't mention what your CPU is - and that should be a
consideration as well. The PVR has a pretty steep CPU requirement to
gain much 3D acceleration. Since one of the main reasons for getting a
3D card is to put off getting a faster CPU, this one has always
confused me. I'd have thought the design would have been modified once
this was determined.
I've almost certainly left out some key points. The PVR
(especially the VR2) has a few good points, but either a Rendition or
3dfx is a much stronger card at this point. Since you've already got a
LS 128 - the 3dfx would be the best choice.
Jim
* Email address is false - "Reply-to:" is correct
Rob
In article <19970601022...@ladder02.news.aol.com>, torma...@aol.com says...
>
>I have a Lightspeed 128 (ET6000) and am going to buy a 3D card. Should I
>
>Thanks,
>
>Jym
>Torma...@aol.com
>
3dfx.
Reasons, in no particular order :
Tombraider
GLQuake
Pandemonium
Wipeout 2097
OUtlaws
POD
Moto Racer - a real cracker
MDK
Terracide
F1 - if you like arcade racers
... and MANY MAN MORE! All available NOW either full version or demo.
and to look forward to:
Dark Forces II
I76 patch
EF2000 G+ - very soon
Flying Nightmares - 'June/July ?'
Flying Corps - patch very soon
Hexen II
Unreal
Lunatic
there are sooo many to list!
Almost everything that has been slated as a 3dfx titles has
appeared more or less on time. The same just cannot be said
for PowerVr.
Rob
Ty
Stephen Fordham wrote:
> *********POWERVR 2*************
> WRONG!!
> The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has all
> the features that where missing in the original PowerVR and it`s now widely
> supported by manufacturers and it`s got some cool features :
Much faster? In what benchmarks? So far I've heard about independent
benching from a few magazines and guess what, the PVr2 is BEHIND the
3Dfx, again.
> - The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
> speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
> so when the faster Pentiums are released the PowerVR will become even more
> powerful!
What does this mean, "..for the future..."? Heck, the PVr2 is ONLY for
the future as it doesn't even have a present as it's NOT OUT YET! Where
do you draw your information about CPU speeds from?
> -The PowerVR 2 comes with these 4 games:
> - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
> fps, bet the 3Dfx can't beat that!)
Guys have said it runs at 20+ fps AND I thought Kalisto said UR for the
3Dfx would look better.
> - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
> runs at 800x600 but in 65,000 colors at 35+ fps)
This one has been promised since the PVR1 (over 6 months ago).
> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
Now this would be nice. What kind of CPU do you need for this?
> - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
Said who? The PVR version wasn't as good as the 3Dfx one.
> Actually the PowerVR is a little bit slower in the Direct3D tests (about
> 1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
> PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
> FRIEND`s 3Dfx!)
Well your benchmarks and those conducted by others don't jive. I also
assume that you've tried renaming your test? What are you stressing
with the D3D test? Tunnel test? Many more agree that the benchmarks
prove the PVR2 is a wash, till they get better drivers anyhow.
> Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
Can't buy it now and it looks like there is ZERO reason to in the near
future as well.
Ty
Ty
Jim Husband wrote:
<SNIP>
> It would also be nice if they quit
> citing results from JPA's retracted benchmark test.
Yes, whatever happened to that test? Did JPA run it again?
Ty
Stephen Fordham
Jim Husband <jim...@computer.net> wrote in article
<3390f062...@204.127.4.22>...
> On 1 Jun 1997 02:29:03 GMT, torma...@aol.com (TormatoJES) wrote:
> >I have a Lightspeed 128 (ET6000) and am going to buy a 3D card. Should
I
> >get a 3Dfx-based card or one based on the Power VR chip?
>
*********POWERVR 2*************
WRONG!!
The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has all
the features that where missing in the original PowerVR and it`s now widely
supported by manufacturers and it`s got some cool features :
- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
so when the faster Pentiums are released the PowerVR will become even more
powerful!
-The PowerVR 2 comes with these 4 games:
- ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
fps, bet the 3Dfx can't beat that!)
- RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
runs at 800x600 but in 65,000 colors at 35+ fps)
- WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
- MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
Actually the PowerVR is a little bit slower in the Direct3D tests (about
1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
FRIEND`s 3Dfx!)
Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
N
"Stephen Fordham" <sbfo...@mail.telepac.pt> scribed:
> The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the
> 3Dfx
Based on what? 3D Winbench? PC Player's D3D Benchmark?
> and has all the features that where missing in the
> original PowerVR
Aside from bilinear filtering, what other features?
> and it`s now widely supported by manufacturers
Uh-huh.
> - The PowerVR 2 is for the future (so is the PowerVR 1),
> it supports CPU speeds up to 400MHZ while the 3Dfx only
> supports CPU speeds up to 200MHZ, so when the faster
> Pentiums are released the PowerVR will become even more
> powerful!
No way! You're telling me all those running the 3DFX
on an overclocked PentiumMMX 225, PPro 233, or a P2 266
are just imagining their 3DFX running? Riiiggght.
> - ULTIM@TE RACE (Racing game that runs at 800x600
> in 16.7M colors at 35+ fps, bet the 3Dfx can't beat
> that!)
Hmm, that's not what I hear from Videologic reps:
Glenn Booth <gl...@videologic.com> wrote on the subject
of "Re: ULTIM@TE RACING" dated 1/17:
>: Which is less than 2 megs, and therefore fits into
>: a 2 meg graphics controllers frame buffer. Note that
>: Apocalypse 3D can do 24 bit per pixel colour depths,
>: but in this case the developer chose 16 bit, since
>: the visual difference between the two is not that great.
> - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK
> style game that also runs at 800x600 but in 65,000
> colors at 35+ fps)
How would you go about running this on host 2D cards
with only 2 megs of RAM?
> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at
> 40+ fps).
Why not at 800x600? How do you access the frame rate
counter on WipeoutXL?
> Actually the PowerVR is a little bit slower in the
> Direct3D tests (about 1.5fps slower), BUT ONLY ON A P133
> and SLOWER.
Don't you have 3D Winbench numbers?
> On a FASTER system the PowerVR OUTPERFORMS the 3Dfx by
> 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A FRIEND`s 3Dfx!)
Try it on 3D Winbench or PC Player's D3D Benchmark.
BTW, you haven't bothered to even mention the benchmark
you use. Odd.
> Your BEST bet is to but the PowerVR 2 ( go to
> www.videologic.com)!
Sounds like we have another brain dead Andy Griffo clone.
Jim Husband
On 2 Jun 1997 23:12:54 GMT, "Stephen Fordham"
<sbfo...@mail.telepac.pt> wrote:
>*********POWERVR 2*************
>WRONG!!
I agree that the PVR2 is wrong - other than that we disagree
quite a lot. <g>
Since you have the card - I hope you enjoy it. Until I see
some verifiable results from glQuake, Descent II, Tomb Raider - I'll
wait and look at the PVR3... if they quit the benchmarking sleaze and
quit claiming to have Descent / Descent II versions "in development."
How long can it take to delete a few entries from the "Upcoming
Titles" table? Been about 6 months. It would also be nice if they quit
citing results from JPA's retracted benchmark test.
Jim
Alan Chiu
Ty wrote:
>
> Stephen Fordham wrote:
> > *********POWERVR 2*************
> > WRONG!!
> > the features that where missing in the original PowerVR and it`s now widely
> > supported by manufacturers and it`s got some cool features :
>
> benching from a few magazines and guess what, the PVr2 is BEHIND the
> 3Dfx, again.
Yes, 3dfx is always the winner!! Just read the following information in
a 3dfx newsgroup:
Germane magazine PC Games tested the new Videologic Apocalypse 3Dx
with the PowerVR PCX2 and other 3d cards.
Results on a Pentium 200 MMX (should be ideal scenario for the PVR2)
direct3d benchmark (d3dtest) polygon throughput
(640x480 16 bit Gourad Shading,Bi-linear filtering,Specular on + default
options on)
monster 3d 310
apocalypse 3dx 210
apocalypse 3d (power vr1) 198
terminator 3d (S3 Virge) 160
and they even used older driver for monster i got 303,55 with the
new 3dfx 2.08 drivers on a p 150
they didnt wrote what happens if u rename the d3dtest.exe neither ...
monster truck madness
monster 3d 24
apocalypse 3dx 15
terminator 3d 12
apocalypse 3d (power vr1) 9
Last test was Direct 3d tunnel test but this test tells nothing about
game performance.
They reported also that the Power VR2 has the same problems with
transparent textures as the Power VR1 e.g. black boxes in the game
moto racer.
--
Alan Chiu
3Dfx games lover
mailto:alc...@glink.net.hk
John L. Hardy IV
Hmm, I guess since I have it running on a 233mhz Pro it must be an
illusion. What a load of shit. Others have it in 266mhz PII's. It scales
upwards just fine thank you. 3dfx is garnering the most developer support
currently, and it shows no sign of abating.
And don't you find it interesting that your so called accelerator requires
such a high end base system to be half-competitive.. most of their
testing/benches have been done w/ Pro200's. Folks are getting better
performance on lower/mid range machines.. so not only is it the speed
leader, it's also the best bang for the buck. Then as others have said,
the duplicitous benchmark issues are enough to sway some.
--
john4 [at] GTE [dot] NET, Seattle, Wa
Asus p6np5 / iPentium Pro 233mhz / 64m / Mitsu. 21" 91txm
Matrox Mill. 8m & Orchid 3dfx / 5.1g Maxtor / Win95b FAT32
Plextor 12x & 8x / Iomega Jaz & Zip / Adaptec 2940uw / SB AWE32
Stephen Fordham wrote in article <01bc6fa1$e7defe20
:- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
ccorpora
In Article<01bc6fa1$e7defe20$0100...@host.telepac.pt>,
> WRONG!!
>
> - The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
> speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
> so when the faster Pentiums are released the PowerVR will become even more
> powerful!
R:
Are you saying a voodoo won't run on a P233 ect. ect. ect.. If so. WRONG!
> -The PowerVR 2 comes with these 4 games:
>
> - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
> fps, bet the 3Dfx can't beat that!)
>
> - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
> runs at 800x600 but in 65,000 colors at 35+ fps)
>
> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
>
> - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
R:
Ok where can I buy it right now. Or is it bundled with Motoracer too!
> Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
R:
If your a gambler then the safe money is on the voodoo. If your a hot-shot go
ahead a blow your wad on a PVR2. I'm far enough into the game where I'm
skipping generation 2.5 for 3rd generation stuff early '98.
Q.B.M.
Rex Nebula
>speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
>so when the faster Pentiums are released the PowerVR will become even more
>powerful!
>
Flash 3D. It must be using "cartoon logic" The Voodoo doesn't know
it can't work with my CPU, so it does until it finds out. Oops, now
it doesn't work anymore.
> - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
>runs at 800x600 but in 65,000 colors at 35+ fps)
>
Didn't they cancel the native port for that?
> - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
>
Which isn't even out yet, so you can't have any idea how fast it is.
I seriously doubt it uses 24bit color too, the speed hit on any card
is too much.
>Actually the PowerVR is a little bit slower in the Direct3D tests (about
>1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
>PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
>FRIEND`s 3Dfx!)
>
OH YEAH, WELL IF I SCREAM I CAN BE RIGHT TOO!
>Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
>
Your best bet is not to listen to this spewed pile of shit, and get a
3Dfx card.
Rex Nebula
neb...@capecod.net
S230...@umassd.edu
Gary Tarolli
On 2 Jun 1997 23:12:54 GMT, "Stephen Fordham"
<sbfo...@mail.telepac.pt> wrote:
>The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
>speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
>so when the faster Pentiums are released the PowerVR will become even more
>powerful!
I think you've been reading too many web pages...
Perhaps the truth is that a PowerVR 2 NEEDS a 400 Mhz cpu to perform well
while a 3Dfx only needs something less than 200 MHz. The above is an
interesting way to turn turn a deficiency into marketting gobbily-gook.
I'm sure all those 3Dfx owners running 233 and 266 Mhz Pentiums will
be surprised to hear their 3Dfx cards don't support their CPUs, I wonder
if their disk drive cards and modems also don't support CPUs over 200 Mhz?
I wonder how to design a card or chip that doesn't support a CPU over 200 Mhz?
Did you ever wonder how silly this sounds?
Where is Andy Rooney when you need him?
Rob
In article <01bc6fa1$e7defe20$0100...@host.telepac.pt>, "Stephen says...
>
>
>
>*********POWERVR 2*************
>
>WRONG!!
>
>The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has all
>the features that where missing in the original PowerVR and it`s now widely
>supported by manufacturers and it`s got some cool features :
>
He he, my turn. Apox.3Dx is faster than Apoc.3d BUT it is still much slower than
3dfx (which has been available for over 6 months now). The only added feature I
know about is bi-linear filtering. Are there others?
>- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
>speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
>so when the faster Pentiums are released the PowerVR will become even more
>powerful!
>
AND POWERVR WILL RULE THE EARTH CRUSHING PUNY 3DFX OWNERS BENEATH ITS MIGHTY
FISTS! Oh pu-leeeeze.
3dfx scales beautifully - framerates DOUBLED going from my old P90 to my new
P200MX. 3dfx benefits just as much running on P.Pro's and P2 - go read some
benchmarks.
>-The PowerVR 2 comes with these 4 games:
>
> - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
>fps, bet the 3Dfx can't beat that!)
>
Edge magazine failed to give this 'demo' a rating due to the complete lack of
longevity. It doesnt run in 16.7M colours. Kalisto are producing a 'Full' game
and there will be a 3dfx version.
<blaa blaa blaa>
>Actually the PowerVR is a little bit slower in the Direct3D tests (about
>1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
>PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
>FRIEND`s 3Dfx!)
>
How about these two real-world benchmark results
MTM :
POwerVR2 15 fps
3dfx 24 fps
MTM a bit too 'old' for you ? PowerVR2 runs better with the next generation of
Direct 3D titles ?? What about...
(from next-generation tests)
Terracide
POwerVR2 10-40 fps
3dfx 40-60 fps
>
>Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
>
As the song says 'Don't believe the Hype!'
Rob.
Jonathan Pearce
TormatoJES wrote:
>
> I have a Lightspeed 128 (ET6000) and am going to buy a 3D card. Should I
>
> Thanks,
>
> Jym
> Torma...@aol.com
>
> E-mail replies are always welcome
PowerVR2 (Apocaypose 3Dx) review at Games Mansion;
http://www.gamesmansion.com/reviews/3dx/3dx.html
They were very impressed.
The card came with 5 games including POD!
Jonathan Pearce
IJdo Dijkstra
neb...@capecod.net (Rex Nebula) wrote:
>Damn! Tell that to my PII 266 which is chugging right along with my
>Flash 3D. It must be using "cartoon logic"
Nope, that's jerk-o-vision, copyrighted by Microsoft ;)
IJdo Dijkstra
aur...@orion.xs4all.nl
IJdo Dijkstra
"Stephen Fordham" <sbfo...@mail.telepac.pt> wrote:
>The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has all
>the features that where missing in the original PowerVR and it`s now widely
>supported by manufacturers and it`s got some cool features :
So widely that except on this newsgroup I haven't heard of it...
(ok, that's overstating things a bit, but it's the idea that counts ;)
)
>- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
>speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
>so when the faster Pentiums are released the PowerVR will become even more
>powerful!
Bullshit. As all current and foreseeable CPU's will still use the
normal 60 - 100 Mhz bus speed, the real speed of the CPU will not
matter to any card.
> - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
>fps, bet the 3Dfx can't beat that!)
Seeing is believing, thank you very much.
> Actually the PowerVR is a little bit slower in the Direct3D tests (about
> 1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
> PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
> FRIEND`s 3Dfx!)
IF the PVR2 is that much better, how come they only manage to squeeze
out a measly 4 more frames?
IJdo Dijkstra
aur...@orion.xs4all.nl
IJdo Dijkstra
Jeff Atwood
Rex Nebula wrote in article <33946034...@snews.zippo.com>...
>On 2 Jun 1997 23:12:54 GMT, "Stephen Fordham"
><sbfo...@mail.telepac.pt> wrote:
>>Actually the PowerVR is a little bit slower in the Direct3D tests (about
>>1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
>>PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH
A
>>FRIEND`s 3Dfx!)
>>
>OH YEAH, WELL IF I SCREAM I CAN BE RIGHT TOO!
>
Heheheheheh. And we all know NGs need MORE SCREAMING! THERE SIMPLY ISN'T
ENOUGH INVECTIVE!
P.S. 3dfx is by far the best (and safest-- guaranteed support due to
critical mass) choice out there. I can't wait to see some of the newer
chipsets from 3dfx and others, though.
Jeff
IJdo Dijkstra
Ty <ty...@aloha.com> wrote:
>> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
>
>Now this would be nice. What kind of CPU do you need for this?
Rumour has it that they nicked Intel's experimental water-cooled 300
Mhz PII for this one ;)
IJdo Dijkstra
aur...@orion.xs4all.nl
}{ardball
>WRONG!!
You're right, you are.
>The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has all
>the features that where missing in the original PowerVR and it`s now widely
>supported by manufacturers and it`s got some cool features :
Really? Care to quote any reliable sources to some benchmarks? Every
benchmark I've seen puts it as slower than 3dfx. "it's now widely
supported by manufacturers"? hahahaha :) Why don't you venture out
beyond Videologic's web pages to garner more information when
comparing these cards.
>- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
>speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to 200MHZ,
>so when the faster Pentiums are released the PowerVR will become even more
>powerful!
PowerVR1 for the future? Ooooookay. Whatever, kid. And where'd you
come up with the idea that "3dfx only supports CPU speeds up to
200Mhz" ? What kind of nonsense is this? Besides, by the time 400Mhz
cpu's become affordable to the masses, the Banshee will be out and
will beat the PowerPR like a stepchild.
> - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at 35+
>fps, bet the 3Dfx can't beat that!)
>
> - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
>runs at 800x600 but in 65,000 colors at 35+ fps)
>
> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
>
> - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
Lessee here. I only see 1 game out of those 4 that doesn't work on a
3dfx as well. So we're supposed to buy a PowerPR card because it has
1 game 3dfx doesn't, while at the same time having much poorer
developer support than 3dfx? Sheah, as if.
>Actually the PowerVR is a little bit slower in the Direct3D tests (about
>1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
>PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
>FRIEND`s 3Dfx!)
Now wait a sec, in your first paragraph you said the VR was, and I
quote, "MUCH faster than the 3Dfx". But now you say it's "a little
bit slower in the Direct3D tests".
I've only seen 1 report that put the VR above 3dfx, and that was the
benchmark fiasco that they recently pulled with JPA, something JPA
have since invalidated.
Sunny Leung
Jonathan Pearce <jee...@interalpha.co.uk> wrote:
>PowerVR2 (Apocaypose 3Dx) review at Games Mansion;
>
>http://www.gamesmansion.com/reviews/3dx/3dx.html
>
>They were very impressed.
>
>The card came with 5 games including POD!
How can a reviewer be trusted when he doesn't even know how to disable
vsync on the Voodoo boards in order to get a proper framerate in
Tunnel?
Regards
--
* Sunny Leung, Sydney Australia. * Discworld Computing Society *
* mailto:su...@fl.net.au * Member - Anthill Inside? *
* http://www.fl.net.au/~sunny * No, Beehive Instead! *
DMan
On Wed, 04 Jun 1997 06:51:54 GMT, su...@fl.net.au (Sunny Leung) blah
blahed:
>Jonathan Pearce <jee...@interalpha.co.uk> wrote:
>
>>They were very impressed.
>>
>>The card came with 5 games including POD!
>
>How can a reviewer be trusted when he doesn't even know how to disable
>vsync on the Voodoo boards in order to get a proper framerate in
>Tunnel?
>
Best bet is to send a polite email to the computer channel
mentioning that and pointing out some recent PR stunts.
Apparently the email address is thecomput...@dial.pipex.com
--
Mark.
DMAN's 3D (website dedicated to 3DFX's Voodoo)
UK : http://www.markbz.demon.co.uk (updated more often)
USA: http://www.magicnet.net/~billk/index.html (mirror)
babyboy
On Wed, 04 Jun 1997 06:51:54 GMT, su...@fl.net.au (Sunny Leung) wrote:
>Jonathan Pearce <jee...@interalpha.co.uk> wrote:
>
>>PowerVR2 (Apocaypose 3Dx) review at Games Mansion;
>>
>>http://www.gamesmansion.com/reviews/3dx/3dx.html
>>
>>They were very impressed.
>>
>>The card came with 5 games including POD!
>
>How can a reviewer be trusted when he doesn't even know how to disable
>vsync on the Voodoo boards in order to get a proper framerate in
>Tunnel?
>
>Regards
>--
Here's a quote from the gamesmansion editor:
Hi Guys,
Firstly we'd like to thank you all for the heartening response to the
review of the 3Dx card. Many of your replies were constructive and
where possible we've replied individually. I'd like to start by saying
that the reviewer of the 3Dx has been taken outside and given a good
kicking by a randomly selected group of football supporters. But here
are some more responses to the feedback:
1. We got someone else to test the 3Dfx with Vsync turned off and it
did better in many of the D3D tests than the Power VR2 card. It was,
by no means however, a clear winner - the revised test results are up
at The Games Mansion (http://www.gamesmansion.com) now.
2. The ZD Winbench 3D benchmark offers patchy results at best - if
it's so good then why don't Ziff Davis's own journalists use it?
3. We're more than happy to do other tests on both cards and we're in
touch with Gary Tavoli (Chief Scientist at 3Dfx) and Colin (Chief
Techy Chap) at Videologic to try and find a test bed that both sides
are happy with.
4. We really don't care which card comes out in front - we're gamers
just like you and we want the best graphics possible as well. The
simple fact is that there are some stand-out titles available for 3Dfx
(GL Quake, the Tomb Raider patch, Pod and Moto Racer to name but four)
and for the 3Dx.
5. Does this really have to descend into a Mac Vs PC, Oasis Vs Blur,
My Willy's Bigger Than Your Willy debate?
6. We appreciate the responses from those of you who took the time to
intelligently argue your case.
7. We do listen.
Cheers,
Andy Hutchinson
(Editor: Games Mansion)
Jonathan Pearce
Rex Nebula wrote:
>
> On 2 Jun 1997 23:12:54 GMT, "Stephen Fordham"
> <sbfo...@mail.telepac.pt> wrote:
> > - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that also
> >runs at 800x600 but in 65,000 colors at 35+ fps)
> >
No - you are thinking of Resident Evil II
Resident Eveil is a pack-in with the PowerVR2
> > - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
> >
> Which isn't even out yet, so you can't have any idea how fast it is.
> I seriously doubt it uses 24bit color too, the speed hit on any card
> is too much.
MechWarrior 2 has been out for some time - it was part of the pack-in
with the PowerVR1.
> >Actually the PowerVR is a little bit slower in the Direct3D tests (about
> >1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
> >PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH A
> >FRIEND`s 3Dfx!)
> >
> OH YEAH, WELL IF I SCREAM I CAN BE RIGHT TOO!
>
> >Your BEST bet is to but the PowerVR 2 ( go to www.videologic.com)!
> >
> Your best bet is not to listen to this spewed pile of shit, and get a
> 3Dfx card.
Your best bet is to watch the Web reviews - not just one or two, but a
many as you can find.
Ignore the 3Dfx users (They are the quiet, polite posters in this
thread!) AND the PowerVR advocates.
Most of the 'paper' magazines seem to have completely lost-it these days
as far as 3D reviews go!
Jonathan
Tim Smith
In article 33959d95...@news.starnetinc.com,[SNIP]
>are some more responses to the feedback:
>1. We got someone else to test the 3Dfx with Vsync turned off and it
>did better in many of the D3D tests than the Power VR2 card. It was,
>by no means however, a clear winner - the revised test results are
>up at The Games Mansion (http://www.gamesmansion.com) now.
Power VR to the CRT like the 3Dfx...better visual results. You don't
cheat on the 3Dfx to give it a chance to ccompete with a cheating
Power VR...you make 'em both behave properly (as the 3Dfx cards
do by default)
>2. The ZD Winbench 3D benchmark offers patchy results at best - if
>it's so good then why don't Ziff Davis's own journalists use it?
Any test using Direct 3D and Win95 is going to be greatly affected by
the particulars of the rev of drivers you use and will not give the
potential customer an accurate view of the HARDWARE's performance.
I somebody buys a card based upon driver speed, he's likely to be
dissapointed at some future date when:
(A) the company may be forced to quietly release newer, more compatible
and slower drivers after winning benchmarks with fast, non-compliant
drivers. (This is not an uncommon practice in the industry...ask the
test folks at PC Mag who've had to re-write benchmarks to overcome
this sort of cheating).
(B) Somebody comes out with a killer app that screams by going around
Direct 3D...Look at Tomb Raider and GLQuake...both stellar 3Dfx
apps that don't use DirectX. 3Dfx cards support several APIs.
[SNIP]
>4. We really don't care which card comes out in front - we're gamers
>just like you and we want the best graphics possible as well. [SNIP]
Your site sure makes this seem untrue...looks like you're on NEC's
payroll...particularly that comment about 3Dfx cards being unable to
run at higher rez than 640x480....untrue. Check the docs in the 3Dfx
dev kit (freely available from 3dfx website). Again...you kick a card based
on current apps/drivers and mislead your readers about the HARDWARE.
[SNIP]
>5. Does this really have to descend into a Mac Vs PC, Oasis Vs Blur,
>My Willy's Bigger Than Your Willy debate?
It's irritating to see a good company like 3Dfx release a fantastic piece
of hardware and be the only ones with enough confidence in it to
release a free software dev kit on their website so that ANYONE can
develop for their card, only to see a company that DOESN'T have such
open policies push their hardware with the assistance of ill-informed
self-appointed journalists who don't accurately relate the facts.
Try again...on a level playing field...to measure hardware performance,
and THEN if the 3Dfx loses to a Commodore 64 that'll be fine with me!
-Tim
Tim Smith
In article 339c736d...@news.xs4all.nl,
aur...@orion.xs4all.nl (IJdo Dijkstra) said:
>Rumour has it that they nicked Intel's experimental water-cooled 300
>Mhz PII for this one ;)
> IJdo Dijkstra
>aur...@orion.xs4all.nl
He gets it in his dreams...Do you think he's running his monitor at
a 40Hz refresh rate (imagine the flicker at THAT rate <grimace>)
or perhaps he's running his monitor at 80Hz? Or is he not syncing
the video card page flips to Vsync...in which case his frame rates
are silly nonsense, since some fromes are not displayed in their
entirety (which is why THAT technique often results in screen
"tearing").
IMHO people who look for frame rates in excess on the monitor
sync rates are playing goofy mand games. The REAL goal of
everyone should be to get hardware that displays the most-
detailed, most complex, highest rez, highest quality images at
a standard video refresh rate...THAT would be a good goal
(of course the benchmarks would be a bit subjective).
-Tim
Simon Fenney
In article <5n5sln$i25$1...@paleale.wanet.net>,
Tim Smith <tims...@tlsmith.wanet.com> wrote:
>
> In article 33959d95...@news.starnetinc.com,
> mymo...@womb.com (babyboy) said:
> >Here's a quote from the gamesmansion editor:
> [SNIP]
> >are some more responses to the feedback:
> >1. We got someone else to test the 3Dfx with Vsync turned off and it
> >did better in many of the D3D tests than the Power VR2 card. It was,
> >by no means however, a clear winner - the revised test results are
> >up at The Games Mansion (http://www.gamesmansion.com) now.
> Sounds like a poor fix for the real problem...you needed to sync the
> Power VR to the CRT like the 3Dfx...better visual results. You don't
> cheat on the 3Dfx to give it a chance to ccompete with a cheating
> Power VR...you make 'em both behave properly (as the 3Dfx cards
> do by default)
I think you've got it the wrong way round. PowerVR always waits for the
VSync, but my understanding was that the 3Dfx supporters were saying
that the tester wasn't disabling it on the 3Dfx to get higher performance.
Simon Fenney
Videologic
-------------------==== Posted via Deja News ====-----------------------
http://www.dejanews.com/ Search, Read, Post to Usenet
Rob
In article <8656032...@dejanews.com>, Simon says...
>
>In article <5n5sln$i25$1...@paleale.wanet.net>,
> Tim Smith <tims...@tlsmith.wanet.com> wrote:
>>
>> In article 33959d95...@news.starnetinc.com,
>> mymo...@womb.com (babyboy) said:
>> >Here's a quote from the gamesmansion editor:
>> [SNIP]
>
>VSync, but my understanding was that the 3Dfx supporters were saying
>that the tester wasn't disabling it on the 3Dfx to get higher performance.
>
>Simon Fenney
>Videologic
>
So how come the framerates in the d3d tests where higher than refresh ?
Jonathan Pearce
Simon Fenney wrote:
> > Sounds like a poor fix for the real problem...you needed to sync the
> > Power VR to the CRT like the 3Dfx...better visual results. You don't
> > cheat on the 3Dfx to give it a chance to ccompete with a cheating
> > Power VR...you make 'em both behave properly (as the 3Dfx cards
> > do by default)
>
> I think you've got it the wrong way round. PowerVR always waits for the
> VSync, but my understanding was that the 3Dfx supporters were saying
> that the tester wasn't disabling it on the 3Dfx to get higher performance.
>
> Simon Fenney
> Videologic
So lets get this clear.
1, The 3Dfx (Voodoo) waits for VSYNC by default.
2, The 3Dx (Apocalypse) also waits for VSYNC by default.
For the 3Dfx to be compared to the 3Dx you
must disable the wait for VSYNC on the 3Dfx,
but NOT for the 3Dx.
And this makes for a fair comparison???????
One obvious question is:
Can the 3Dx be made ignore VSYNC also?
Jonathan
Dave Gibbons
On Fri, 06 Jun 1997 22:09:26 +0100, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
>Simon Fenney wrote:
>
>> I think you've got it the wrong way round. PowerVR always waits for the
>> VSync, but my understanding was that the 3Dfx supporters were saying
>> that the tester wasn't disabling it on the 3Dfx to get higher performance.
>>
>> Simon Fenney
>> Videologic
>
>So lets get this clear.
>
>1, The 3Dfx (Voodoo) waits for VSYNC by default.
>2, The 3Dx (Apocalypse) also waits for VSYNC by default.
>
>For the 3Dfx to be compared to the 3Dx you
>must disable the wait for VSYNC on the 3Dfx,
>but NOT for the 3Dx.
>
>And this makes for a fair comparison???????
>
>
>One obvious question is:
>
>Can the 3Dx be made ignore VSYNC also?
>
>Jonathan
This has to be the biggest load of crap I have ever seen.
I have NEVER seen any of the D3D tests (run on a variety of cards with
a variety of sync rates) exceed the frame rate of the monitor without
turning VSYNC waiting off. Period.
Gary Tarolli
Simon Fenney wrote:
>
> I think you've got it the wrong way round. PowerVR always waits for the
> VSync, but my understanding was that the 3Dfx supporters were saying
> that the tester wasn't disabling it on the 3Dfx to get higher performance.
If it's always waiting for VSync, then how can it produce 100+ frames per
second?
}{ardball
>Your best bet is to watch the Web reviews - not just one or two, but a
>many as you can find.
Which (except Videologic's/NEC's pages, of course) all seem to give
the impression that 3dfx is by far the better choice between the two.
Outside of D3D (and the UK), PowerPR has little support at all.
Jim Husband
Possibly by using a refresh of 120? I didn't see where refresh rates
were mentioned in the article - and I don't know what refresh rates the PVR2
supports. Seems like a bit of a stretch though. A rate of 113 with VSync on
would mean 90% of the time the fps was 120 or greater and %10 it was somewhere
below 120.
If the PVR2 can get 113 fps with VSync on - then the z-Buffer issue
and support are the only barriers they have left. If their D3D performance is
that good - then they may actually succeed.
I'm still interested in whether a z-Buffer can be implemented in D3D
using the PVR/2 drivers. Simon?
Jim
* Email address is false - "Reply-to:" is correct
Scott Pedersen
Jonathan Pearce wrote:
> So lets get this clear.
>
> 1, The 3Dfx (Voodoo) waits for VSYNC by default.
> 2, The 3Dx (Apocalypse) also waits for VSYNC by default.
Supposidly. Obviously though, the 3dx figures _were not_ with the card
set to a 60hrz moniter (there is no 'debate' in it- they simply _CANT_
be.), while the 3dfx figures were. I dont know why the 3dx used wasnt
waiting for the vsync if its supposed to by default- perhaps this was an
accident on the users part, or maybe incorrect installation..<shrug>.
Im sure theirs dozens of explanations, but 'nice' and 'nasty' (ie.
cheating).
> For the 3Dfx to be compared to the 3Dx you
> must disable the wait for VSYNC on the 3Dfx,
> but NOT for the 3Dx.
That is not and never was the case. The initial 3dfx figures _were_
with the card set to the "60hrz" moniter refresh rate, while the 3dx
were not however.
> And this makes for a fair comparison???????
No, though if you substitute the "3dfx" and "3dx" (damn i wish they
could come up with their own name!), that _is_ the situation which
occured with the first set of tests.
> One obvious question is:
>
> Can the 3Dx be made ignore VSYNC also?
It is already. Unless you have some fantastical explanation for their
card getting rates like '110' with a moniter set to 60Hrz updates?
(their use of this moniter refresh setting is proven with the first 3dfx
tests) "Nope" you say, well there you go..
Scott Pedersen
N
jim...@computer.net (Jim Husband) scribed:
>> If it's always waiting for VSync, then how can it
>> produce 100+ frames per second?
> Possibly by using a refresh of 120? I didn't see where
> refresh rates were mentioned in the article - and I don't
If they did use an Apocalypse with a refresh of 120Hz with
vsync on, a frame rate of about 100 still wouldn't be
possible with Tunnel Test. It would have to be a divisible
rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
Anything above 60 fps but below 120 wouldn't be possible
without disabling vsync.
You can try this with your 3DFX. With all the performance
hungry features enabled - such as CTRL+A/D/L, F5, F6 -
at 85Hz refresh, you should see a Tunnel rate of 85/2 =
~42 fps. At a 75Hz refresh, you should see a rate of
75/2 = ~37 fps.
N
Jonathan Pearce <jee...@interalpha.co.uk> scribed:
> Can the 3Dx be made ignore VSYNC also?
The Gamesmansion review didn't have the Apocalypse wait
for vsync while they did with the 3DFX, giving an
unfair advantage to the Apocalypse. A later revision
was done with the 3DFX not waiting for vsync. They
failed to mention that the Apocalypse was also not
waiting for vsync from the very beginning. BTW, I
use Apocalypse because it can get a bit confusing using
3Dx, 3DFX, 3Dx, 3DFX, over and over again, which I
gather was NEC/Videologic's goal.
Jonathan Pearce
N wrote:
>
> jim...@computer.net (Jim Husband) scribed:
>
> >> If it's always waiting for VSync, then how can it
> >> produce 100+ frames per second?
>
> > Possibly by using a refresh of 120? I didn't see where
> > refresh rates were mentioned in the article - and I don't
>
> If they did use an Apocalypse with a refresh of 120Hz with
> vsync on, a frame rate of about 100 still wouldn't be
> possible with Tunnel Test. It would have to be a divisible
> rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
> Anything above 60 fps but below 120 wouldn't be possible
> without disabling vsync.
>
Thats not correct! If the refresh rate is 120Hz, that means 120 screen
refreshes per second. If a 3D card waits for VSYNC, but can only
generate 110 frames per second then the display will be updated 110
times out of the possible 120.
During the 10 remaining refreshes the card will not update the screen.
It simply does not have the data available in time, so it skips one
update!
(Those 10 'missed' updates are of course spread throughout the 1 second
period.)
I think the difference between the 3Dfx and the Apocalypse might be
explained in the position and function of the FRAME BUFFER!
The 3Dfx uses part of its 4MB for the TWO frame buffers - one displayed
one being updated.
The Apocaypse uses the frame buffers of the external 2D card. If the
Apocaypse can control which buffer is being displayed - ie stopping it
alternating the standard 1,2,1,2,1,2 then it could run the missed frames
as repeats 1,2,1,1,2,1,2.
You simply get two identical frames, which at 120Hz is impossible to
see.
So still 120Hz, but only 110 updates per second and still waiting for
VSYNC.
It could be that this functionality is not possible with the memory
access design of the 3Dfx.
Jonathan
Jonathan Pearce
Dave Gibbons wrote:
>
> On Fri, 06 Jun 1997 22:09:26 +0100, Jonathan Pearce
>
> >Simon Fenney wrote:
> >
>
> >> I think you've got it the wrong way round. PowerVR always waits for the
> >> VSync, but my understanding was that the 3Dfx supporters were saying
> >> that the tester wasn't disabling it on the 3Dfx to get higher performance.
> >>
> >> Videologic
> >
> >So lets get this clear.
> >
> >1, The 3Dfx (Voodoo) waits for VSYNC by default.
> >2, The 3Dx (Apocalypse) also waits for VSYNC by default.
> >
> >must disable the wait for VSYNC on the 3Dfx,
> >but NOT for the 3Dx.
> >
> >
> >
> >
> >Can the 3Dx be made ignore VSYNC also?
> >
>
> This has to be the biggest load of crap I have ever seen.
> I have NEVER seen any of the D3D tests (run on a variety of cards with
> a variety of sync rates) exceed the frame rate of the monitor without
> turning VSYNC waiting off. Period.
>
> Dave
> da...@flavious.prestel.co.uk
Very True! But what if they were using 120HZ screen refresh rate!
This would still explain the 3Dfx 60fps limit (according to other 3Dfx
owners)
Jonathan
Dave Gibbons
On Sat, 07 Jun 1997 13:50:21 +0100, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
>Dave Gibbons wrote:
>>
>> This has to be the biggest load of crap I have ever seen.
>> I have NEVER seen any of the D3D tests (run on a variety of cards with
>> a variety of sync rates) exceed the frame rate of the monitor without
>> turning VSYNC waiting off. Period.
>>
>> Dave
>> da...@flavious.prestel.co.uk
>
>Very True! But what if they were using 120HZ screen refresh rate!
>This would still explain the 3Dfx 60fps limit (according to other 3Dfx
>owners)
>
>Jonathan
So?
You were saying something about fair comparison I seem to remember?
You can have greater than 60hz refresh on the 3Dfx.
Jonathan Pearce
What I was implying is that the 3Dfx needs to turn VSYNC waiting off.
If this makes no difference to the screen appearance (which shows just
how quick the 3Dfx is!) then this is a fair comparison. But if you get
visible tearing (as in Outlaws! - from one 3Dfx owner) then it is not a
fair comparison.
The Apocaypse can I believe, update at any rate upto the refresh rate,
without disabling the VSYNC waiting.
The 3Dfx must work at the Frame rate or sub division there of! 120,60
etc (with VSYNC waiting ON)
A benchmark program should push the cards to the limit. You now have to
judge the results not only in actual numbers, but what is viewed on the
screen during the benchmark OR produce two sets of results for the VSYNC
ON/OFF for the 3Dfx.
The cards are very different in design!
Jonathan
Jim Husband
On Sat, 07 Jun 1997 13:38:04 +0100, Jonathan Pearce <jee...@interalpha.co.uk>
wrote:
>N wrote:
>>
>> If they did use an Apocalypse with a refresh of 120Hz with
>> vsync on, a frame rate of about 100 still wouldn't be
>> possible with Tunnel Test. It would have to be a divisible
>> rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
>> Anything above 60 fps but below 120 wouldn't be possible
>> without disabling vsync.
>>
>
>Thats not correct! If the refresh rate is 120Hz, that means 120 screen
>refreshes per second. If a 3D card waits for VSYNC, but can only
>generate 110 frames per second then the display will be updated 110
>times out of the possible 120.
>During the 10 remaining refreshes the card will not update the screen.
>
I agree that "non-intregal" fps rates are possible - since we don't
know the frequency with which the fps is updated.
However - if you're assuming a *constant* generated rate of 110 fps -
then the output will be 60 fps with VSync waiting turned on. Assuming the card
just missed a VSync and has to wait for the next scan - no processing can
occur during this time. So the "start" of each frame's generation always
occurs immediately following the VSync - and every generated frame will be
stuck waiting the same amount of time for the next VSync.
However - fps is almost never constant. During a particular 1/10
second - maybe the card was able to maintain 120 fps; and during the next 1/10
second it consistently missed the VSync and was only able to get 60 fps. The
average can result in non-intergral rates even though the "instantaneous" fps
has to be multiple of the refresh rate.
I was really amazed by this when Dave Nagy first explained it - I was
under the same impression as you that 110 fps in a 120 refresh mode would
simply give you 110 fps. He also explains it a bit better than I do - so if
it's unclear maybe he'll straighten it out.
It also turns out that a higher refresh rate does not always result in
a faster fps when waiting for VSync. Again - assuming a constant 75 fps (which
is probably the unrealistic part of this) then a monitor at 75hz results is a
perfect 75 fps. However a monitor at 85hz will result in only 42.5 fps and one
at 120hz will result in only 60 fps.
Doubling (or tripling if that's possible) a monitor's refresh rate can
never hurt VSync performance.
Dave Gibbons
On Sat, 07 Jun 1997 15:27:16 +0100, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
Dave Gibbons wrote:
What are you on about? It can update upto the refresh rate without
disabling the VYSNC? Whats that supposed to mean?
The monitor is refreshing at 120hz (120th of second per frame)
note here:- 120th of a second occurs sooner than 110th of a second
so:-
Assuming you are rendering to the invisible frame buffer.
The process goes something like this:-
SOFTWARE
DISPLAY/MONITOR
Start rendering: Start refresh:
continue rendering: raster traversing
continue rendering: Hit Vysnc:
NOTES: rendering misses VSYNC hit (now into 60HZ)
continue rendering: raster traversing
Wait for VSYNC raster traversing
NOTES: Must wait or tearing *MAY* be obtrusive
Screen swapped Hit Vsync
Rendering restarted raster traversing
NOTES: Two Vsyncs have occured since we
started first render. hence 60hz
Listen:- the Apoclypse is a different hardware
solution than the 3DFX - it has a strange way of
Zbuffering - and it pumps frame data over the PCI
bus to the 2D card. BUT, and its a BIG BUT it must
interact with the frame buffer in the same way - we
are talking about a monitor here you know that thing
you are looking at right now. It creates its picture
through rasterisation - no matter how different the
implementation the Apocalyspe is it can't get round
this basic fact.
So ok the Apocalypse renders 110 frames per second - if it has to wait
for VSYNC at 120hz then 110 frames is MEANINGLESS it will wait
for the next sync at 120hz. Triple buffering can improve performance
in this respect. (Oh shit I've given you another idea now!)
>A benchmark program should push the cards to the limit. You now have to
>judge the results not only in actual numbers, but what is viewed on the
>screen during the benchmark OR produce two sets of results for the VSYNC
>ON/OFF for the 3Dfx.
>The cards are very different in design!
>Jonathan
I'm sorry, but what are you whittering on about?
Clutching at straws or what?
Dave Gibbons
On Sat, 07 Jun 1997 14:42:55 GMT, jim...@computer.net (Jim Husband)
wrote:
> However - if you're assuming a *constant* generated rate of 110 fps -
>then the output will be 60 fps with VSync waiting turned on. Assuming the card
>just missed a VSync and has to wait for the next scan - no processing can
>occur during this time. So the "start" of each frame's generation always
>occurs immediately following the VSync - and every generated frame will be
>stuck waiting the same amount of time for the next VSync.
>
>Jim
>* Email address is false - "Reply-to:" is correct
Exactly! Christ is it me, or is this guy clutching at straws!
Jim Husband
On 7 Jun 1997 06:01:34 GMT, n...@one.here (N) wrote:
>jim...@computer.net (Jim Husband) scribed:
>
>>> If it's always waiting for VSync, then how can it
>>> produce 100+ frames per second?
>
>> Possibly by using a refresh of 120? I didn't see where
>> refresh rates were mentioned in the article - and I don't
>
>vsync on, a frame rate of about 100 still wouldn't be
>possible with Tunnel Test. It would have to be a divisible
>rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
>Anything above 60 fps but below 120 wouldn't be possible
>without disabling vsync.
>
The "instantaneous" fps has to be a multiple - but the fps rating is
done over time (probably only updated once a second or so). Since the test is
not static (some areas have higher fps than others) - you can easily end up
with 114 fps with VSync on and a refresh of 120.
If 90% of the time the fps is 120 and 10% of the time it is 60 - then
you end up with ((90 * 120) + (10 * 60)) / 100 = 114.
That said - I don't believe the PVR2 is capable of this, so I think
the test was performed with VSync turned off. Have I mentioned how poorly I
thought the article was written? <grin> GamesMansion ought to consider hiring
one or more of the sharper people in this group to keep them from writing such
a confusing review.
By the time they correct their review (again) - either they'll be
embarrassed by the outcome (if 3dfx is still ahead of PVR2) - or (if the PVR2
actually outperformed the 3dfx) they'll no longer have a "scoop."
Dave Gibbons
On Sat, 07 Jun 1997 13:38:04 +0100, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
N wrote:
>
> jim...@computer.net (Jim Husband) scribed:
>
> >> If it's always waiting for VSync, then how can it
> >> produce 100+ frames per second?
>
> > Possibly by using a refresh of 120? I didn't see where
> > refresh rates were mentioned in the article - and I don't
>
> If they did use an Apocalypse with a refresh of 120Hz with
> vsync on, a frame rate of about 100 still wouldn't be
> possible with Tunnel Test. It would have to be a divisible
> rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
> Anything above 60 fps but below 120 wouldn't be possible
> without disabling vsync.
>
>Thats not correct! If the refresh rate is 120Hz, that means 120 screen
>refreshes per second. If a 3D card waits for VSYNC, but can only
>generate 110 frames per second then the display will be updated 110
>times out of the possible 120.
>During the 10 remaining refreshes the card will not update the screen.
>It simply does not have the data available in time, so it skips one
>update!
>(Those 10 'missed' updates are of course spread throughout the 1 second
>period.)
You are much mistaken, there are not '10 refreshes left over'.
It is infact in defecit. By the time the card has finished rendering
the scene 120hz has been and gone and we are in defecit. It will then
have to wait for the next refresh - hence 60hz (see my previous
explanation with the order of events)
Simple explantion: 120hz (120 fps) is quicker than 110hz.
Dave
da...@flavious.prestel.co.uk
Tim Smith
In article 8656032...@dejanews.com,
Simon Fenney <sim...@videologic.com> said:
[SNIP]
>I think you've got it the wrong way round. PowerVR always waits for the
>VSync, but my understanding was that the 3Dfx supporters were saying
>that the tester wasn't disabling it on the 3Dfx to get higher
>performance.
As I replied to you via E-mail...It's a good thing if your product waits for
VSYNC....now let's see a free, fully-functional SDK on your web site...
3Dfx did it, with no ill effects...the only reason for other companies to
not follow suit is if they have something to hide.
BTW: What monitor sync rates are you guys using? If the frame rates
reported are not an even multiple of the VSYNC rate, then some frames
are not being displayed completely before the follow-on frames, or some
frames are being repeated, or the card is not syncing its page flips to the
monitor VSYNC. I'd be interested to see a side-by-side comparison
of the two cards rendering the exact same scene in the same "mode"
(i.e. resolution, color depth, video sync rate, and rendering features).
Don't get me wrong here...my primary interest is in open hardware systems
with maximum performance at minimum price. Right now, that appears to
be 3Dfx. If somebody else ever releases THEIR SDK while matching or
exceeding the performance of the 3Dfx chipset, I'll cheer them on too.
-Tim
N
Jonathan Pearce <jee...@interalpha.co.uk> scribed:
> Very True! But what if they were using 120HZ screen refresh rate!
> This would still explain the 3Dfx 60fps limit (according to other
> 3Dfx owners)
Wrong again - follow the thread more carefully. One of their
direct3D test produced a score of 150 fps on the 3DFX. If
in fact the screen refresh was set at 120Hz with vsync on, the
rate would be 120 fps, not 60. They clearly had the 3DFX
refresh rate set at 60Hz.
N
jim...@computer.net (Jim Husband) scribed:
> The "instantaneous" fps has to be a multiple - but the fps
> rating is done over time (probably only updated once a
> second or so). Since the test is not static (some areas
> have higher fps than others) - you can easily end up
> with 114 fps with VSync on and a refresh of 120.
Here's what I've noticed running it on my system: With
vsync off and all options enabled using the keys
CTRL+A/D/L, F5/F6, my P166/83 and R3D at 61MHz maxes out
at 67 fps in Tunnel Test. Now assuming you're correct,
I should see 67 fps or thereabouts with vsync on at a
refresh rate of 75Hz. This isn't the case. I see a
steady 37 fps or 75/2. At a refresh rate of 85 Hz, Tunnel
produces 42 fps or 85/2.
> Have I mentioned how poorly I thought the article was
> written? <grin>
Doesn't everyone? <g>
N
Jonathan Pearce <jee...@interalpha.co.uk> scribed:
> The Apocaypse can I believe, update at any rate upto the
> refresh rate, without disabling the VSYNC waiting.
Uhh no, you cannot believe. Unless the Apocalypse is
using triple buffering - which it isn't - it CANNOT
update at any rate with vsync on.
> The 3Dfx must work at the Frame rate or sub division
> there of! 120,60 etc (with VSYNC waiting ON)
ALL double buffered cards, including the Apocalypse,
work under this restriction. Can you explain how
the Apocalypse violates this rule without resorting
to triple buffering?
> The cards are very different in design!
You need to read up a bit more.
N
Jonathan Pearce <jee...@interalpha.co.uk> scribed:
> Thats not correct! If the refresh rate is 120Hz, that means
> 120 screen refreshes per second. If a 3D card waits for
> VSYNC, but can only generate 110 frames per second then the
> display will be updated 110 times out of the possible 120.
> During the 10 remaining refreshes the card will not update
> the screen.
Wrong yet again. Sheesh, you're on a roll here.
1) Display is updated 120 times a second.
2) Let's say a 3D scene takes just over a frame to render
and display. Now waiting for vsync would mean that despite
the scene taking just over a frame to render, it would be
forced to wait for two full screen refreshes to render.
This would cut the effective frame rate to 120/2 or 60Hz.
> It simply does not have the data available in time, so
> it skips one update!
And one update after that, and one update after that,
cutting the effective rate to 120/2.
> (Those 10 'missed' updates are of course spread throughout
> the 1 second period.)
Huh? We're not talking triple-buffering here. You
can't just spread the 10 missed updates evenly over
the 1 second period.
> The Apocaypse uses the frame buffers of the external 2D
> card. If the Apocaypse can control which buffer is
> being displayed - ie stopping it alternating the
> standard 1,2,1,2,1,2 then it could run the missed frames
> as repeats 1,2,1,1,2,1,2.
The Apocalypse's use of the 2D card's frame buffer is
NO DIFFERENT than any other 2D or 3D board with double
buffering and vsync on.
> You simply get two identical frames, which at 120Hz is
> impossible to see.
No, the following frames are rendered just following
a refresh, resulting once again in 120/2.
> So still 120Hz, but only 110 updates per second and
> still waiting for VSYNC.
Look, I have a feeling all this theory won't convince
you at all. So here's a really simple suggestion:
try it YOURSELF. Turn vsync off, execute Tunnel Test,
and use some of the performance robbing options so
you see about 100 - 110 fps. Now enable waiting for
vsync. You'll see 60 fps at a refresh rate of 120Hz.
Try it.
> It could be that this functionality is not possible
> with the memory access design of the 3Dfx.
It could be your memory access has gone bad. The 3DFX
functions no differently than the Apocalypse in this
respect.
M. R. Embury
Here's a suggestion:
How about changing the subject of this thread to: 3Dfx vs 3Dx refresh
rates IRT Vsync
This discussion is getting far too technical and geek specific for the
average 3D gamer that is trying to make a decision between these two
products. The thread does not resemble: 3Dfx or Power VR? Why? unless
your a VSNYC guru.
Thanks a bunch!
TormatoJES
I couldn't agree more! As the originator of this thread, I was dismayed
to see the direction it was going. Please keep it on a level for those of
us who aren't at you level yet.
Thanks,
Scott Pedersen
Scott Pedersen wrote:
> Exactly. If the maximum is within the level the refresh rate is set at
> then 'vsync' hardly makes a difference. 120 when the maximum is '~110'
> is fine (though it would still be more sensible to remove refresh sinc),
> '60' when the maximum update rate is '>110-130' definately is _not_.
Actually i take that '120 when the maximum is '~110' is fine' bit back.
Ive just read over a couple of posts explaining the intervals regarding
refresh rates, and this is obvoiusly not the case. The simple fact is
the "3dx" was not waiting for vsync in the tests while the 3dfx was.
_PERIOD_. End of discussion. Give up Jonathon, however you
add/divide/crumple them, the numbers simply do not add up to any other
resolution. And please dont say 'the 3dx obviously doesnt need the to
tie to vsync' or some such nonsence, because this to is nonsence.
(proven by Simon of Videologic already stating in this newsgroup that
the 3dx is supposed to wait for vsync - (even though it wasnt in those
tests))
Scott Pedersen
Scott Pedersen
Dave Gibbons wrote:
> So?
>
> You were saying something about fair comparison I seem to remember?
>
> You can have greater than 60hz refresh on the 3Dfx.
Exactly. If the maximum is within the level the refresh rate is set at
then 'vsync' hardly makes a difference. 120 when the maximum is '~110'
is fine (though it would still be more sensible to remove refresh sinc),
'60' when the maximum update rate is '>110-130' definately is _not_.
Scott Pedersen
Dave Gibbons
>Thanks,
>Jym
Torma...@aol.com
Sorry guys, the thread turned this way because the PVR mob,
all 3 of them including one Rep from Videologic are pissing in the
wind. Desperately trying to come up with arguments as to why a set
of very dubious test figures show PVR2 'the only 3D card worth
bothering about' . The simple fact is that these tests and the review
are completely worthless and totally flawed.
PVR2 may be OK hardware, but at present it just ain't in the same
ballpark as 3Dfx as regards support -it certainly seems to be the
developers favourite. Lets say PVR2 is faster than 3DFX (which I doubt
- there are is a mags that concluded that it isn't), it ain't gonna
help if you don't have any software.
PVR has a extremely bad rep here, and it don't help them by
compounding it with some of the very dubious comments.
PVR2 may take off, and that's fair enough. 3Dfx has already. At
present you can be safe in the knowledge that you have the best
support, the fastest native hardware ports and the fastest D3D
software. PVR2 still has to come up with the titles to allow a fair
comparison with native ports and D3D looks like being a 3Dfx winner.
BTW don't trust any benchmarks involving Tunnel test & D3D test
as they are more or less worthless given PVR (actually known as PPR
round here for obvious reasons) record in cheating in these tests.
Even without cheating the worth of these tests are very questionable
when there are far better tests out there.
PVR and PVR2 are sold on the basis of promises.
They have a good way to go to match to user base and support of
the 3Dfx boards.
Its your money and your choice.
Dave
da...@flavious.prestel.co.uk
Jonathan Pearce
Jim Husband wrote:
>
> The "instantaneous" fps has to be a multiple - but the fps rating is
> done over time (probably only updated once a second or so). Since the test is
> not static (some areas have higher fps than others) - you can easily end up
> with 114 fps with VSync on and a refresh of 120.
>
> If 90% of the time the fps is 120 and 10% of the time it is 60 - then
> you end up with ((90 * 120) + (10 * 60)) / 100 = 114.
>
> That said - I don't believe the PVR2 is capable of this, so I think
> the test was performed with VSync turned off. Have I mentioned how poorly I
> thought the article was written? <grin> GamesMansion ought to consider hiring
> one or more of the sharper people in this group to keep them from writing such
> a confusing review.
>
> By the time they correct their review (again) - either they'll be
> embarrassed by the outcome (if 3dfx is still ahead of PVR2) - or (if the PVR2
> actually outperformed the 3dfx) they'll no longer have a "scoop."
>
> Jim
>
> * Email address is false - "Reply-to:" is correct
OK I think we have done this one to Death!
I agree with your explanation. The remaining question is whether the
Apocalypse does or does not wait for VSYNC.
I think our discussions have show just how difficult it is to review
these cards!
I stand corrected on some parts!
Thanks to eveyone who posted.
Jonathan.
Jonathan Pearce
Well I have designed high speed asynchronous-synchronous hardware.
If you are right about the way these cards work then the design
is very poor.
After all, what is the point in being able to generate 110 fps
if you can only 'clearly' display 60fps.
The fact that the 3Dfx can handle not waiting for VSYNC without
visible update-effects is a credit to it.
Jonathan.
Filiep Geeraert
Jim Husband wrote:
>
> Have I mentioned how poorly I
> thought the article was written? <grin> GamesMansion ought to consider hiring
> one or more of the sharper people in this group to keep them from writing such
> a confusing review.
I think the review sounded very biased, and read almost like one of
those PR articles Andy Griffith wrote in this newsgroup.
To me it looked like someone else told them what to write and they
changed the words a bit to make it look like it would be their own
experience.
Jonathan Pearce
N wrote:
>
> jim...@computer.net (Jim Husband) scribed:
>
> >> If it's always waiting for VSync, then how can it
> >> produce 100+ frames per second?
>
> > Possibly by using a refresh of 120? I didn't see where
> > refresh rates were mentioned in the article - and I don't
>
> If they did use an Apocalypse with a refresh of 120Hz with
> vsync on, a frame rate of about 100 still wouldn't be
> possible with Tunnel Test. It would have to be a divisible
> rate - e.g. either 120 fps, 60 fps, 40 fps, 30 fps, etc.
> Anything above 60 fps but below 120 wouldn't be possible
> without disabling vsync.
>
> hungry features enabled - such as CTRL+A/D/L, F5, F6 -
> at 85Hz refresh, you should see a Tunnel rate of 85/2 =
> ~42 fps. At a 75Hz refresh, you should see a rate of
> 75/2 = ~37 fps.
Well I have to agree with you - finally!
In the name of fairness I tried using the Tunnel test at different
refresh rate on my PowerVR1 card (Not the reviewed PowerVR2 card);
60Hz - 121fps
75Hz - 121fps
100Hz - 121fps
I dont think this card can be waiting for VSYNC!
Jonathan
Dave Gibbons
On Sun, 08 Jun 1997 12:37:15 +0100, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
Jim Husband wrote:
>
<snip>
>>
>> By the time they correct their review (again) - either they'll be
>> embarrassed by the outcome (if 3dfx is still ahead of PVR2) - or (if the PVR2
>> actually outperformed the 3dfx) they'll no longer have a "scoop."
>>
>> Jim
>>
>> * Email address is false - "Reply-to:" is correct
>OK I think we have done this one to Death!
>
>I agree with your explanation. The remaining question is whether the
>Apocalypse does or does not wait for VSYNC.
>
>I think our discussions have show just how difficult it is to review
>these cards!
>
>I stand corrected on some parts!
>
>Thanks to eveyone who posted.
>
>Jonathan.
OK Jonathon. But I don't agree there is any question. If the PVR(2)
is to avoid tearing it *must* wait for vsync. The question I think you
mean is - did they test without vsync in the games mansion review?
The answer to this if their figures are cosher is almost certainly no,
they weren't waiting for vsync on the PVR2.
No hard feelings.
Stephen Fordham
}{ardball <hard...@alt.net.REMOVETHIS> wrote in article
<3396d20a...@news.alt.net>...
> On 2 Jun 1997 23:12:54 GMT, "Stephen Fordham"
> <sbfo...@mail.telepac.pt> wrote:
>
> >WRONG!!
>
> You're right, you are.
>
> >The NEW PowerVR 2 (Apocalypse 3Dx) is MUCH faster than the 3Dfx and has
all
> >the features that where missing in the original PowerVR and it`s now
widely
> >supported by manufacturers and it`s got some cool features :
>
> Really? Care to quote any reliable sources to some benchmarks? Every
> benchmark I've seen puts it as slower than 3dfx. "it's now widely
> supported by manufacturers"? hahahaha :) Why don't you venture out
> beyond Videologic's web pages to garner more information when
> comparing these cards.
>
> >- The PowerVR 2 is for the future (so is the PowerVR 1), it supports CPU
> >speeds up to 400MHZ while the 3Dfx only supports CPU speeds up to
200MHZ,
> >so when the faster Pentiums are released the PowerVR will become even
more
> >powerful!
>
> PowerVR1 for the future? Ooooookay. Whatever, kid. And where'd you
> come up with the idea that "3dfx only supports CPU speeds up to
> 200Mhz" ? What kind of nonsense is this? Besides, by the time 400Mhz
> cpu's become affordable to the masses, the Banshee will be out and
> will beat the PowerPR like a stepchild.
>
> > - ULTIM@TE RACE (Racing game that runs at 800x600 in 16.7M colors at
35+
> >fps, bet the 3Dfx can't beat that!)
> >
> > - RESIDENT EVIL (MDK/TOMB RAIDER/ALONE IN THE DARK style game that
also
> >runs at 800x600 but in 65,000 colors at 35+ fps)
> >
> > - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
> >
> > - MECHWARRIOR 2 (Runs at 640x480 in 16.7M colors at 40+ fps).
>
> Lessee here. I only see 1 game out of those 4 that doesn't work on a
> 3dfx as well. So we're supposed to buy a PowerPR card because it has
> 1 game 3dfx doesn't, while at the same time having much poorer
> developer support than 3dfx? Sheah, as if.
>
> >Actually the PowerVR is a little bit slower in the Direct3D tests (about
> >1.5fps slower), BUT ONLY ON A P133 and SLOWER. On a FASTER system the
> >PowerVR OUTPERFORMS the 3Dfx by 4fps (I KNOW BECAUSE I`VE TRIED IT WITH
A
> >FRIEND`s 3Dfx!)
>
> Now wait a sec, in your first paragraph you said the VR was, and I
> quote, "MUCH faster than the 3Dfx". But now you say it's "a little
> bit slower in the Direct3D tests".
>
> I've only seen 1 report that put the VR above 3dfx, and that was the
> benchmark fiasco that they recently pulled with JPA, something JPA
> have since invalidated.
>
READ THIS AND WEEP !!!!!!!!!!
HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
Dave Gibbons
On 8 Jun 1997 19:51:08 GMT, "Stephen Fordham"
<sbfo...@mail.telepac.pt> wrote:
Are you an asshole or what - that review has zero credibility.
What the hell do you think all the posting here has been talking
about.
BTW shouting makes you look twice as stupid.
N
In article <01bc71de$9c285aa0$0100...@host.telepac.pt>, "Stephen Fordham"
<sbfo...@mail.telepac.pt> wrote:
> READ THIS AND WEEP !!!!!!!!!!
> HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
Got your caps lock stuck again, eh?
> http://www.gamesmansion.com/reviews/3dx/3dx.html
This is the same incompetent review that first had
the 3DFX waiting for vsync, but not the PowerVR2.
They then revised the 3DFX scores without waiting
for vsync, but didn't change their conclusion or
tone at all. This is also the same incompetent
people that can't get 3D Winbench working simply
because of a conflict with their Mystique, and who
argue that 3D Winbench isn't any good because none
of the Ziff Davis publications use it. Of course,
it doesn't matter to them that 3D Winbench was released
just over a month ago, and that the latest PC Magazine
actually makes use of it.
N
In article <339A99...@interalpha.co.uk>, Jonathan Pearce
<jee...@interalpha.co.uk> wrote:
> I think our discussions have show just how difficult it is to review
> these cards!
Which is why more benchmarks should follow 3D Winbench's lead.
It disables waiting for vsync for all cards. You can enable
vsync, but it doesn't produce a score with it on.
Jonathan Pearce
I agree - we need a better benchmark program.
Two problems;
1, I dont know how to disable/enable VYNC on the PowerVR1! (Nothing
obvious)
2, In the readme with 3D Winbench they say that you cannot quote PowerVR
results because they are not valid - too many of the tests fail.
Jonathan
Gary Tarolli
Jonathan Pearce wrote:
>
> I agree - we need a better benchmark program.
>
> Two problems;
> 1, I dont know how to disable/enable VYNC on the PowerVR1! (Nothing
> obvious)
> 2, In the readme with 3D Winbench they say that you cannot quote PowerVR
> results because they are not valid - too many of the tests fail.
That in itself says something, doesn't it?
Tim Smith
[SNIP]
>>This discussion is getting far too technical and geek specific for the
>>average 3D gamer that is trying to make a decision between these two
>I couldn't agree more! As the originator of this thread, I was dismayed
>to see the direction it was going. Please keep it on a level for those
>Jym
>Torma...@aol.com
O.K....here goes...(and, BTW, this is why I *hate* these types of performance
arguments):
1. Owners of various bits of hardware generally like to convince themselves
that they bought the right/best stuff...so they tend to jump into these debates
weather they know what they're saying or not.
2. raster graphics (like on a PC, as opposed to vector graphics like the ones in
the old arcade game "Asteroids") consist of a series of images flashed rapidly
before your eyes at a fixed rate. A TV, for instance uses a rate of 60Hz, or
60 times per second (unlrelated to your PC note: TV uses an interlaced sceme
that shows every other line on each update...so the effective rate is 30Hz.)
3. A 3D graphics card can render images at a give rate in frames per second (FPS)
depending upon scene complexity, screen resolution, color depth, etc.) This means
that the 3D card could theoretically move that many frames per second to screen
memory for display.
4. If the graphics card FPS does not exactly match the refresh, or VSYNC (vertical
sync signal) rate of the CRT...on of two things occurs:
(a) the card must ignore VSYNC and display the new page while a current page is
only partially displayed. (this causes image degredation of a type called "tearing").
(b) the card must wait for the next vsync before displaying the newly rendered image
and in doing so, it wastes time that could be used to render the next frame.
*note* triple buffering could get around some of this, but some cards cannot do it and
those that can, like the 3Dfx cards, trade off some memory-intensive features if you
enable triple buffering.
5. Because some people want to brag about superior frame rates, they tell their cards
to ignore VSYNC and simply crank out the max FPS it can. The number can become
meaningless since the user is not seeing some frames in their entirety...but that doesn't
matter if your goal in life is to brag about frame rate numbers.
6. This whole thing becomes very complex because most graphics cards have different
features...so you're not really even seeing the same images with 2 different cards...and
because computer monitors, unlike TVs, can change their screen Vsync,refresh rates...
so a benchmark addict can tweak his video refresh rate to achieve better numbers.
Hope that helps clear things up :-)
-Tim
jon
In article <339a93c...@news.prestel.co.uk>, da...@flavious.prestel.co.uk (Dave Gibbons) wrote:
>On 8 Jun 1997 02:19:25 GMT, torma...@aol.com (TormatoJES) wrote:
>
>
>>>Here's a suggestion:
>>>
>>>How about changing the subject of this thread to: 3Dfx vs 3Dx refresh
>>>rates IRT Vsync
>>>
>>>average 3D gamer that is trying to make a decision between these two
>>>your a VSNYC guru.
>>>
>>>Thanks a bunch!
>
>>to see the direction it was going. Please keep it on a level for those of
>>us who aren't at you level yet.
>
Carl Mueller
n...@one.here (N) wrote:
>In article <339A99...@interalpha.co.uk>, Jonathan Pearce
><jee...@interalpha.co.uk> wrote:
>
>> I think our discussions have show just how difficult it is to review
>> these cards!
>
>Which is why more benchmarks should follow 3D Winbench's lead.
>It disables waiting for vsync for all cards. You can enable
>vsync, but it doesn't produce a score with it on.
Actually, instead of disabling wait-for-vsync, the benchmarks
should provide enough polygons & pixels such that the wait
makes little difference. Benchmarks that are frame-rate
limitied are not very useful, given the trend towards more
and more complex scenery.
-Carl (mue...@cs.unc.edu)
Draekon
> READ THIS AND WEEP !!!!!!!!!!
> HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
>
What a bunch of crap! I've never read a more biased article in my life. The
PowerVR2 may beat the 3dfx in some areas, but for a 2nd generation card, it
doesn't do it by very much. Let's look at the facts.
-=[ 3dfx ]=- -=[ PowerVR2 ]=-
Support: (adv 3dfx)
Microsoft Direct3D, Winglide, Microsoft Direct3D, PowerSGL
Reality Lab 2.0, Intel 3DR, (Full OpenGL in future? or just GLQuake)
Criterion Renderware
(Full Open GL support in works)
Memory: (adv PowerVR2)
Voodoo-4Mb EDO (2 texture, 2 framebuffer) 4Mb SGRAM (texture memory)
Voodoo Rush- 6Mb EDO (2 texture,
4 frame buffer)
Features: (winner? I'd say 3dfx but I'll say draw since PowerVR2 is not
out. I can't verify everything and
I see different names for equivalent features)
Texture modulation, Bilinear Filtering,
Perspective-correct texture mapping, Perspective-correction,
Z-buffering (16-bit), (equivalent to Z-Buffering 32-bit),
Level-of-detail MIP mapping, Anti-aliased texture mapping,
Bi-linear and advanced texture filtering, Light volumes,
Texture compositing and morphing, Real time shadows,
Animated textures, True logarithmic colored fog,
Anti-aliasing (edges & textures), Translucency,
Gouraud shading, Smooth shading,
Sub-pixel correction, 16 millions colors (+)
Per-pixel alpha blending effects,
16-bit (65,5536)
Voodoo- up to 640x480 w/z-buffer up to 1280x1024x256 colors (+)
up to 800x600 w/o z-buffer
Voodoo Rush- up to 800x600 w/z-buffer
up to 1024x768 w/o z-buffer
3d acceleration in window
Software Support: (adv 3dfx; Native & Direct 3d; OpenGL support in the
future closer than PowerVR2)
The GamesMansion article:
Wipeout XL: (adv PowerVR2)
(ot available in US) "incredible frame rates" up to 1024x768
Pod: (adv 3dfx)
"The 3Dfx card did appear a hell of lot smoother, no doubt thanks to a
higher frame rate."
Direct 3D Tests: (draw; one thing to mention is that Vsync is a function
that matches the frame rate with your monitor's refresh rate. I don't know
what why it is shown to be such a big deal since the PowerVR2 does not have
video syncing; I also wish they renamed the benchmark filename, because of
some rumors)
Fill Rate- (draw: 1.3% difference)
30.88 31.28
Polygon throughput- (draw- 1.1% diff)
446 451
Intersections- (adv 3dfx: 22% diff)
4.93 4.03
Twist Test- (adv 3dfx: 33% diff)
151.51 101.01
Tunnel Test- (draw: 1.2% diff)
111.48 113.63
If you want to flame me saying in the PowerVR2 beat the 3dfx in the "draw"
instances, I will just say that a 1.1-1.3% difference is not statistically
viable to be considered a difference. I would say that the 2 cards are
evenly matched in 3 of 5 tests, but the 3dfx wins by a 20%+ margin in other
tests.
Price: (draw: maybe 3dfx when you look at Hercules' one card solution &
Voodoo prices are still dropping)
$150-250 US Retail on streets $199 MSRP or ~$130-150 Retail
(Flash3d-Stingray 128 3D)
Stingray OEM $200
In other words, it is NOT "a hell of a lot cheaper than the Voodoo cards
(Rush or otherwise)." I've never seen so much BS in my life and I love
their scientific way of determining how much better PowerVR2 is than 3dfx.
Somehow, I feel that there are people out there trying to make the PowerVR2
seem better than it really is.
Do I smell a conspiracy or am I paranoid?
}{ardball
>READ THIS AND WEEP !!!!!!!!!!
>
>HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
>
>http://www.gamesmansion.com/reviews/3dx/3dx.html
If I weep it's because I'm laughing.
I don't know how lagged your newsfeed is over in .pt (wherever that
is) but I suggest you read the article again. The person writing the
review had not a clue what he was doing.
ccorpora
In Article<33996F...@interalpha.co.uk>, <jee...@interalpha.co.uk> writes:
> What I was implying is that the 3Dfx needs to turn VSYNC waiting off.
> If this makes no difference to the screen appearance (which shows just
> how quick the 3Dfx is!) then this is a fair comparison. But if you get
> visible tearing (as in Outlaws! - from one 3Dfx owner) then it is not a
> fair comparison.
And Outlaws is only minor tearing. I fool with Outlaws with vsych off still.
Q.B.M.
ccorpora
In Article<33987C...@interalpha.co.uk>, <jee...@interalpha.co.uk> writes:
> So lets get this clear.
>
> 1, The 3Dfx (Voodoo) waits for VSYNC by default.
> 2, The 3Dx (Apocalypse) also waits for VSYNC by default.
>
> For the 3Dfx to be compared to the 3Dx you
> must disable the wait for VSYNC on the 3Dfx,
> but NOT for the 3Dx.
>
> And this makes for a fair comparison???????
>
>
> One obvious question is:
>
> Can the 3Dx be made ignore VSYNC also?
R:
Here's the realworld situation.
You can turn the vsynch off on the voodoo and get extra frames without
tearing. So what's the point in having it on?
The Vr2 I guess doesn't allow it to be turned off, or I still not clear on
this.
So if turning an option on/off doesnt' penalize a system why not use it to
your benifit?
I have no problem benchinc the 3dfx both ways on and off. But only a dolt
wouldn't take the extra frames and leave the synch. off.
Hey if the Vr2 allows you to turn it off easily and it helps without noticable
tearing then I say bench it that way as well.
Q.B.M.
Gary Tarolli
ccorpora wrote:
> You can turn the vsynch off on the voodoo and get extra frames without
> tearing. So what's the point in having it on?
Not quite true. If you are double buffered (as most games are), then you
will get tearing. The tearing may not be that noticeable in some games,
especially ones that pan left and right. But if you are in a airplane and
roll, you will see the tearing quite readily.
Unless of course you are triple-buffered, in which case the hardware really
is waiting for VSYNC, but it can continue rendering even while waiting (to
some extent). This makes it possible for a game that renders at exactly 50 Hz
say, to run at an AVERAGE of 50 Hz while waiting for VSYNC. The way this
works is that most frames swap after 1 VSYNC but once in a while the hardware
waits 2 VSYNC to swap. It's very hard to explain....
Anyway, the limitation is that you cannot run 640x480 triplebuffered on
Vooodoo if you are using a zbuffer. There's just not enough memory.
John Reynolds
In article <339ac16...@news.prestel.co.uk>,
Dave Gibbons <da...@flavious.prestel.co.uk> wrote:
% OK Jonathon. But I don't agree there is any question. If the PVR(2)
% is to avoid tearing it *must* wait for vsync.
Not that I think the PVR2 is doing it but I feel I should point out
something that a lot of people seem to be missing:
It IS theoretically possible to avoid tearing without waiting for the
vsync.
All it requires is enough framebuffer to handle triple buffering and
either a vertical blank interrupt generated by the card, or hardware
on the card to flip frames at the next vsync.
It's just a pity that most current 3D cards only have 2MB of
framebuffer, which makes this impossible at a reasonable resolution
with a z-buffer.
Cheers,
--
John M Reynolds
mailto:john.r...@ccc.ox.ac.uk
Corpus Christi College, Oxford, UK
John Reynolds
In article <339A9C...@interalpha.co.uk>,
Jonathan Pearce <jee...@interalpha.co.uk> wrote:
% After all, what is the point in being able to generate 110 fps
% if you can only 'clearly' display 60fps.
%
% The fact that the 3Dfx can handle not waiting for VSYNC without
% visible update-effects is a credit to it.
Eh?
With the wait for vsync switched off, the tearing is really quite
noticeable. I'd certainly call that a visible update-effect. Of
course some people are more bothered by this than others.
I suspect the reason people playing games on PCs don't mind this is
that most PC games tear quite badly, as syncing properly is difficult
due to the lack of a vertical blank interrupt.
Tim Smith
In article 01bc71de$9c285aa0$0100...@host.telepac.pt,
"Stephen Fordham" <sbfo...@mail.telepac.pt> said:
>
>READ THIS AND WEEP !!!!!!!!!!
>
>HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
>
>http://www.gamesmansion.com/reviews/3dx/3dx.html
>
was done by an individual who was clearly quite confused about what he was
doing...he essentailly compared apples to oranges and gave the award to the
grapefruit..
Subsequent posts to this group have provided evidence that the PowerVR
cheats all the time by not waiting for VSYNC even when you tell it to...but that's
not my chief argument...
After one of the guys from the company that pushes the powerVR posted
here, I challenged them to release their software developer's kit onto their
website, making it freely available to developers everywhere as the folks at
3Dfx have done......deafening silence! no free PowerVR SDK. They have no
confidence in their product, and the product is dead-end technology. After all
the registered developers have released their PowerVR ports and decided that
the real money is in the NEXT generation of card (as always happens), the
smaller developers won't be able to develop for your card, and nobody will be
able to develop any patches, updates, emulations or anything...you'll be stuck.
When 3DFx users have all the good software and all you have are the demos
that came with your card, I think you'll be the one weeping..
-Tim
Simon Fenney
In article <339cc335...@204.127.4.22>,
jim...@worldnet.att.net (Jim Husband) wrote:
>
> On Fri, 06 Jun 1997 20:20:33 -0400, Gary Tarolli <tar...@3dfx.com> wrote:
> >Simon Fenney wrote:
> >>
> >> I think you've got it the wrong way round. PowerVR always waits for the
> >> VSync, but my understanding was that the 3Dfx supporters were saying
> >> that the tester wasn't disabling it on the 3Dfx to get higher performance.
> >
> >If it's always waiting for VSync, then how can it produce 100+ frames per
> >second?
>
> Possibly by using a refresh of 120? I didn't see where refresh rates
> supports. Seems like a bit of a stretch though.
To a certain extent the refresh rate would depend on what the graphics
card can support. With both PCX1 and PCX2, the 3D card just sees is a
block of memory in PCI space which it then writes to. If the complexity
of the scene is not an issue (which is the case with the simple D3D
benchmarks), it all comes down to the bandwidth across PCI.
If you assume, say, that 50% of the theoretical PCI bandwidth is actually
available (i.e. 66Mb/s) for writing rendered pixels, then I guess you
could write 107 640x480x16bit images per second.
As you said, it seems like a bit of a stretch, but is theoretically (and
obviously practically) possible, though you'd need a good 2D graphics
card!
> A rate of 113 with VSync on
> would mean 90% of the time the fps was 120 or greater and %10 it was somewhere
> below 120.
Working backwards, I guess 120fps would need 56% of PCI bandwidth just for
writing pixels.
> If the PVR2 can get 113 fps with VSync on - then the z-Buffer issue
> and support are the only barriers they have left. If their D3D performance is
> that good - then they may actually succeed.
>
> I'm still interested in whether a z-Buffer can be implemented in D3D
> using the PVR/2 drivers. Simon?
Part of the idea behind PowerVR is to eliminate the traditional Z buffer.
A lot of applications want per-pixel Z sorting which PCX1 and PCX2
provide. They aren't interested in what is actually in the Z buffer. Why
use the memory for a Z buffer when games companies keep wanting more
texture memory?
Simon Fenney
Design Manager
-------------------==== Posted via Deja News ====-----------------------
http://www.dejanews.com/ Search, Read, Post to Usenet
ccorpora
In Article<339C2D...@3dfx.com>, <tar...@3dfx.com> writes:
> Path:
> ccorpora wrote:
>
> > You can turn the vsynch off on the voodoo and get extra frames without
> > tearing. So what's the point in having it on?
GARY:
> Not quite true. If you are double buffered (as most games are), then you
> will get tearing. The tearing may not be that noticeable in some games,
> especially ones that pan left and right. But if you are in a airplane and
> roll, you will see the tearing quite readily.
R:
Obviously you know best. Were waiting for that game where we can roll! :)
Come to think of it I can in Descent 2 and still notice almost zero tearing.
But again I don't argue that tearing could come into play, but with the
software available right now it's a very very minor issue. That's what I was
basing it off of.
Q.B.M.
ccorpora
In Article<5ndnsb$o2i$1...@paleale.wanet.net>, <tims...@tlsmith.wanet.com>
writes:
> 5. Because some people want to brag about superior frame rates, they tell
their cards
> to ignore VSYNC and simply crank out the max FPS it can. The number can
become
> meaningless since the user is not seeing some frames in their
entirety...but that doesn't
> matter if your goal in life is to brag about frame rate numbers.
R:
I think you can substitue brag with inform. At least in my case. Hey I've
helped out many with maximizing voodoo's performance. It's for others benifits
as to why I right. I could careless what people percieve of me since this
forum is very crude in getting a persons feelings and tones out. All my posts
are genuine.
>
> 6. This whole thing becomes very complex because most graphics cards have
different
> features...so you're not really even seeing the same images with 2
different cards...and
> because computer monitors, unlike TVs, can change their screen
Vsync,refresh rates...
> so a benchmark addict can tweak his video refresh rate to achieve better
numbers.
R:
The easiest way to generalize features is to just see what is being offered
and compare it. If one is obviously visually superior in a few different apps.
then your safe to assume it's equal to or better. At that time you look at
support and apply the same rules.
Q.B.M.
M. R. Embury
> Regarding this, where the hell can you get updated drivers for the
> PVR?
Is that with Vsnyc on or off! :-)
Werner Punz
corp...@sable.ox.ac.uk (John Reynolds) wrote:
>I suspect the reason people playing games on PCs don't mind this is
>that most PC games tear quite badly, as syncing properly is difficult
>due to the lack of a vertical blank interrupt.
>
Ahem, DirectDraw (the 2d API not the 3d) has solved this problem. So
if a DirectDraw driver is implemented correctly no tearing in 2d games
and videos should occur.(Although it doesn't bother me at all)
Werner
---
mailto://we...@inflab.uni-linz.ac.at
http://witiko.ifs.uni-linz.ac.at/~werpu/
FOR A REPLY PLEASE REMOVE THE NOSPAM FROM MY E-MAIL
ADDRESS.
Dave Gibbons
On 9 Jun 1997 14:34:05 GMT, corp...@sable.ox.ac.uk (John Reynolds)
wrote:
In article <339ac16...@news.prestel.co.uk>,
Dave Gibbons <da...@flavious.prestel.co.uk> wrote:
>% OK Jonathon. But I don't agree there is any question. If the PVR(2)
>% is to avoid tearing it *must* wait for vsync.
>Not that I think the PVR2 is doing it but I feel I should point out
>something that a lot of people seem to be missing:
>It IS theoretically possible to avoid tearing without waiting for the
>vsync.
Yes I know this.
I did comment on it in the thread that this thread spewed from :-
}}So ok the Apocalypse renders 110 frames per second - if it has to
}}wait for VSYNC at 120hz then 110 frames is MEANINGLESS it will wait
}}for the next sync at 120hz. Triple buffering can improve performance
}} in this respect. (Oh shit I've given you another idea now!)
Even tho' you don't end up waiting as such for vysnc in a triple
buffered rendering system you still need it for flipping pages to
*eliminate* tearing. Its a little problematic with the PC though with
the lack of interrupts. One of my pet hates is wildly fluctuating
frame rates - I prefer an average rather than a superlative that
happens once in a while. Some people think that 3x is a panacea
- it can be, but sometimes people go too far.
Dave
da...@flavious.prestel.co.uk
Steve
On 6 Jun 1997 07:06:35 GMT, Tim Smith <tims...@tlsmith.wanet.com>
wrote:
>>Ty <ty...@aloha.com> wrote:
>>>> - WIPEOUT 2097XL (Runs at 640x480 in 32,000 colors at 40+ fps).
<SNIP>
>He gets it in his dreams...Do you think he's running his monitor at
>a 40Hz refresh rate (imagine the flicker at THAT rate <grimace>)
>or perhaps he's running his monitor at 80Hz? Or is he not syncing
>the video card page flips to Vsync...in which case his frame rates
>are silly nonsense, since some fromes are not displayed in their
>entirety (which is why THAT technique often results in screen
>"tearing").
He could be running his monitor at the standard 60Hz and skipping
every third frame without tearing. The FPS doesn't have to match the
vertical refresh rate to avoid tearing.
>IMHO people who look for frame rates in excess on the monitor
>sync rates are playing goofy mand games.
40Hz is well below the VESA rate of 72/75Hz, let alone the refresh
rates supported by high end monitors of upto 120Hz (Iiyama 17" Vision
Master Pro, spec from evesham micros).
> The REAL goal of
>everyone should be to get hardware that displays the most-
>detailed, most complex, highest rez, highest quality images at
>a standard video refresh rate...
Video refresh rate, as in video video?? Which country would you call
standard? uk/europe = PAL, 50Hz...us/japan = NTSC, 60Hz
>THAT would be a good goal
>(of course the benchmarks would be a bit subjective).
>
>-Tim
What would look better than any of the above is...
1) True hardware anti-aliasing without a huge performance hit. (Voodoo
25-50% drop in performance with software bodged A-A turned on.)
2) Depth of field effect, will it ever be possible in realtime??
These would give you an incredible sense of realism but 1) is not here
yet and 2) won't be for years.
Cheers
Steve
Simon Fenney
In article <5nirve$cco$1...@paleale.wanet.net>,
Tim Smith <tims...@tlsmith.wanet.com> wrote:
>
> In article 01bc71de$9c285aa0$0100...@host.telepac.pt,
> "Stephen Fordham" <sbfo...@mail.telepac.pt> said:
> >
> >READ THIS AND WEEP !!!!!!!!!!
> >
> >HERE IS A REVIEW THAT PUTS THE PowerVR 2 IN FRONT OF THE 3DFX :
> >
> >http://www.gamesmansion.com/reviews/3dx/3dx.html
> >
> That so-called "review" is what started the whole VSYNC debate! The review
> was done by an individual who was clearly quite confused about what he was
> doing...he essentailly compared apples to oranges and gave the award to the
> grapefruit..
>
> Subsequent posts to this group have provided evidence that the PowerVR
> cheats all the time by not waiting for VSYNC even when you tell it to...but
that's
> not my chief argument...
A good thing it's not your "chief argument", because it is completely
incorrect. The card will wait for the VSync - unless the application has
set up triple buffering on the 2D card.
And now I must get back to work.
Simon Fenney
Design Manager
Videologic
Ty
Tim Smith wrote:
>
> Agreed...with DRAM so cheap, I wish 3Dfx cards were available with
> 24MB of EDO...8 for frames, 8 for textures, 8 for Zbuff/ alpha, or
> whatever...
Just curious, what would you need 8 megs for the frame buffer for or
even the Zbuffer?
Ty
ccorpora
> And now I must get back to work.
>
> Simon Fenney
> Design Manager
> Videologic
R:
Well this is a step in the right direction. :)
Sorry Simon.
Q.B.M.
John Reynolds
In article <33a9c280...@news.uni-linz.ac.at>,
Werner Punz <we...@inflab.uni-linz.ac.at> wrote:
% corp...@sable.ox.ac.uk (John Reynolds) wrote:
%
% >I suspect the reason people playing games on PCs don't mind this is
% >that most PC games tear quite badly, as syncing properly is difficult
% >due to the lack of a vertical blank interrupt.
%
% Ahem, DirectDraw (the 2d API not the 3d) has solved this problem. So
% if a DirectDraw driver is implemented correctly no tearing in 2d games
% and videos should occur.(Although it doesn't bother me at all)
I stand corrected.
Out of interest, how DOES DirectDraw manage that trick?
ttammi
Tim Smith <tims...@tlsmith.wanet.com> wrote:
>5. Because some people want to brag about superior frame rates, they
>tell their cards to ignore VSYNC and simply crank out the max FPS it
>can. The number can become meaningless since the user is not seeing
>some frames in their entirety...but that doesn't matter if your goal
>in life is to brag about frame rate numbers.
Or to benchmark 3D cards' performance. Wasn't that the issue here?
Benchmarking 3D cards with vsync enabled can be pretty meaningless in
lots of cases.
Peter O'Boyle
ccorpora <ccor...@neo.lrun.com> wrote:
Credit where credit's due - I used Tweek last night on F1. The speedup
is <very> noticeable. Ninja rocks!
Thanks cc...
Peter
Tim Smith
In article 339E86...@aloha.com,
1. high resolution (x,y, and colorspace) plus triple buffering.
( (1024x768x16bit = 1.5+MB) * 3 buffers) = 4.5+MB
2. Some blit or overlay space...it would be nice to have an image
buffer for instrument panels and the like that could be overlayed
over the 3D stuff on-the-fly by the video chip based on some
alpha buffer. (This would need an improved chipset of course).
3. A larger Zbuffer would (with an enhanced chipset) permit larger
(or finer-resolution) Z-buffer settings...I seem to recall that the current
hardware uses a 16-bit number for each entry in the Z-buffer.
4. Because of current RAM limitations on the card/chipset, some
features of the 3Dfx chipset cannot be used simultaneously...that's
a shame.
The dream of a huge texture buffer is, of course, obvious.
-Tim
Tim Smith
In article 339de5ed...@news.dial.pipex.com,
postm...@127.0.0.1 (Steve) said:
>On 6 Jun 1997 07:06:35 GMT, Tim Smith <tims...@tlsmith.wanet.com>
>wrote:
>> The REAL goal of
>>everyone should be to get hardware that displays the most-
>>detailed, most complex, highest rez, highest quality images at
>>a standard video refresh rate...
>Video refresh rate, as in video video?? Which country would you call
>standard? uk/europe = PAL, 50Hz...us/japan = NTSC, 60Hz
No, no, no,...hadn't you noticed that video refresh rates on computers
are not tied to AC-line frequencies? (They ARE in television, where NTSC
systems are 60Hz in the U.S. and PAL sets are 50Hz in Europe).
Many video card/monitors run 60Hz, 72Hz, 75Hz, 80Hz, etc in different
modes. People prefer the higher rates because the images tend to flicker
less and are much easier on the eyes.
>What would look better than any of the above is...
>1) True hardware anti-aliasing without a huge performance hit. (Voodoo
>25-50% drop in performance with software bodged A-A turned on.)
Yes, this would LOOK better, but the arguments were about benchmarking
speed. All I was saying is that we need a benchmark that says something
like "Each card shal render Scene X with all the features in List Y at
a particular resolution and monitor sync rate". As long as each card in
a benchmark renders different images at different resolutions with different
features and at different monitor refresh/sync rates, we CAN'T have any
honest, meaningful benchmarks.
>2) Depth of field effect, will it ever be possible in realtime??
Agreed! That's one I'd LOVE to see!
-Tim
Tim Smith
In article VA.0000010...@me.netlife.fi,
Yes, it's completely meaningless if the cards are running at different VSYNC
rates. If, however, you force two cards to wait for VSYNC, and then force both
to run at the same VSYNC rate then the effects of that part of the benchmark
argument fall away and allow you to examine other aspects of their performance.
When the monitors are running at different rates, or one or both cards are not
waiting for VSYNC, the results of your benchmarks say as much about these
issues as they do about the actual performance of the cards.
-Tim
Tim Smith
In article 8660246...@dejanews.com,
Simon Fenney <sim...@videologic.com> said:
[SNIP]
>A good thing it's not your "chief argument", because it is completely
>incorrect. The card will wait for the VSync - unless the application has
>set up triple buffering on the 2D card.
>
>
>Simon Fenney
>Design Manager
>Videologic
1. Then how do you account for the benchmark results posted here several
days ago indicating that the card benched the same with VSYNC turned on
while running at a handful of different video rates...the rate changes should
have had SOME impact upon the benchmarks. This leads to an interesting
question: I hear that your card writes into the video buffer of the user's 2D
card...if this is the case, HOW do you manage the VSYNC wait trick. Does
your card/software know the nasty details of the registers of all the 2D cards
on the market in order to read/wait for the ACTUAL vsync, or do you rely on
the old 6845 control registers that most chipsets emulate or shadow? I seem
to recall that the VSYNC in some chipsets has no direct relation to the actual
sync rate and is just there to support old code.
2. I have still heard no reply to my previous query to you about your company's
SDK...any chance your'll ever make one generally available like your 3Dfx
competitors have done?
BTW: don't take this wrong...I greatly appreciate your taking the time and
effort to participate in this group...but I'm really wary of hardware/software
companies that try to create new standards, then keep small developers and
users from having access to the docs. (and yes, I'm wary of Microsoft doing this
too...)
-Tim
David Nagy
I just exchanged emails with Simon Fenney, and the process helped me at
least understand this double vs. triple-buffering concept better.
I retract my earlier statement that triple buffering won't really help
framerates in real-world situations. It *does* de-couple the rendering
from the screen retrace in the sense that it gives the renderer someplace
to "work" while the screen is being redrawn *and* the 'other'
accumulation buffer is already full.
This is very useful because in many cases the render rate is just a
little slower than the refresh rate. Without the extra buffer you will
repeatedly be in a situation where you finish rendering into the
accumulation buffer just a *little bit* after the latest refresh starts.
Without another buffer to work on, the renderer has no choice but to stop
work until the buffers swap at the end of the current refresh.
With two buffers, (in addition to the frame buffer that is used to write to
the screen) the renderer will only stall if its already ahead of the
game. And in that case, it deserves a little break. :)
Dave Nagy
Dave Gibbons
On Fri, 13 Jun 1997 23:24:33 GMT, dave...@netcom.com (David Nagy)
wrote:
<snip>
>game. And in that case, it deserves a little break. :)
>
>
> Dave Nagy
Nobody has said as far as I'm aware, that it isn't a worhwhile
measure. Its that it is only better than 2x under certain
circumstances. It doesn't absolve you from taking vysnc into account.
Which I think was the proposal that was put forward - with relation to
the gamesmansion tests.
David Nagy
Dave Gibbons (da...@flavious.prestel.co.uk) wrote:
[re: triple buffering]
: Nobody has said as far as I'm aware, that it isn't a worhwhile
: measure.
Well, I said that... Which is why I wanted to set things straight. :)
: Its that it is only better than 2x under certain
: circumstances. It doesn't absolve you from taking vysnc into account.
Yes, but I can see that it would be pretty darn useful *most* of the time.
Or, at least a lot of the time.
Let's pretend that the 'average' computer can run the average game at
30FPS if vsync is ignored. (3D accelerated of course) If the average
refresh rate is somewhere around 70Hz, then if we do wait for the retrace
we'll just miss being able to draw on every other redraw. (35Hz) Waiting
for the next one stalls the rendering if there's only one accumulation
buffer, and knocks the instantaneous framerate down to 23.3FPS. (Every
third retrace)
That's fairly brutal drop just to avoid tearing. With the addition of
another buffer, at least the renderer can stay busy all the time. My
math isn't up to figuring out what sort of average rate could be
maintained, but it should be somewhat higher than 23.3FPS. Since the
renderer is working constantly, it seems like the average rate should be
approaching the 'un-synced' rate. Shouldn't it?
Dave Nagy
Dave Gibbons
On Sat, 14 Jun 1997 06:28:46 GMT, dave...@netcom.com (David Nagy)
wrote:
Dave Gibbons (da...@flavious.prestel.co.uk) wrote:
[re: triple buffering]
: Nobody has said as far as I'm aware, that it isn't a worhwhile
: measure.
>Well, I said that... Which is why I wanted to set things straight. :)
OK sorry, I must have missed that :-)
: Its that it is only better than 2x under certain
: circumstances. It doesn't absolve you from taking vysnc into account.
>Yes, but I can see that it would be pretty darn useful *most* of the time.
>Or, at least a lot of the time.
Yes, to a degree (and its all about degrees) , I just think your being
a bit optimistic. We'll have to agree to disagree :-)
B.K.Myskja
Let`s say that we can run a program at 5fps with Vsync _on_ at 75Hz
refresh rate. Now, if we turn Vsync _off_, the program will _still_
run at 5fps, right?
B.K.Myskja