68030/68882 vs 80486 (wa
Marcus Finlay
Gaven Miller wrote to All:
GM> 386/33 is 5 MIPS
GM> 030/33 is 5.5 MIPS
GM>
GM> 486 DX2/66 is 44 MIPS
GM> 040/66 ("DX2") is 49 MIPS
GM>
GM> Pentium 60 is supposedly 100MIPS
GM>
GM> So, the 386/030 comparison is more valid.
GM> Likewise 040/486 is a more valid comparison.
Hmm....figures I have rate a 486DX/50 at 22 Mips
486DX2/66 is 26 Mips
030/20 is 8 Mips so a 030/32 would be around 10 Mips
040/33 is 26 Mips
A 040 is nearly twice a 486DX @ the same clock rate
A PPC (not sure which model, but a 32bit job) is more then 90% faster
than a Pentium @ the same clock rate.
You and I both know that Motorola have always been ahead of intel :-)
GM> > You're not wrong by much. In a 'real' world test performed by my TT
GM> > owning friend, (4meg ST/4meg TT RAM) who compiled POV Ray 2.2 for himself
GM> > and compared render times on the same scene file at the same (sizeable)
GM> > resolution, we found that his TT was matching 486DX50 machines with
GM> > more RAM. That to me says that the 68030/68882 pairing, given 32-bit wide
GM> > RAM access (TT-ram) is actually faster at a given clock speed than a
GM> > 486DX chip. Imagine how much quicker the 68040 is than a 486DX. 040 is
GM> > reputed to command thrice the grunt of the 030/882 at the same clock
GM> > frequency!! Food for thought eh?...
GM>
GM> Was that a _GENUINE_ DX50, or a clock-doubled DX2 25/50.
I ported the Unix version of POV to my TT, and it was very close to the
DX2 50 (Sorry Chris) at rendering the same image at the same size
When I got TT RAM it was matching it. I was also "embarressing" my
40Mhz 040 Quadra owning friend, it was too much slower (can't remember
how much slower) but it had him "worried".
Cheers
Marcus
___
* Kivi (unregistered) 1.41 *
Brent
MF>Hi Gaven,
MF>Gaven Miller wrote to All:
GM> 386/33 is 5 MIPS
GM> 030/33 is 5.5 MIPS
GM>
GM> 486 DX2/66 is 44 MIPS
??? According to the figures I have sitting here in front of me "INTEL
DX2-66Mhz POWER METER V1.81 MIPS 20.8 MIPS", the closest figure I can
find to your 44 MIPS is the AMD DX4-100+ at 39.0 MIPS, these are specs
are from a motherboard manual trying to sell this mother board
interesting ??
GM> 040/66 ("DX2") is 49 MIPS
GM>
GM> Pentium 60 is supposedly 100MIPS
I would doubt that as in my experence a Pentium 60 is about the same
speed as a DX2-80 around 23.3 MIPS ( perhaps slower ), note this is
how it appeares using a machine running 8 Mb ram ( both machines ) not
based on actual tests, but the usability of this is what we in the
real world are more interested in ( I hope ) not figures made up by
some manufacturer to claim their processor is greater that the
competitions.
GM> So, the 386/030 comparison is more valid.
GM> Likewise 040/486 is a more valid comparison.
MF>Hmm....figures I have rate a 486DX/50 at 22 Mips
MF>486DX2/66 is 26 Mips
Is that for a 486DX-50 or DX2-50 ? if DX2 then my figures rate that
processor at 15.9 MIPS for an Intel
MF>030/20 is 8 Mips so a 030/32 would be around 10 Mips
MF>040/33 is 26 Mips
I seem to recollect once hearing ( perhaps reading ) the 68000 was
around 8 MIPS
Marcus could perhaps confirm, I think its in the processor manual, if
you still have your copy maybe you could look it up, if not I can tell
you were to find my copy :-)
MF>A 040 is nearly twice a 486DX @ the same clock rate
MF>A PPC (not sure which model, but a 32bit job) is more then 90%
MF>faster than a Pentium @ the same clock rate.
MF>You and I both know that Motorola have always been ahead of intel
MF>:-)
Tell me about it, and the processors dont have have as much trouble
just a shame the major manufactures computers using their processors
are usless at marketing and/or removing the stigma of being a games
machine. Interesting point to note what are most clones ( 80x86 based
) machines used for now, I would like to conjecture games would be the
answer :-)
MF>I ported the Unix version of POV to my TT, and it was very close to
MF>the DX2 50 (Sorry Chris) at rendering the same image at the same
MF>size When I got TT RAM it was matching it. I was also
MF>"embarressing" my 40Mhz 040 Quadra owning friend, it was too much
MF>slower (can't remember how much slower) but it had him "worried".
Marcus I am about to lay my hands on a copy of C perhaps we should
compile it on my DX4-100, and do a comparision, I have a felling that
optomizations are performed in the clone versions including some
assembly where as yours was full C code or did you complile the clone
one as well ??, I also recon different compilers make a difference,
mind you we have seen that one ourselves
Cheers
havoc
Gaven Miller
> MF>Hi Gaven,
> MF>Gaven Miller wrote to All:
> GM> 386/33 is 5 MIPS
> GM> 030/33 is 5.5 MIPS
> GM>
> GM> 486 DX2/66 is 44 MIPS
> ??? According to the figures I have sitting here in front of me "INTEL
> DX2-66Mhz POWER METER V1.81 MIPS 20.8 MIPS", the closest figure I can
> find to your 44 MIPS is the AMD DX4-100+ at 39.0 MIPS, these are specs
> are from a motherboard manual trying to sell this mother board
> interesting ??
I didn't have the correct figures on hand at the time, so I simply scaled
the "DX/33" figures up. (Bad practice, I know).
Having just done the calcs again, I see that my "44" was wrong. I simply
scaled the 20 MIPS (Intel's claim at the time of the 486 launch was "15
MIPS for them 486/25, 20 MIPS for the 486/33") up by 1.7. But something
went wrong somewhere.
The best figure I could get was 34 MIPS. ("Dyslexia of the typing fingers"
is what I plead)
Should have gone searhing for the correct figures. Sorry.
> GM> 040/66 ("DX2") is 49 MIPS
> GM>
> GM> Pentium 60 is supposedly 100MIPS
040/25 was IIRC 22MIPS, so linearly scaling it to 33MHz, then scaling it
by 1.7 (to "clock double" it) gives (22*(33/25))*1.7=49.3 MIPS. (I am not
too happy about the "linear scaling" of the 25Mhz MIPS rate however.
(I have a niggling suspicion that 040/25 was 18 MIPS, 040/33 was 22. In
this case the figure would be 22*1.7=37.4. But, I have been wrong in the
past with such suspicions)
Re: P5/60 = 100:
That came from a number of PC computer mags [hangs head in shame at
reading and then reciting opposition propaganda] at the time of the
Pentium release. The 66Mhz was "over 100Mhz", and then Intel would release
a 60Mhz "entry level" model, which also met Intel's "100Mhz" target.
> I would doubt that as in my experence a Pentium 60 is about the same
> speed as a DX2-80 around 23.3 MIPS ( perhaps slower ), note this is
> how it appeares using a machine running 8 Mb ram ( both machines ) not
> based on actual tests, but the usability of this is what we in the
> real world are more interested in ( I hope ) not figures made up by
> some manufacturer to claim their processor is greater that the
> competitions.
P5/60 equalling dx2/80? Crap motherboard/chipset in the P5?
> GM> So, the 386/030 comparison is more valid.
> GM> Likewise 040/486 is a more valid comparison.
> MF>Hmm....figures I have rate a 486DX/50 at 22 Mips
> MF>486DX2/66 is 26 Mips
> Is that for a 486DX-50 or DX2-50 ? if DX2 then my figures rate that
> processor at 15.9 MIPS for an Intel
> MF>030/20 is 8 Mips so a 030/32 would be around 10 Mips
> MF>040/33 is 26 Mips
> I seem to recollect once hearing ( perhaps reading ) the 68000 was
> around 8 MIPS
> Marcus could perhaps confirm, I think its in the processor manual, if
> you still have your copy maybe you could look it up, if not I can tell
> you were to find my copy :-)
> MF>A 040 is nearly twice a 486DX @ the same clock rate
> MF>A PPC (not sure which model, but a 32bit job) is more then 90%
> MF>faster than a Pentium @ the same clock rate.
> MF>You and I both know that Motorola have always been ahead of intel
> MF>:-)
But unfortunately Intel is catching up fast. And Motorola have given up
on enhancing the 68k line in favour of the PPC.
--
Quote For The Month:
"If she's a lady, I'm a Pernicious Knid"
Tony Smolar
>DX2-66Mhz POWER METER V1.81 MIPS 20.8 MIPS", the closest figure I can
>find to your 44 MIPS is the AMD DX4-100+ at 39.0 MIPS, these are specs
>are from a motherboard manual trying to sell this mother board
>interesting ??
Accurate MIPS figures are hard to come by, but the AMD DX4-100 is around
50 MIPS, maybe more, depending on who you believe.
>I would doubt that as in my experence a Pentium 60 is about the same
>speed as a DX2-80 around 23.3 MIPS ( perhaps slower ), note this is
>I seem to recollect once hearing ( perhaps reading ) the 68000 was
>around 8 MIPS
An 8 Mhz 68000 is between 0.8 and 1.2 MIPS. I've used the same CPU benchmarks
on my 8mhz Atari STe, and my AMD DX4-100, and the DX4 consistantly comes
out 40-50 times faster than the STe, so if the 68000 really was 8 MIPS as you
say, than my DX4 would be at least 320 MIPS, and I KNOW that's not the case.
(It'd be nice though :) )
Brent
Ok getting of topic here namly Atari advocacy
re:
BY> I would doubt that as in my experence a Pentium 60 is about the
BY>same speed as a DX2-80 around 23.3 MIPS ( perhaps slower ), note
BY>this is how it appeares using a machine running 8 Mb ram ( both
BY>machines ) not based on actual tests, but the usability of this is
BY>what we in the real world are more interested in ( I hope ) not
BY>figures made up by some manufacturer to claim their processor is
BY>greater that the competitions.
GM>P5/60 equalling dx2/80? Crap motherboard/chipset in the P5?
Hmm possably, wont name the manufactuer (sp?), but it interesting to
note that many of us in sales and usage of these beige boxes, we
compare a DX4-100 to a Pentium 75, the processor speed is comparable
with the Pentium only slightly on top in most in most instances except
1 area ( admitidly becoming more important as games players want
better more relistic graphics ), video.
GM>But unfortunately Intel is catching up fast. And Motorola have
GM>given up on enhancing the 68k line in favour of the PPC.
Interestingly though only catching up on the 68k line, and not willing
to drop that older system support ( DOS ) from processors up until now
ie the Pentium pro, interestily enough though no one is willing ot go
out on a ledge and use it yet due to so many years of backward
compatability
Anyhow I should bug out now as this is being done on one of those
horrid biege boxes. Unfortionatly my trusted Atari was kill due to
some rain + 1* idiot fellow worker, whilst using for an audio task I
would not trust to a clone, long story cut short :-(
Cheers
havoc
Johan Klockars
Gaven Miller <gmi...@inca.co.nz> wrote:
>Brent (ha...@earthlight.co.nz) wrote:
>> Marcus Finlay <marc...@pointless.gen.nz> wrote:
>
>> MF>Gaven Miller wrote to All:
>040/25 was IIRC 22MIPS, so linearly scaling it to 33MHz, then scaling it
>by 1.7 (to "clock double" it) gives (22*(33/25))*1.7=49.3 MIPS. (I am not
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
You have to shave that 1.7 factor back off again, since there is no such
thing as a "clock doubled" '040. The 33 MHz chip you're talking about is
_exactly_ the same thing that some deluded people are now calling 66 MHz.
>P5/60 equalling dx2/80? Crap motherboard/chipset in the P5?
The P5 should be a lot better than that, especially for recompiled code.
>> MF>030/20 is 8 Mips so a 030/32 would be around 10 Mips
The figure I've seen is about 3.5 for a 16MHz '030.
Of course MIPS can mean any number of things, so comparisons are very
unlikely to give useful results.
If you count actual instructions executed per second (about the least
useful thing to do), a 20MHz '030 would top at 10 MIPS (or 6.5 if there
are no 2 cycle instructions, I can't recall at the moment).
Most instructions take much longer than that, though, and then there's
memory to worry about.
>> I seem to recollect once hearing ( perhaps reading ) the 68000 was
>> around 8 MIPS
Maybe if you could mange to run it at 66 MHz or so...
--
Chalmers University | Why are these | e-mail: ra...@cd.chalmers.se
of Technology | .signatures | jo...@rand.thn.htu.se
| so hard to do | WWW/ftp: rand.thn.htu.se
Gothenburg, Sweden | well? | (MGIFv5, QLem, BAD MOOD)
Johan Klockars
It's not always a good idea to post in the middle of the night...
(I managed to confuse the DX50 vs DX2/66 with the SX50 vs DX50 somehow.)
In article <60331203653$6F...@pointless.gen.nz>,
Marcus Finlay <marc...@pointless.gen.nz> wrote:
>Hi Gaven,
>
>Gaven Miller wrote to All:
>
>
>>030/20 is 8 Mips so a 030/32 would be around 10 Mips
>>040/33 is 26 Mips
>>
>>A 040 is nearly twice a 486DX @ the same clock rate
Performance per MegaHerz is a totally irrelevant number.
You can get the 486 at three or four times the clock frequency, you know.
>GM> > You're not wrong by much. In a 'real' world test performed by my TT
>GM> > owning friend, (4meg ST/4meg TT RAM) who compiled POV Ray 2.2 for himself
>GM> > and compared render times on the same scene file at the same (sizeable)
>GM> > resolution, we found that his TT was matching 486DX50 machines with
>GM> > more RAM. That to me says that the 68030/68882 pairing, given 32-bit wide
>GM> > RAM access (TT-ram) is actually faster at a given clock speed than a
Something must've been _very_ wrong with your tests.
A 32MHz '030+'882 should be far behind a 50MHz 486DX at normal floating
point work, not to mention integer.
>GM> > 486DX chip. Imagine how much quicker the 68040 is than a 486DX. 040 is
At non-fp work an '040 is perhaps 50% faster per clock than a 486.
The 486's are available at _much_ higher clock speeds, though.
>GM> > reputed to command thrice the grunt of the 030/882 at the same clock
It could actually be a lot more than three times as fast at fp.
>When I got TT RAM it was matching it. I was also "embarressing" my
>40Mhz 040 Quadra owning friend, it was too much slower (can't remember
>how much slower) but it had him "worried".
That must've been an '040 without an FPU, or your tests were really
_completely_ screwed up.
Chris Wilkinson
: I hope the cancelling of my previous version of this message worked.
: It's not always a good idea to post in the middle of the night...
Sometimes one has to post in the middle of the night (only because
NetscRape users hog net bandwidth during the day) :-)
: >GM> > resolution, we found that his TT was matching 486DX50 machines with
: >GM> > more RAM. That to me says that the 68030/68882 pairing, given 32-bit wide
: >GM> > RAM access (TT-ram) is actually faster at a given clock speed than a
:
: Something must've been _very_ wrong with your tests.
: A 32MHz '030+'882 should be far behind a 50MHz 486DX at normal floating
: point work, not to mention integer.
My mistake there. I posted the original about POVRay comparisons. The
machine that the TT030 matched was a DX2/50. I still believe that 030/882
can munch int and fp as fast (or pretty damn close) as a 486DX...
: >When I got TT RAM it was matching it. I was also "embarressing" my
: >40Mhz 040 Quadra owning friend, it was too much slower (can't remember
: >how much slower) but it had him "worried".
: That must've been an '040 without an FPU, or your tests were really
: _completely_ screwed up.
Maybe the Mac was using a slow port of POV, but Quadra 840AV were _NEVER_
shipped with 68LC040 'economy' version of the '040. They all were shipped
with the full 68040RC chip AFAIK.
Ciao...
Chris W.
/|\