Apple to ditch IBM, switch to Intel chips (CNET - This is real)
ELVIS2000
Randy Howard
@ElvisLives.com says...
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
Oh boy, did I ever pick the wrong time to buy a Mac. :-(
And Intel bagged them, even though AMD clearly has a better product
offering. Not encouraging. I wonder what the last version of OS X
will be to support my sudden boat-anchor?
--
Randy Howard (2reply remove FOOBAR)
"I don't really care about being right you know,
I just care about success." --Steve Jobs
Lloyd Parsons
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <gp72a1h2v682hnb7r...@4ax.com>, elvis2000
> @ElvisLives.com says...
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-
> > 5731398.html?part=rss&tag=5731398&subj=news
>
> Oh boy, did I ever pick the wrong time to buy a Mac. :-(
>
> And Intel bagged them, even though AMD clearly has a better product
> offering. Not encouraging. I wonder what the last version of OS X
> will be to support my sudden boat-anchor?
I wouldn't start worrying yet. This particular rumor has popped up many
times in the past.
imout...@mac.com
> In article <gp72a1h2v682hnb7r...@4ax.com>, elvis2000
> @ElvisLives.com says...
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
>
> Oh boy, did I ever pick the wrong time to buy a Mac. :-(
>
> And Intel bagged them, even though AMD clearly has a better product
> offering. Not encouraging. I wonder what the last version of OS X
> will be to support my sudden boat-anchor?
10.4 officially supports "firewire" Macs and above. That's 5-6 years
back. So I would expect, if Apple were to abandon PPC, your Mac would
be a first-class OS X box for the next 5 years at least.
Peter Ammon
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
Yeah, clearly the PowerPC has no future, which is why Sony, Microsoft,
and Nintendo are all using PowerPC chips in their next generation
consoles. (Ok, all we know about Nintendo so far is that IBM is making
the chip, but c'mon.)
-Peter
--
Pull out a splinter to reply.
Randy Howard
imout...@mac.com says...
Will there be an OS 11, with a replacement for Cocoa, or will they
just pick up an x86-64 Linux distro and slap aqua on top of it,
or will they try and keep up the "one OS fits all" silliness with
an that supports multiple native processor architectures with
something like the transitive product?
I wonder how much of the lack of polish on 10.4 is because their
best have already been pulled off to work on OS 11?
spinning... :-)
Paul McCahan
ELVIS2000 <elvi...@ElvisLives.com> wrote:
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-57
> 31398.html?part=rss&tag=5731398&subj=news
this isn't going to happen, it's a red herring designed to throw people
off. steve p doesn't speak until tuesday, so the article is wrong from
the beginning.
imout...@mac.com
> Will there be an OS 11, with a replacement for Cocoa,
there's nothing about cocoa that requires PPC. Remember, it started on
68K, and was ported to SPARC and x86 in the mid-90s. All the cruddy #if
__X86 compiler macros in the frameworks have been retained, and
throught 10.x's development x86 builds have been done to ensure the
thing as a whole at least runs on x86.
> or will they
> just pick up an x86-64 Linux distro and slap aqua on top of it,
They have their own distro and it's called Darwin.
> or will they try and keep up the "one OS fits all" silliness with
> an that supports multiple native processor architectures with
> something like the transitive product?
>
> I wonder how much of the lack of polish on 10.4 is because their
> best have already been pulled off to work on OS 11?
Like I said, if Apple goes x86 there won't be any user-level
differences. Developers will have to construct fat-binaries that run on
either platform. This is a solved problem for Apple/NeXT.
brianleahy
> this isn't going to happen, it's a red herring
I hope you're right.
> if Apple were to abandon PPC, your Mac would be a first-class OS X box
> for the next 5 years at least.
..while all the cutting edge development, of both hardware and
software, would go into the (gag) new Intel Macs....
--
brianleahy
10.4 Good Buddy :cool:
G5 Dual 2GHZ / 160GB / 1GB RAM / Superdrive
Apple 20" Cinema Display
SmartDrive 120GB Firewire HD
Maxtor 250GB SATA
Visit my wife's eBay store - hand crafted gifts!
http://stores.ebay.com/Catchy-Creations-by-brendaonline
------------------------------------------------------------------------
brianleahy's Profile: http://www.macosx.com/forums/member.php?userid=456
View this thread: http://www.macosx.com/forums/showthread.php?t=222556
macosx.com - The Answer to Mac Support - http://www.macosx.com
Timberwoof
ELVIS2000 <elvi...@ElvisLives.com> wrote:
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-57
> 31398.html?part=rss&tag=5731398&subj=news
At this point, this is real speculation.
--
Timberwoof <me at timberwoof dot com> http://www.timberwoof.com
If Macintosh is a luxury cruise ship,
then Linux is a freighter with wood paneling in the officers' quarters.
Charles Jo
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
As seen on News.com TalkBack:
Pretty kewel! And complementary:
Intel Inside/Apple Outside
Blue Men Dancing/Other Bright Silhouettes Dancing
Wonder what the pundits will call this duo? Aptel? Inpple? Intelapple?
AI? OSx86?
____________________________
Related links:
http://apple.slashdot.org/article.pl?sid=05/06/04/0238235&tid=118&tid=3&tid=137
http://www.geektimes.com/michael/techno/computing/hardware/products/apple/macintosh/misc/project-star-trek.html
Unrelated link:
http://www.shawnevans.com/
Blatant plug:
CharlesJo.com
Seeker1
ELVIS2000 <elvi...@ElvisLives.com> wrote:
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-57
> 31398.html?part=rss&tag=5731398&subj=news
People are tripping all over this article.
First off, all we know is (supposedly) Apple and Intel are in talks.
What about? Why does that mean that Apple is going to shift all its
machines to x86 CPU's?
I agree Intel fabbing PPC's is unlikely - that can be tabled.
But I remember an explosion of these kinds of rumors when people heard
Apple was going to be using AMD chips. Yes, it did. It used AMD chips --
IN ITS BASE STATIONS.
Maybe Apple is in talks with Intel to use XScale processors for some new
tablet or phone-type device? Or is going to be using Intel chips for
some other kind of DLDevice?
Maybe they want to use the Itanium (only) in a special version of the
Xserve? Or heck maybe the Xserve line is going all-Itanium. Who knows?
It's a lot easier to do these kinds of transitions on your server line
than on your desktop line.
Maybe they threw up their hands at getting a super-liquid-cooled G5 into
the Powerbooks and are going Pentium-M.
Even those two above scenarios do not mean their entire desktop CPU
lineup is going x86, nor does it mean OS X is being ported to generic
x86, nor does it mean future Macs will run Windows natively.
Sure, IBM hasn't exactly been burning up the world with speed-bumping
the G5, but last I checked, both it and Freescale are making *some*
headway with PPC.
There's sooooo many reasons a transition to an all x86 lineup don't make
sense.
1. You've just gotten your developers to deal with a processor
transition (1994), then a OS transition (2000), and now you're going to
hit them with ANOTHER processor transition (2005)? C'mon, get real.
2. The PowerPC is about to finally take off in a big way -- the fact
that all the next-gen game consoles are using it should tell you
something.... this is not 1999, with Moto in a tailspin going nowhere
fast and Intel moving like lightning. This is 2005, with the G5 making
modest but appreciable gains, and Intel increasingly hitting a brick
wall from (shock) thermal problems.
3. Last I heard, Apple was still touting 64-bit processing for the
consumer desktop, and Altivec. Bye bye to both once they go x86 for the
desktops. Remember, the path of 64-bit is not exactly clear in the x86
space, still, this would be the hugest about-face of all....
4. They will not do it unless they make closed machines. You will not be
able to run x86-OSX on your Dell, nor your Dell's copy of Windows on
your x86-Mac. That's just not Apple. I can't see Apple giving up Open
Firmware and adopting the PC BIOS. So they will probably go back to the
Old World days and use a boot ROM... therefore, IMHO, the huge
advantages so many tout from this transition (cost reductions, use of
PC-space video cards and addons) won't exist.
5. I agree they might gain performance from a switch to Intel, but I
can't see why at this point that performance gain will offset the loss
of users and developers to the architecture switch. Honestly, it needs
to keep being said, but it's not like the G5, stuck at dual 2.7 GHz,
can't do anything consumers want to throw at it....
imout...@mac.com
> > if Apple were to abandon PPC, your Mac would be a first-class OS X box
> > for the next 5 years at least.
>
> ..while all the cutting edge development, of both hardware
? Randy already bought his hardware. There's very little Apple would be
doing to update it for him now, even if they stick with IBM.
Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
Randy's G5 will still remain supported regardless of the CPU/chipset
Apple moves to.
> and
> software, would go into the (gag) new Intel Macs....
Going x86's not that big a deal on the user end, aside from legacy app
compatibility. It's a significant but not immense PITA for developers,
tho, but if developers write to Carbon or Foundation, then get
x-platform support for free and x86 binaries are only a recompile away.
Randy Howard
char...@gmail.com says...
> ELVIS2000 wrote:
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
>
> As seen on News.com TalkBack:
>
> Pretty kewel! And complementary:
>
> Intel Inside/Apple Outside
> Blue Men Dancing/Other Bright Silhouettes Dancing
>
> Wonder what the pundits will call this duo? Aptel? Inpple? Intelapple?
> AI? OSx86?
MacIntRash
iWowICouldHaveHadAnOpteron
iSwitchedThenISwitchedBack
Charles Jo
iNtel
Randy Howard
see...@NOSPAMmac.com says...
> Maybe Apple is in talks with Intel to use XScale processors for some new
> tablet or phone-type device? Or is going to be using Intel chips for
> some other kind of DLDevice?
Smash together an iPod and a PDA, running what? Darwin on the Xscale?
Doubt it.
> Maybe they want to use the Itanium (only) in a special version of the
> Xserve? Or heck maybe the Xserve line is going all-Itanium. Who knows?
Have you been asleep for the last five years or so? Itanic is dead.
It hit the iceberg a while back and was sinking, but then Opteron came
along and torpedoed all the remaining lifeboats. No survivors. HP
is still pretending like there is a market for Itanium, because Intel
is basically paying them to keep it alive, but most of the HP engineers
that were working for it have all been hired by Intel now, not a good
sign.
> It's a lot easier to do these kinds of transitions on your server line
> than on your desktop line.
Especially when you don't have a server line, but one supposed server
which is really a Powermac in a 1U chassis with the video card de-
popped.
> Even those two above scenarios do not mean their entire desktop CPU
> lineup is going x86, nor does it mean OS X is being ported to generic
> x86, nor does it mean future Macs will run Windows natively.
I suspect they'll be more than willing to come up with a hardware
design that has Intel CPUs but conforms to none of the PC standards
for things like SMBIOS, ACPI, boot spec, PCnn, all the lame
"Designed for Microsoft Windows..." requirements, etc. They *might*
try and sell OS X to everybody, but I suspect given their history,
they'll balk at that. That will all boil down to what chipset
decisions they make to go along with the CPU. If they announce
anything next week, that means that they've been running it in the
lab for a long time already. Somebody needs to be digging through
the trashcans in cupertino looking for boxes with "intel inside"
on them. :-)
> There's sooooo many reasons a transition to an all x86 lineup don't make
> sense.
Prediction: Dell is going AMD, and Intel is giving Apple the pricing
of a lifetime to shore up 1/3 of that loss in unit shipments, then will
gradually ratchet them up over time. Dell/AMD's just doing a better job
of keeping it a secret than Apple/Intel.
Source? None, just a suspicion. :-)
> 2. The PowerPC is about to finally take off in a big way -- the fact
> that all the next-gen game consoles are using it should tell you
> something.... this is not 1999, with Moto in a tailspin going nowhere
> fast and Intel moving like lightning. This is 2005, with the G5 making
> modest but appreciable gains, and Intel increasingly hitting a brick
> wall from (shock) thermal problems.
If Apple wanted to speed up their computers they'd look at their
file system code, storage controller hardware and associated drivers
and stop blaming all their performance problems on the clock speed.
Then they'd stop pretending like 256MB was a reasonable minimum
memory configuration for OS X.
Then they'd figure out why gcc 4.0.0 does such a horrible job
optimizing for the G5 and fix that.
SJ is still ticked at IBM though because he got made out to be a dork
after announcing the PPC unobtanium. That means that any crazy
think can happen, and the RDF is on full-charge as we speak.
> 3. Last I heard, Apple was still touting 64-bit processing for the
> consumer desktop, and Altivec. Bye bye to both once they go x86 for the
> desktops.
Wrong. By the time they get there, if they even go, they'll all be 64-
bit in Intel land. And it looks like OS X will still be 32-bit and 64-
bit code duct taped together.
> Remember, the path of 64-bit is not exactly clear in the x86
> space, still, this would be the hugest about-face of all....
Sure it is. Both Intel and AMD have it splashed all over their server,
workstation, desktop and mobile roadmaps.
> 4. They will not do it unless they make closed machines. You will not be
> able to run x86-OSX on your Dell, nor your Dell's copy of Windows on
> your x86-Mac. That's just not Apple.
Yeah.
> I can't see Apple giving up Open Firmware and adopting the PC BIOS.
Good. PC BIOS is stretched almost to the breaking point. That was the
one nice thing about Itanium boxes, but it never took off.
> So they will probably go back to the
> Old World days and use a boot ROM...
If they don't care about running Windows on Apple hardware, or selling
OS Whatever to PC customers, they can do whatever they want in that
regard.
> therefore, IMHO, the huge advantages so many tout from this transition
> (cost reductions, use of PC-space video cards and addons) won't exist.
I don't see how the "so's" and "therefore's" are being strung together.
Too many leaps that don't seem connected.
I can't imagine Apple will build their own chipsets to glue in the Intel
processor, so they'll have to go somewhere else for that, but if so, they
may at least get PCI-e finally, basically for free as a consolation
prize.
> 5. I agree they might gain performance from a switch to Intel, but I
> can't see why at this point that performance gain will offset the loss
> of users and developers to the architecture switch.
Users will get real unhappy if there is a problem getting all the
big apps to run on both platforms. The specialized industry users
running some uber-expensive custom software written for the PPC will
mess their shorts. Most of the "mom and pop" users will just keep
trucking along and probably not even know what the hooplah is about.
> Honestly, it needs to keep being said, but it's not like the G5,
> stuck at dual 2.7 GHz, can't do anything consumers want to throw at it....
Read from a SATA drive as fast as a PC can read from an IDE drive would
be one.
:-)
Seeker1
Timberwoof <timbe...@stimpberawoofm.com> wrote:
> In article <gp72a1h2v682hnb7r...@4ax.com>,
> ELVIS2000 <elvi...@ElvisLives.com> wrote:
>
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-
> > 57
> > 31398.html?part=rss&tag=5731398&subj=news
>
> At this point, this is real speculation.
Strange thing: at the same time MacRumors and CNET (and even the WSJ)
are discussing this rumor, a very *different* rumor of a very
*different* kind is on the front page of ThinkSecret, which has a
reputation for accuracy.
http://www.thinksecret.com/news/0506quasar.html
hile IBM has been facing some difficulties achieving its performance
goals with the PowerPC 970-based processors found in Power Mac G5
systems, the company is driving ahead with plans that will dictate the
PowerPC's direction into the next decade.
[snip]
While not specifically part of eClipz, an UltraLite derivative of the
processor, P6UL, could very well find its way into Macs, albeit not
until 2007 or so when the project is completed.
So, interestingly, both rumors have the same timeframe. Where will Macs
be in 2007? On Intel or on Power6Lite/G6?
Doesn't look like IBM is giving up on the PPC at all, that is why the
switch-to-Intel story is even more strange.
Randy Howard
imout...@mac.com says...
> brianleahy wrote:
> > > if Apple were to abandon PPC, your Mac would be a first-class OS X box
> > > for the next 5 years at least.
> >
> > ..while all the cutting edge development, of both hardware
>
> ? Randy already bought his hardware. There's very little Apple would be
> doing to update it for him now, even if they stick with IBM.
Thanks for cheering me up. :-)
> Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> Randy's G5 will still remain supported regardless of the CPU/chipset
> Apple moves to.
Yet as new features get poured into new operating system releases and
applications, fewer and fewer of them will work on my suddenly-found
boat anchor.
> Going x86's not that big a deal on the user end, aside from legacy app
> compatibility. It's a significant but not immense PITA for developers,
> tho, but if developers write to Carbon or Foundation, then get
> x-platform support for free and x86 binaries are only a recompile away.
Not to mention that the much more finely tuned optimizer for x86-64 will
give an immediate performance boost to a lot of things just by doing
that one step.
imout...@mac.com
> > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > Randy's G5 will still remain supported regardless of the CPU/chipset
> > Apple moves to.
>
> Yet as new features get poured into new operating system releases and
> applications, fewer and fewer of them will work on my suddenly-found
> boat anchor.
How do you figure? Intel doesn't bring anything to the table, other
than PCIe, and Apple is already shipping another Intel solution, AGP
8x, that isn't THAT much worse.
Now, if Apple were to announce going with IBM's CELL architecture this
summer, THEN you would be in boo-hoo-hoo territory, since a $2000 CELL
box compared to your 2003-era technology would be a scary, scary thing
indeed.
> > Going x86's not that big a deal on the user end, aside from legacy app
> > compatibility. It's a significant but not immense PITA for developers,
> > tho, but if developers write to Carbon or Foundation, then get
> > x-platform support for free and x86 binaries are only a recompile away.
>
> Not to mention that the much more finely tuned optimizer for x86-64 will
> give an immediate performance boost to a lot of things just by doing
> that one step.
Yup. More like the x86 is doing so much behind the scenes that the
compiler just gives up and lets the CPU do the optimization, which,
oddly enough, is probably the best approach.
Lars Träger
Randy Howard wrote:
> In article <1117864978.5...@z14g2000cwz.googlegroups.com>,
> char...@gmail.com says...
> > ELVIS2000 wrote:
> > > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
> >
> > As seen on News.com TalkBack:
> >
> > Pretty kewel! And complementary:
> >
> > Intel Inside/Apple Outside
> > Blue Men Dancing/Other Bright Silhouettes Dancing
> >
> > Wonder what the pundits will call this duo? Aptel? Inpple? Intelapple?
> > AI? OSx86?
>
> MacIntRash
>
> iWowICouldHaveHadAnOpteron
iSorryUOnlyGetXeon
Lars T.
Randy Howard
Lars.T...@epost.de says...
>
> > MacIntRash
> >
> > iWowICouldHaveHadAnOpteron
>
> iSorryUOnlyGetXeon
:-)
Randy Howard
imout...@mac.com says...
> Randy Howard wrote:
> > > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > > Randy's G5 will still remain supported regardless of the CPU/chipset
> > > Apple moves to.
> >
> > Yet as new features get poured into new operating system releases and
> > applications, fewer and fewer of them will work on my suddenly-found
> > boat anchor.
>
> How do you figure? Intel doesn't bring anything to the table, other
> than PCIe,
bigger cache, faster clock, dual core now, hyperthreading (yeah, I know
HT is BS but marketing loves it), and compilers that actually optimize.
> and Apple is already shipping another Intel solution, AGP
> 8x, that isn't THAT much worse.
Except the adapters are about a year behind to be kind on the video side,
plus SLI is a big deal if you really want graphics performance. PCI-e
x16 is up there pretty good already, and it's not even broken in good
yet. AGP isn't ever going to get much faster, because nobody (but apple)
cares about it anymore.
> Now, if Apple were to announce going with IBM's CELL architecture this
> summer, THEN you would be in boo-hoo-hoo territory, since a $2000 CELL
> box compared to your 2003-era technology would be a scary, scary thing
> indeed.
Not so scary, i'll just buy something else and give this thing to the
wife, although she isn't all that impressed with it so far. :-)
> Yup. More like the x86 is doing so much behind the scenes that the
> compiler just gives up and lets the CPU do the optimization, which,
> oddly enough, is probably the best approach.
Not really. Turn off the optimizer on CPU-intensive code on x86-64
and watch what happens. The optimizer is doing a lot over there, and
there's actually a bit of an "optimizer war" going on, between the
Intel, MS and gcc compilers.
I've run into a few cases where gcc 4.0.0 on the PPC actually slows
code down when you turn the optimizer up. Sad really.
John Slade
"Paul McCahan" <mcahann@-NOSPAM-cs.com> wrote in message
news:m1boe.277$BK2....@news.uswest.net...
I don't know, the story seems pretty reliable to me. However if it is
true, Apple should just start making an OS for PCs. I mean if they have such
a great product, it should really give Microsoft fits through competition.
Apple will start off making their own chipset for Intel CPUs. It will be
sort of futile. It will be done for nothing but ego. Just to say that the
Mac isn't a PC but at it's heart it really will be. They will eventually
phase that out and become an OS and clone maker.
Who is to blame for the PPC dying on the Mac platform? I think Apple
bears most of the blame but not all. IBM clearly put the video game consoles
over Apple's needs when they had the chip shortage. IBM was also slow to
develop the PPCs for Apple. They had no problems advancing the versions for
video game consoles. IBM was making CPUs for a company that competed with
their own server line. That's probably why IBM wasn't all that eager to
develop the PPC for Macs any further. It seems like they just screwed Apple.
But Apple was asking for it. Apple should have developed their own CPU when
they dumped the Motorola 680x0 line.
Anyway if this switch is just a rumor, then all of the above is
meaningless bullshit until Apple finally does switch to the x86.
John
John Slade
"Seeker1" <see...@NOSPAMmac.com> wrote in message
news:seeker1-A7BD11...@comcast.dca.giganews.com...
Not strange at all. How long has the PPC been stagnant? No real speed
increase in how many years? Then they seem to overclock some G5s for a .3Ghz
boost. What stupidity. IBM can't be depended upon to produce faster chips in
sufficient quantities. Intel on the other hand has no problems like this.
They are moving ahead in leaps and bounds compared to IBM and are producing
those more advanced chips in high quantities.
John
John Slade
"Charles Jo" <char...@gmail.com> wrote in message
news:1117864978.5...@z14g2000cwz.googlegroups.com...
> ELVIS2000 wrote:
> >
http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
>
> As seen on News.com TalkBack:
>
> Pretty kewel! And complementary:
>
> Intel Inside/Apple Outside
> Blue Men Dancing/Other Bright Silhouettes Dancing
>
> Wonder what the pundits will call this duo? Aptel? Inpple? Intelapple?
> AI? OSx86?
Since they call the Windows-Intel combination "Wintel", I suspect they
will use the same naming convention for OS X and Intel. I like the term
"Oxtail" mainly because they taste so good. But I think "X-Tel" will be the
winner.
John
imout...@mac.com
> In article <1117867974.3...@o13g2000cwo.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > > > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > > > Randy's G5 will still remain supported regardless of the CPU/chipset
> > > > Apple moves to.
> > >
> > > Yet as new features get poured into new operating system releases and
> > > applications, fewer and fewer of them will work on my suddenly-found
> > > boat anchor.
> >
> > How do you figure? Intel doesn't bring anything to the table, other
> > than PCIe,
>
> bigger cache,
IBM could do this, but it has the monster FSB bandwidth (well, compared
to x86) so doesn't need to.
>faster clock
Truly Irrelevant, cf. AMD vs. x86.
>dual core now
Fake dual core with separate L2 caches and a shared FSB to the memory
controller. Yuck!
>, hyperthreading (yeah, I know HT is BS but marketing loves it)
Well, a ~10% gain here & there isn't that much BS.
> and compilers that actually optimize.
True. But not really a deciding thing.
> > and Apple is already shipping another Intel solution, AGP
> > 8x, that isn't THAT much worse.
>
> Except the adapters are about a year behind to be kind on the video side,
huh? Apple is shipping 6800U, the best prosumer graphics card you can
get. Apple doesn't have 'workstation' spec cards, but IME the
top-of-the-line NVIDIA has most if not all of the quadro features
turned on, at least it did in GeF4 days, dunno about now.
> plus SLI is a big deal if you really want graphics performance.
Not that big a deal, really. SLI helps with fill-rate, but Apple
doesn't sell G5s to play games on so this is largely irrelevant to
them.
> PCI-e x16 is up there pretty good already, and it's not even broken in good
> yet. AGP isn't ever going to get much faster, because nobody (but apple)
> cares about it anymore.
AGP really really blows, but PCIe sucks compared to a real graphics
architecture like the PS3 or xbox2.
For games, there is no great benchmarkabe difference between AGP and
PCIe, since the CPU <-> GPU communication is through an asynchronous
one-way FIFO, and the GPU can't drain the FIFO faster than the CPU can
fill it anyway.
In the future PCIe has the promise of offloading more processing onto
the GPU, eg. with CoreImage, which is why I haven't felt the need to
buy a G5 yet, but for even these tasks where PCIe is more useful I
don't think there will be a significant performance delta between AGP
8x and PCIe.
> > Now, if Apple were to announce going with IBM's CELL architecture this
> > summer, THEN you would be in boo-hoo-hoo territory, since a $2000 CELL
> > box compared to your 2003-era technology would be a scary, scary thing
> > indeed.
>
> Not so scary, i'll just buy something else and give this thing to the
> wife, although she isn't all that impressed with it so far. :-)
Well, why the hell did you blow $2000 on something you have no need for
other than benchmarking and generally complaining about sundry
failings???
> > Yup. More like the x86 is doing so much behind the scenes that the
> > compiler just gives up and lets the CPU do the optimization, which,
> > oddly enough, is probably the best approach.
>
> Not really. Turn off the optimizer on CPU-intensive code on x86-64
> and watch what happens. The optimizer is doing a lot over there, and
> there's actually a bit of an "optimizer war" going on, between the
> Intel, MS and gcc compilers.
yeah, I know. Clearly there are some things the compiler can do. I just
think it can go too far and gum things up since what's happening inside
the CPU is so much different from the crappy x86 compatibility
interface.
> I've run into a few cases where gcc 4.0.0 on the PPC actually slows
> code down when you turn the optimizer up. Sad really.
Well, nobody really pushes gcc PPC except Apple apparently, and
2001-2004 Apple was under one bitch of a hiring freeze, sadly limiting
its ability to expand such critical teams as the compiler team.
Odd that. $5B in the bank and they can't hire people.
Randy Howard
> Randy Howard wrote:
> IBM could do this, but it has the monster FSB bandwidth (well, compared
> to x86) so doesn't need to.
Maybe IBM should put a couple good compiler guys to work on gcc 4.0.1
then. :-)
> >faster clock
>
> Truly Irrelevant, cf. AMD vs. x86.
In straight number terms yeah. In Terms of being firewalled with little
hope of going faster...
> >dual core now
>
> Fake dual core with separate L2 caches and a shared FSB to the memory
> controller. Yuck!
Better than the above...
> >, hyperthreading (yeah, I know HT is BS but marketing loves it)
>
> Well, a ~10% gain here & there isn't that much BS.
Except it doesn't work out that way in practice, only in very customized
code just to make HT look good. For heavy I/O, like a typical server
load, it actually slows things down. Think bottleneck on the freeway,
then look at a HT block diagram. Same thing.
Disk I/O, network throughput, memory blasting, you name it, if it's
I/o bound, HT puts the breaks on things badly. Turning off HT in
setup on servers that support will usually buy you a pretty nice
performance increase. Try it sometime.
> > and compilers that actually optimize.
>
> True. But not really a deciding thing.
It just makes the other issues worse than they would be otherwise.
> > PCI-e x16 is up there pretty good already, and it's not even broken in good
> > yet. AGP isn't ever going to get much faster, because nobody (but apple)
> > cares about it anymore.
>
> AGP really really blows, but PCIe sucks compared to a real graphics
> architecture like the PS3 or xbox2.
I'm not a graphics dude at all, and it sounds like you are, so I take
your word for it, although I have seen some early PCI-e RAID
controllers do some amazing things, like 2X the throughput on the
exact same set of hard drives as the fastest previous PCI-X 133 card
could do. Massive performance improvement. First time I saw it,
I thought something was wrong with the clock chip in the system,
too good to believe.
Especially since the throughput on PCI-X controllers wasn't saturating
the bus. The design simply makes bulk transfers happen much more
efficiently on PCI-e. I was hoping that would carry over to video, but
maybe not.
> > Not so scary, i'll just buy something else and give this thing to the
> > wife, although she isn't all that impressed with it so far. :-)
>
> Well, why the hell did you blow $2000 on something you have no need for
> other than benchmarking and generally complaining about sundry
> failings???
I meant that if something better came along, it wouldn't be a the "boo
hoo" scenario you described, despite it costing more than $2K. Cool out.
If nothing else, I like having a more recent box around to make sure no
endian bugs are hiding anywhere.
> > I've run into a few cases where gcc 4.0.0 on the PPC actually slows
> > code down when you turn the optimizer up. Sad really.
>
> Well, nobody really pushes gcc PPC except Apple apparently,
Other than IBM, who would? Frankly I doubt Apple has the right kind of
staff around for that anyway, whereas IBM probably has them by the
truckload.
> 2001-2004 Apple was under one bitch of a hiring freeze, sadly limiting
> its ability to expand such critical teams as the compiler team.
>
> Odd that. $5B in the bank and they can't hire people.
Forward planning for the big crash in sales from the switch off of PPC.
:-)
imout...@mac.com
> In article <1117875143....@z14g2000cwz.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > IBM could do this, but it has the monster FSB bandwidth (well, compared
> > to x86) so doesn't need to.
>
> Maybe IBM should put a couple good compiler guys to work on gcc 4.0.1
> then. :-)
ya got that right. Moto used to do this with their stuff for Apple.
> > >faster clock
> >
> > Truly Irrelevant, cf. AMD vs. x86.
>
> In straight number terms yeah. In Terms of being firewalled with little
> hope of going faster...
? IBM is assumedly upclocking to 3.2Ghz by the end of the year (xbox2).
Not bad.
Talk about being stuck, two years ago I paid ~$220 for my 2.6Ghz
Northwood B, now this pricepoint is occupied by a 3.2Ghz part. 23%
upclocking in 24 months, so much for Moore's Law (I know Moore was
talking about capacities not speeds per se). In the interim, Apple's
mid-grade has gone from 1.8 to 2.3, a 28% speedup. Granted, that's not
much to write home about either, but this generation has been pretty
screwy really, and at least Apple has pushed the FSB bus up across the
line, while Intel has been sitting at 200MhzxQDR for a very long time
now.
> > >dual core now
> >
> > Fake dual core with separate L2 caches and a shared FSB to the memory
> > controller. Yuck!
>
> Better than the above...
Given that a dualcore 3.2 Pentium D costs $100 more than two single 3.2
P4s, I REALLY don't see the advantage here. Granted, Xeon pricing is a
ripoff but that doesn't affect IBM directly.
> > >, hyperthreading (yeah, I know HT is BS but marketing loves it)
> >
> > Well, a ~10% gain here & there isn't that much BS.
>
> Except it doesn't work out that way in practice, only in very customized
> code just to make HT look good. For heavy I/O, like a typical server
> load, it actually slows things down. Think bottleneck on the freeway,
> then look at a HT block diagram. Same thing.
Dunno. There's real-world measurements where HT works well:
http://www.hardwareanalysis.com/content/article/1797.1/
Granted HT can slow things down, too.
> Disk I/O, network throughput, memory blasting, you name it, if it's
> I/o bound, HT puts the breaks on things badly. Turning off HT in
> setup on servers that support will usually buy you a pretty nice
> performance increase. Try it sometime.
Don't run an high-load x86 server, don't have to.
> > > and compilers that actually optimize.
> >
> > True. But not really a deciding thing.
>
> It just makes the other issues worse than they would be otherwise.
Significant, yes. But software is in fact the one thing completely in
Apple's power to fix.
> > > PCI-e x16 is up there pretty good already, and it's not even broken in good
> > > yet. AGP isn't ever going to get much faster, because nobody (but apple)
> > > cares about it anymore.
> >
> > AGP really really blows, but PCIe sucks compared to a real graphics
> > architecture like the PS3 or xbox2.
>
> I'm not a graphics dude at all, and it sounds like you are, so I take
> your word for it, although I have seen some early PCI-e RAID
> controllers do some amazing things, like 2X the throughput on the
> exact same set of hard drives as the fastest previous PCI-X 133 card
> could do. Massive performance improvement. First time I saw it,
> I thought something was wrong with the clock chip in the system,
> too good to believe.
Sure. PCI's got all kinds of nastiness with bridges and while I know
jack about storage one would think the two-way transactional nature of
it would benefit from a better interface.
> Especially since the throughput on PCI-X controllers wasn't saturating
> the bus. The design simply makes bulk transfers happen much more
> efficiently on PCI-e. I was hoping that would carry over to video, but
> maybe not.
It's all about how the games that push the graphics card dump megabytes
of command traffic onto the drivers. The drivers are coalescing these
commands behind the app's back and pushing them off to the card
asynchronously. When (not if) the GPU falls too far behind in
processing this command traffic on its end the drivers block further
graphics calls (this will happen when the app gets two frames ahead of
the current frame).
This asynchronous batch processing means the bottleneck is the GPU, not
the interface. Same thing was found with AGP 4x and 8x, no big diff in
the benchmarks.
PCIe has the advantage of much better two-way latencies (AGP seems to
suck at this for some reason) and of course SLI or just dual-head
capability, which is cool if you want to do a four-way environment too
with dual-head video cards.
> > > I've run into a few cases where gcc 4.0.0 on the PPC actually slows
> > > code down when you turn the optimizer up. Sad really.
> >
> > Well, nobody really pushes gcc PPC except Apple apparently,
>
> Other than IBM, who would? Frankly I doubt Apple has the right kind of
> staff around for that anyway, whereas IBM probably has them by the
> truckload.
My knowledge on this is running down its half-life, but at one point in
the not-so-distant past a single well-timed shotgun blast could take
out the entire Apple compiler and low-level perf team members (the
compiler team is heavily leveraged onto gcc/gdb of course, all they do
is a pretty GUI to GNU tools really, and the few low-level guys are all
NeXT veterans).
Cameron Kaiser
>>this isn't going to happen, it's a red herring
>I hope you're right.
So do I. This is horrible news. At least back in the 68K days, we knew we
were getting a *better* processor as an upgrade when the PPC switch came.
--
Cameron Kaiser * cka...@floodgap.com * posting with a Commodore 128
personal page: http://www.armory.com/%7Espectre/
** Computer Workshops: games, productivity software and more for C64/128! **
** http://www.armory.com/%7Espectre/cwi/ **
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <gp72a1h2v682hnb7r...@4ax.com>, elvis2000
> @ElvisLives.com says...
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-
> > 5731398.html?part=rss&tag=5731398&subj=news
>
> Oh boy, did I ever pick the wrong time to buy a Mac. :-(
>
> And Intel bagged them, even though AMD clearly has a better product
> offering. Not encouraging. I wonder what the last version of OS X
> will be to support my sudden boat-anchor?
That is all true - assuming that you're silly enough to believe a
nonsense rumor that has reappeared every couple of years for at least 20
years - and just isn't going to happen.
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <1117867974.3...@o13g2000cwo.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > > > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > > > Randy's G5 will still remain supported regardless of the CPU/chipset
> > > > Apple moves to.
> > >
> > > Yet as new features get poured into new operating system releases and
> > > applications, fewer and fewer of them will work on my suddenly-found
> > > boat anchor.
> >
> > How do you figure? Intel doesn't bring anything to the table, other
> > than PCIe,
>
> bigger cache, faster clock, dual core now, hyperthreading (yeah, I know
> HT is BS but marketing loves it), and compilers that actually optimize.
PPC doesn't need a bigger cache as much as Intel given the immense
memory bandwidth of the 1.35 GHz bus.
PPC is faster clock speed than AMD - which you insist is better than
Intel.
Dual core isn't far away. Meanwhile, Apple uses two separate chips -
which achieves nearly all the benefit.
Hyperthreading doesn't add much. It certainly doesn't add anywhere near
as much as the dual chips in PowerMacs. I'm not interested in marketing
gimmicks.
Compilers. Something of an issue, but better than in the past. And since
high end Macs manage to perform pretty well vs. high end PCs, Macs are
doing OK in spite of compilers that aren't as optimized as PCs.
>
> > and Apple is already shipping another Intel solution, AGP
> > 8x, that isn't THAT much worse.
>
> Except the adapters are about a year behind to be kind on the video side,
> plus SLI is a big deal if you really want graphics performance. PCI-e
> x16 is up there pretty good already, and it's not even broken in good
> yet. AGP isn't ever going to get much faster, because nobody (but apple)
> cares about it anymore.
Switching their processor chip to Intel wouldn't have any effect on
that. It's a matter of writing and optimizing drivers, creating new
motherboards, and testing everything. You still need to do that
regardless of what CPU you use.
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <1117875143....@z14g2000cwz.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > IBM could do this, but it has the monster FSB bandwidth (well, compared
> > to x86) so doesn't need to.
>
> Maybe IBM should put a couple good compiler guys to work on gcc 4.0.1
> then. :-)
Nice attempt to change the topic.
>
> > >faster clock
> >
> > Truly Irrelevant, cf. AMD vs. x86.
>
> In straight number terms yeah. In Terms of being firewalled with little
> hope of going faster...
Really? In the past 2 years, Intel has gone from 3.2 to 3.8 GHz. IBM's
G5 has gone from 2.0 to 2.7 GHz. Do the math.
>
> > >dual core now
> >
> > Fake dual core with separate L2 caches and a shared FSB to the memory
> > controller. Yuck!
>
> Better than the above...
Not better than two separate processors with 1.35 GHz memory bus.
>
> > >, hyperthreading (yeah, I know HT is BS but marketing loves it)
> >
> > Well, a ~10% gain here & there isn't that much BS.
>
> Except it doesn't work out that way in practice, only in very customized
> code just to make HT look good. For heavy I/O, like a typical server
> load, it actually slows things down. Think bottleneck on the freeway,
> then look at a HT block diagram. Same thing.
>
> Disk I/O, network throughput, memory blasting, you name it, if it's
> I/o bound, HT puts the breaks on things badly. Turning off HT in
> setup on servers that support will usually buy you a pretty nice
> performance increase. Try it sometime.
So then why are you bringing up HT as an advantage?
TravelinMan
"John Slade" <hitm...@pacbell.net> wrote:
> Not strange at all. How long has the PPC been stagnant? No real speed
> increase in how many years?
Well, during the time the G5 has gone from 2.0 to 2.7 GHz, the P4 has
gone from 3.2 to 3.8 GHz. Do the math.
C Lund
Seeker1 <see...@NOSPAMmac.com> wrote:
> I agree Intel fabbing PPC's is unlikely - that can be tabled.
FWIW, I think an Intel PPC is more likely than a x86 Mac.
Actually, an Intel PPC might not be a bad idea at all. IBM and
Motorola have both had trouble delivering the bulk Apple needs. I
can't remember hearing about Intel having problems like that.
But we'll see on Monday. B)
--
C Lund, www.notam02.no/~clund
Andrew J. Brehm
> Maybe Apple is in talks with Intel to use XScale processors for some new
> tablet or phone-type device? Or is going to be using Intel chips for
> some other kind of DLDevice?
That is likely.
> Maybe they want to use the Itanium (only) in a special version of the
> Xserve? Or heck maybe the Xserve line is going all-Itanium. Who knows?
> It's a lot easier to do these kinds of transitions on your server line
> than on your desktop line.
That makes a lot of sense.
> Maybe they threw up their hands at getting a super-liquid-cooled G5 into
> the Powerbooks and are going Pentium-M.
I don't think the Pentium-M has that many advantages over G3s and G4s
and their derivatives.
--
Andrew J. Brehm
Marx Brothers Fan
PowerPC/Macintosh User
Supporter of Chicken Sandwiches
TravelinMan
C Lund <cl...@notam02SPAMBLOCK.no> wrote:
> In article <seeker1-D36243...@comcast.dca.giganews.com>,
> Seeker1 <see...@NOSPAMmac.com> wrote:
>
> > I agree Intel fabbing PPC's is unlikely - that can be tabled.
>
> FWIW, I think an Intel PPC is more likely than a x86 Mac.
>
> Actually, an Intel PPC might not be a bad idea at all. IBM and
> Motorola have both had trouble delivering the bulk Apple needs. I
> can't remember hearing about Intel having problems like that.
Intel has had plenty of problems with the 90 nm process, as well. I
doubt if an Intel PPC would be significantly faster than IBM's.
Chad Irby
"John Slade" <hitm...@pacbell.net> wrote:
> How long has the PPC been stagnant? No real speed
> increase in how many years?
Better watch that "stagnant" bit - when you compare the PPC to the Intel
chips, the PPC is increasing clock at a higher rate, and getting better
preformance increases from that clock increase.
It took Intel two and a half years to go from 3 GHz to 3.8 GHz, and took
IBM only one and a half years to go from 2.0 to 2.7 GHz with the G5.
--
I don't have a lifestyle.
I have a lifeCSS.
Lars Träger
John Slade wrote:
> "Charles Jo" <char...@gmail.com> wrote in message
> news:1117864978.5...@z14g2000cwz.googlegroups.com...
> > ELVIS2000 wrote:
> > >
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-5731398.html?part=rss&tag=5731398&subj=news
> >
> >
> > Wonder what the pundits will call this duo? Aptel? Inpple? Intelapple?
> > AI? OSx86?
>
> Since they call the Windows-Intel combination "Wintel", I suspect they
> will use the same naming convention for OS X and Intel. I like the term
> "Oxtail" mainly because they taste so good. But I think "X-Tel" will be the
> winner.
I'll vote for BullShit.
Lars T.
Chad Irby
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <1117867974.3...@o13g2000cwo.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > > > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > > > Randy's G5 will still remain supported regardless of the CPU/chipset
> > > > Apple moves to.
> > >
> > > Yet as new features get poured into new operating system releases and
> > > applications, fewer and fewer of them will work on my suddenly-found
> > > boat anchor.
> >
> > How do you figure? Intel doesn't bring anything to the table, other
> > than PCIe,
>
> bigger cache,
Adding cache isn't exactly hard, you know. Intel *needs* the bigger
cache because of all of their other issues coming from the higher clock
speeds.
> faster clock,
...but, as you know, clock speed is turning out to be a moderately
non-issue sort of issue. The AMD chips are beating the Intel chips,
while using much lower clock speeds.
> dual core now,
This isn't the issue most people think, and dual core G5s are probably
about two days off.
> hyperthreading (yeah, I know
> HT is BS but marketing loves it),
So all Apple has to do is use some similar phrase for the sorts of
things they've already been doing..
> and compilers that actually optimize.
...for old benchmarks that don't have anything to do with anything.
With all of those "advantages" you cite above, Intel isn't getting any
real performance improvements...
Meanwhile, Intel clock speeds are *really* stagnant, when compared to
Chad Irby
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <1117875143....@z14g2000cwz.googlegroups.com>,
> imout...@mac.com says...
> > Randy Howard wrote:
> > IBM could do this, but it has the monster FSB bandwidth (well, compared
> > to x86) so doesn't need to.
>
> Maybe IBM should put a couple good compiler guys to work on gcc 4.0.1
> then. :-)
>
> > >faster clock
> >
> > Truly Irrelevant, cf. AMD vs. x86.
>
> In straight number terms yeah. In Terms of being firewalled with little
> hope of going faster...
In those terms, Intel is *screwed*.
It's taken them two and a half years from the introduction of the 3.0
GHz chips to get to 3.8 GHz. That's an increase of 27% in 30 months.
The G5 hit 2 GHz one and a half years ago, and took a year and a half to
go to 2.7 GHz. That's an increase of 35% in 18 months. In other words,
the G5 is increasing clock speed over twice as fast as the top-clocked
Intel chips...
Chad Irby
TravelinMan <Now...@spamfree.com> wrote:
...but there could be a *lot* more high-end G5 chips available, at lower
prices, and the low-end chips like the G4 could get a real price
advantage...
TravelinMan
Chad Irby <ci...@cfl.rr.com> wrote:
Maybe.
I don't think it's that simple. IBM has plenty of fab capacity. There's
apparently a limitation in the process that prevents them from going any
faster. Note that when they actually HAVE released a faster chip,
availability isn't that much of a problem.
bob
"TravelinMan" <Now...@spamfree.com> wrote in message news:Nowhere-
(snip)
> > > How do you figure? Intel doesn't bring anything to the table, other
> > > than PCIe,
> >
> > bigger cache, faster clock, dual core now, hyperthreading (yeah, I know
> > HT is BS but marketing loves it), and compilers that actually optimize.
>
> PPC doesn't need a bigger cache as much as Intel given the immense
> memory bandwidth of the 1.35 GHz bus.
I don't agree entirely with this statement. Generally speaking, RISC
machines require higher bandwidths and larger caches than CISC machines. An
equivalent RISC program is usually larger that a CISC. If a cache is smaller
then, obviously, less of the program resides in cache. If the program is
larger then there will be more cache misses.
No doubt bus bandwidth is important factor, but much of the performance
impact is the latency to get the first word out of memory. Especially on
cache reads misses since the processor may be stalled waiting for data. So
even a fast bus has to wait on the memory controller.
Seeker1
> > tablet or phone-type device? Or is going to be using Intel chips for
> > some other kind of DLDevice?
>
> Doubt it.
I don't see why this scenario I'm suggesting is *less* likely. I believe
the rumors are true that Apple and Intel have been in talks, I just
think the CNET scenario is less likely than something else. I believe
every scenario I've presented (using Intel chips for a new
device/DLD/peripheral, using *some* form of x86 for servers or maybe a
new Xstation workstation ONLY.... are more likely than Apple moving its
ENTIRE line of *CPUs* over to x86.)
> > It's a lot easier to do these kinds of transitions on your server line
> > than on your desktop line.
>
> Especially when you don't have a server line, but one supposed server
> which is really a Powermac in a 1U chassis with the video card de-
> popped.
Dude, I'm tired of this discussion. I realize the Xserve is not some
peoples' idea of a perfect server (because of certain redundancy
features it lacks), but it *is* a server. Maybe not the server you would
want to run the nation's defenses on, but it is still a server, which is
good for a certain market where 100% uptime is not life or death and can
live with 95% uptime.
I still maintain that even if the CNET article is essentially correct,
it could still be correct, and Apple could only be moving PART of its
lineup (servers, laptops, new high-end Xstation workstations) to x86,
and possibly keeping PPC alive in other sectors. Yes, it requires
supporting two kinds of OSX at once, a PPC-OSX and an x86-OSX, but I
think they can do the trick; NeXTStep did it.
> If Apple wanted to speed up their computers they'd look at their
> file system code, storage controller hardware and associated drivers
> and stop blaming all their performance problems on the clock speed.
> Then they'd stop pretending like 256MB was a reasonable minimum
> memory configuration for OS X.
>
> Then they'd figure out why gcc 4.0.0 does such a horrible job
> optimizing for the G5 and fix that.
So you agree with me - the two things people tout for this "switch"
(price reduction, performance) they could get in OTHER ways.
1. they can eke out more performance without going to x86. we agree.
2. I also don't see prices dropping on Macs that dramatically, either.
Cheaper CPUs may mean $100-200 cheaper computers, but it's not like the
PowerMac is going to go from $2999 to $999 just by going x86.
> SJ is still ticked at IBM though because he got made out to be a dork
> after announcing the PPC unobtanium. That means that any crazy
> think can happen, and the RDF is on full-charge as we speak.
This switch would have made far more sense around 1999-2000 when
Motorola was in full stall. IBM is stumbling and so sure they embarassed
the Jobster with his 3GHz prophecy, but they, unlike Moto, appear to
have a roadmap and still be on their feet. That's why breaking off from
Big Blue at this point doesn't make sense.
Unless, again, there are things behind the scenes we don't know about. I
still do find it interesting that ThinKSecret, known for its accuracy,
is predicting Apple staying with IBM till 2007 at least, which is
exactly contrary to all the other rumors sites.
> Wrong. By the time they get there, if they even go, they'll all be 64-
> bit in Intel land.
Sure. But the x86 world will be going in two different directions.
Opteron makes a lot more sense than any of Intel's x86-64 options if
Apple wants to stay in 64-bit-land. That's why an *Intel* switch really
makes *less* sense than AMD.
And you didn't even deal with Altivec... yes developers developers
developers aren't optimizing for it, but Apple's never abandoned it
yet... they insisted Big Blue staple it to the G5.
> I can't imagine Apple will build their own chipsets to glue in the Intel
> processor, so they'll have to go somewhere else for that, but if so, they
> may at least get PCI-e finally, basically for free as a consolation
> prize.
But I think we are all in agreement that Apple can get a lot of the
things you and HM want (PCIe, better disk controllers, more optimized
code/Filesystem, etc.) without switching the CPU. So *what* are the
advantages, other than, yea, P4 is at 3.8 GHz, and stuck?
Remember what I just said -- if they are going to be using a boot ROM,
they will *not* be using generic found-in-the-bin x86 mobo's from Taiwan
in their x86 Macs. So then there's not even a guarantee that this
decision means that Apple users get all the tech out there in the PC
space...
> Most of the "mom and pop" users will just keep
> trucking along and probably not even know what the hooplah is about.
I thought they managed the 0x0 to PPC transition pretty well, but I'm
really unclear on why some people think this one (PPC-x86) will go more
smoothly, considering the processors are MORE dissimilar.
Yes, I know NeXTStep was x86, there is a secret "Star Wars" version of
OS X probably running on x86 already, blah blah, the bottom line is you
know developers are going to take a dump if this announcement is true
EXACTLY as CNET stated it. It ain't gonna be so easy.
bob
"Cameron Kaiser" <cka...@floodgap.com> wrote in message
news:42a192b5$0$16185$bb4e...@newscene.com...
> brianleahy <brianlea...@nomx.macosx.com> writes:
>
> >>this isn't going to happen, it's a red herring
>
> >I hope you're right.
>
> So do I. This is horrible news. At least back in the 68K days, we knew we
> were getting a *better* processor as an upgrade when the PPC switch came.
Depends on your definition of *better*. If it really was *better* then this
thread wouldn't exist. I always was under the impression that most MAC users
hated MS Windows more than Intel processors. I wonder what sales would be
the if the choices were PPC and Windows or Intel and OSX(or pick your Apple
OS)?
Tim Smith
> tablet or phone-type device? Or is going to be using Intel chips for
> some other kind of DLDevice?
How about iPods? The processor currently used is kind of slow--maybe
they need more for a video iPod?
--
--Tim Smith
Tim Smith
How about a PCI card for Macs that contains an Intel processor, and
bundled software to run Windows on it, with good integration with OS X,
for people who find VPC too slow?
There were products like this back in the 68k days from third parties.
--
--Tim Smith
Timberwoof
"John Slade" <hitm...@pacbell.net> wrote:
> "Paul McCahan" <mcahann@-NOSPAM-cs.com> wrote in message
> news:m1boe.277$BK2....@news.uswest.net...
> > In article <gp72a1h2v682hnb7r...@4ax.com>,
> > ELVIS2000 <elvi...@ElvisLives.com> wrote:
> >
> > >
> http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-57
> > > 31398.html?part=rss&tag=5731398&subj=news
> >
> > this isn't going to happen, it's a red herring designed to throw people
> > off. steve p doesn't speak until tuesday, so the article is wrong from
> > the beginning.
>
> I don't know, the story seems pretty reliable to me.
Reliable? It's all speculation and attributions to unnamed sources.
> However if it is
> true, Apple should just start making an OS for PCs. I mean if they have such
> a great product, it should really give Microsoft fits through competition.
> Apple will start off making their own chipset for Intel CPUs. It will be
> sort of futile. It will be done for nothing but ego. Just to say that the
> Mac isn't a PC but at it's heart it really will be. They will eventually
> phase that out and become an OS and clone maker.
>
> Who is to blame for the PPC dying on the Mac platform? I think Apple
> bears most of the blame but not all. IBM clearly put the video game consoles
> over Apple's needs when they had the chip shortage. IBM was also slow to
> develop the PPCs for Apple. They had no problems advancing the versions for
> video game consoles. IBM was making CPUs for a company that competed with
> their own server line. That's probably why IBM wasn't all that eager to
> develop the PPC for Macs any further. It seems like they just screwed Apple.
> But Apple was asking for it. Apple should have developed their own CPU when
> they dumped the Motorola 680x0 line.
>
> Anyway if this switch is just a rumor, then all of the above is
> meaningless bullshit until Apple finally does switch to the x86.
It's all meaningless bullshit.
--
Timberwoof <me at timberwoof dot com> http://www.timberwoof.com
If Macintosh is a luxury cruise ship,
then Linux is a freighter with wood paneling in the officers' quarters.
Randy Howard
Now...@spamfree.com says...
> > bigger cache, faster clock, dual core now, hyperthreading (yeah, I know
> > HT is BS but marketing loves it), and compilers that actually optimize.
>
> PPC doesn't need a bigger cache as much as Intel given the immense
> memory bandwidth of the 1.35 GHz bus.
PPC gets exactly the same memory throughput out of DDR 3200 as does
the Intel. They both get exactly 3200MB/s, as expected, with dual
processors. Problem is, with a Xeon, and a single processor you
can also get that same 3200, but PPC gets about 500MB/s less than
that, for no good reason.
Cache misses are still hugely expensive, on both. Given that PPC uses
more instructions to do the same work, cache is arguably more important.
An Opteron dual processor box will get about 5900MB/s using the
same ram due to hypertransport. When you go to 4-way, it's
almost 10GB/s. Nothing scales memory throughput w/extra processors
the way opteron does today.
> PPC is faster clock speed than AMD - which you insist is better than
> Intel.
It is. clock speed as a specific number isn't as important as it
is relative to its competition.
> Hyperthreading doesn't add much. It certainly doesn't add anywhere near
> as much as the dual chips in PowerMacs. I'm not interested in marketing
> gimmicks.
Dual, quad and 8-way PC boxes do exactly the same thing. But you are
correct that HT is basically a marketing gimmick, but it did get
developers that wouldn't have otherwise thinking about parallelism
in their programs a few years earlier, which is a good thing long
term.
> > Except the adapters are about a year behind to be kind on the video side,
> > plus SLI is a big deal if you really want graphics performance. PCI-e
> > x16 is up there pretty good already, and it's not even broken in good
> > yet. AGP isn't ever going to get much faster, because nobody (but apple)
> > cares about it anymore.
>
> Switching their processor chip to Intel wouldn't have any effect on
> that. It's a matter of writing and optimizing drivers, creating new
> motherboards, and testing everything. You still need to do that
> regardless of what CPU you use.
Oh really? Where is the Mac w/PCI-express? Pretty hard to write a
driver for a bus interface that doesn't exist.
Randy Howard
says...
> > memory bandwidth of the 1.35 GHz bus.
>
> I don't agree entirely with this statement. Generally speaking, RISC
> machines require higher bandwidths and larger caches than CISC machines. An
> equivalent RISC program is usually larger that a CISC. If a cache is smaller
> then, obviously, less of the program resides in cache. If the program is
> larger then there will be more cache misses.
Exactly right.
Chad Irby
<reply_in_group-2A3...@news1.west.earthlink.net>,
Tim Smith <reply_i...@mouse-potato.com> wrote:
That's been mentioned over the last few weeks, and is one of the
frontrunners in the "what the heck is going on" sweepstakes.
Bootstrap Bill
"Chad Irby" <ci...@cfl.rr.com> wrote in message
news:cirby-CA3AB3....@news-server1.tampabay.rr.com...
Apple should create something similar to the LifeDrive.
Bootstrap Bill
"Tim Smith" <reply_i...@mouse-potato.com> wrote in message
news:reply_in_group-B76...@news1.west.earthlink.net...
How about a way for Windows XP to run concurrently with the new Intel
version of OS X? Apple could license XP and integrate it into future
versions of OSX. End users would probably need to buy an XP license, but it
would be there for those who want it.
The biggest advantage - no need for companies to support two versions of the
same program. XP users would be able to run OSX programs and vice versa.
Microsoft still makes their $100 per head for an XP license and so does
Apple.
Steve Hix
Paul McCahan <mcahann@-NOSPAM-cs.com> wrote:
> In article <gp72a1h2v682hnb7r...@4ax.com>,
> ELVIS2000 <elvi...@ElvisLives.com> wrote:
>
> > http://news.com.com/Apple+to+ditch+IBM%2C+switch+to+Intel+chips/2100-1006_3-
> > 57
> > 31398.html?part=rss&tag=5731398&subj=news
>
> this isn't going to happen, it's a red herring designed to throw people
> off. steve p doesn't speak until tuesday, so the article is wrong from
> the beginning.
Anyone willing to bet that it's an attempt by someone to manipulate
Apple's stock price for a day or two?
TravelinMan
<reply_in_group-B76...@news1.west.earthlink.net>,
Tim Smith <reply_i...@mouse-potato.com> wrote:
I used to use one of those cards.
Apple also had one model which had a DOS card standard.
Timberwoof
"Bootstrap Bill" <william...@gmail.com> wrote:
> "Tim Smith" <reply_i...@mouse-potato.com> wrote in message
> news:reply_in_group-B76...@news1.west.earthlink.net...
> >
> > How about a PCI card for Macs that contains an Intel processor, and
> > bundled software to run Windows on it, with good integration with OS X,
> > for people who find VPC too slow?
> >
> > There were products like this back in the 68k days from third parties.
> >
> Mac Charlie
>
> How about a way for Windows XP to run concurrently with the new Intel
> version of OS X? Apple could license XP and integrate it into future
> versions of OSX. End users would probably need to buy an XP license, but it
> would be there for those who want it.
That's like buying a bicycle for your pet fish.
> The biggest advantage - no need for companies to support two versions of the
> same program. XP users would be able to run OSX programs and vice versa.
> Microsoft still makes their $100 per head for an XP license and so does
> Apple.
And there would be no reason for OS X.
Timberwoof
Steve Hix <se...@NOSPAMspeakeasy.netINVALID> wrote:
It was badly timed for that: It was released on Friday at 5:00, after close of
trading. By the time Monday rolls around, people will realize that all they have
to do is wait for the keynote speech to finish and they'll know.
But I wonder what happened on Friday:
http://finance.yahoo.com/q/bc?t=5d&s=AAPL&l=on&z=m&q=l&c=&c=%5EIXIC
Looks like someone sold an awful lot of shares.
imout...@mac.com
> In article <sehix-E7C69F....@news.isp.giganews.com>,
> Steve Hix <se...@NOSPAMspeakeasy.netINVALID> wrote:
>
> But I wonder what happened on Friday:
> http://finance.yahoo.com/q/bc?t=5d&s=AAPL&l=on&z=m&q=l&c=&c=%5EIXIC
> Looks like someone sold an awful lot of shares.
News/rumor that the iPod shuffle was not selling through or something.
heheh, now that I'm getting free batteries for my 1st and 2nd
generation iPods I may not need to buy a new one for another 2-3
years...
Steve Hix
TravelinMan <Now...@spamfree.com> wrote:
Sun has been doing something similar for several years.
The main problem being that the card pretty much costs as much as a PC
of about the same performance; production volumes are too low to make
for useful cost reductions.
TravelinMan
Steve Hix <se...@NOSPAMspeakeasy.netINVALID> wrote:
At the time that Apple offered the card, it wasn't as much as a full PC.
You save the cost of a case, video card, power supply, optical disk and
hard disk. Essentially, the card is a motherboard with RAM and CPU - and
nothing else.
George Graves
"bob" <dv...@aolnospam.com> wrote:
> "Cameron Kaiser" <cka...@floodgap.com> wrote in message
> news:42a192b5$0$16185$bb4e...@newscene.com...
> > brianleahy <brianlea...@nomx.macosx.com> writes:
> >
> > >>this isn't going to happen, it's a red herring
> >
> > >I hope you're right.
> >
> > So do I. This is horrible news. At least back in the 68K days, we knew we
> > were getting a *better* processor as an upgrade when the PPC switch came.
>
> Depends on your definition of *better*. If it really was *better* then this
> thread wouldn't exist. I always was under the impression that most MAC users
> hated MS Windows more than Intel processors. I wonder what sales would be
> the if the choices were PPC and Windows or Intel and OSX(or pick your Apple
> OS)?
I, for one having NOTHING whatsoever against Intel's x86 processors.
Even if they were better than the PPC, they are pretty close to the end
of their life as far as far a speed increases are concerned, they are
not 64-bit processors, they don't scale well in multiprocessor
applications. I also wouldn't want a computer with a BIOS chip! However,
Apple would likely design around around that archaic silliness. Now, If
Apple went for an AMD Opteron, it might make sense. It's almost twice as
fast as a Xenon Galatian, and considerably faster than a 2.7 GHz G5 (by
about 35%). But the Opteron is AMD, not Intel. The G5 is considerably
quicker, on average, than is either Xenon, so I don't see the point of
Apple switching. Intel's fastest 64-bit processors are not as fast as
the IBM PPC, so why would Apple change?
http://www.anandtech.com/mac/showdoc.aspx?i=2436&p=5
Would it be to break the 3 GHz "mindshare" barrier? Since this "barrier"
has nothing whatsoever to do with actual processor performance, I simply
cannot see the logic behind such a move.
I would be flabbergasted by such a move and am surprised that the
"computer press" would start such a rumor without similarly surmising
it's likelihood.
--
George Graves
------------------
A sports car makes the journey more fun than the destination.
George Graves
Harri Mellin <atz02-...@netscape.net> wrote:
> In article <1117865211.9...@o13g2000cwo.googlegroups.com>,
> imout...@mac.com wrote:
>
> > Going x86's not that big a deal on the user end, aside from legacy app
> > compatibility. It's a significant but not immense PITA for developers,
> > tho, but if developers write to Carbon or Foundation, then get
> > x-platform support for free and x86 binaries are only a recompile away.
>
> or they only tell you to get vpc with windows if you want to use their
> programs (cheaper for them)
This is my biggest fear, I'm afraid. If Apple changed to X86
architecture, it would mean that developers would have to have TWO
versions of their software running on the same 'platform' and still
include (for a while, anyway) PPC versions of each title. They might
balk at that and tell X86 Mac owners that if they want to run these
apps, they'll have to install Windows either as a second OS, or as VPC
(easier to do without the big performance hit since no loner will the
processor have to be emulated). This would do what M$ has been unable to
do for more than 20 years - kill-off the Mac.
Jobs couldn't help but understand the ramifications of such a move.
Apple will not be changing to X86 processors for a variety of reasons
both hardware and software.
George Graves
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <1117865211.9...@o13g2000cwo.googlegroups.com>,
> imout...@mac.com says...
> > brianleahy wrote:
> > > > if Apple were to abandon PPC, your Mac would be a first-class OS X box
> > > > for the next 5 years at least.
> > >
> > > ..while all the cutting edge development, of both hardware
> >
> > ? Randy already bought his hardware. There's very little Apple would be
> > doing to update it for him now, even if they stick with IBM.
>
> Thanks for cheering me up. :-)
>
> > Firewire 800, 802.11b/g, USB 2.0, PCI, AGP 8x, etc etc expansion for
> > Randy's G5 will still remain supported regardless of the CPU/chipset
> > Apple moves to.
>
> Yet as new features get poured into new operating system releases and
> applications, fewer and fewer of them will work on my suddenly-found
> boat anchor.
>
> > Going x86's not that big a deal on the user end, aside from legacy app
> > compatibility. It's a significant but not immense PITA for developers,
> > tho, but if developers write to Carbon or Foundation, then get
> > x-platform support for free and x86 binaries are only a recompile away.
>
> Not to mention that the much more finely tuned optimizer for x86-64 will
> give an immediate performance boost to a lot of things just by doing
> that one step.
Except that the only 64-bit "X-86/64" that's actually faster than the
2.7 GHz G5 isn't built by Intel, it's made by AMD and called the Opteron.
George Graves
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-0AA000...@news.central.cox.net>,
> Now...@spamfree.com says...
> > > bigger cache, faster clock, dual core now, hyperthreading (yeah, I know
> > > HT is BS but marketing loves it), and compilers that actually optimize.
> >
> > PPC doesn't need a bigger cache as much as Intel given the immense
> > memory bandwidth of the 1.35 GHz bus.
>
> PPC gets exactly the same memory throughput out of DDR 3200 as does
> the Intel. They both get exactly 3200MB/s, as expected, with dual
> processors. Problem is, with a Xeon, and a single processor you
> can also get that same 3200, but PPC gets about 500MB/s less than
> that, for no good reason.
>
> Cache misses are still hugely expensive, on both. Given that PPC uses
> more instructions to do the same work, cache is arguably more important.
>
> An Opteron dual processor box will get about 5900MB/s using the
> same ram due to hypertransport. When you go to 4-way, it's
> almost 10GB/s. Nothing scales memory throughput w/extra processors
> the way opteron does today.
But Opteron is AMD, not Intel.
TravelinMan
<gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
George Graves <gmgra...@pacbell.net> wrote:
And it's not faster by enough to justify a wholesale platform change.
Randy Howard
Now...@spamfree.com says...
It absolutely is, particularly for an x86-64 dual processor Powermac.
Memory bandwidth by the truckload relative to anything else out there.
Chad Irby
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-4F0323...@news.central.cox.net>,
> Now...@spamfree.com says...
> > In article
> > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > George Graves <gmgra...@pacbell.net> wrote:
>
> > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the Opteron.
> >
> > And it's not faster by enough to justify a wholesale platform change.
>
> It absolutely is, particularly for an x86-64 dual processor Powermac.
> Memory bandwidth by the truckload relative to anything else out there.
...and that's about it, really. The early benchmarkers for the new
dual-core models are coming up with about a ten percent performance
boost over the rest of the processors on the market. Considering that
the Opterons they're testing are over $2500 per *chip*, that's not too
impressive...
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-4F0323...@news.central.cox.net>,
> Now...@spamfree.com says...
> > In article
> > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > George Graves <gmgra...@pacbell.net> wrote:
>
> > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the Opteron.
> >
> > And it's not faster by enough to justify a wholesale platform change.
>
> It absolutely is, particularly for an x86-64 dual processor Powermac.
> Memory bandwidth by the truckload relative to anything else out there.
Evidence?.....
Randy Howard
> In article <MPG.1d0d617e2...@news.verizon.net>,
> Randy Howard <randy...@FOOverizonBAR.net> wrote:
>
> > In article <Nowhere-4F0323...@news.central.cox.net>,
> > Now...@spamfree.com says...
> > > In article
> > > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > > George Graves <gmgra...@pacbell.net> wrote:
> >
> > > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the Opteron.
> > >
> > > And it's not faster by enough to justify a wholesale platform change.
> >
> > It absolutely is, particularly for an x86-64 dual processor Powermac.
> > Memory bandwidth by the truckload relative to anything else out there.
>
> Evidence?.....
Are you really going to play that lame card again?
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-21B783...@news.central.cox.net>,
> Now...@spamfree.com says...
> > In article <MPG.1d0d617e2...@news.verizon.net>,
> > Randy Howard <randy...@FOOverizonBAR.net> wrote:
> >
> > > In article <Nowhere-4F0323...@news.central.cox.net>,
> > > Now...@spamfree.com says...
> > > > In article
> > > > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > > > George Graves <gmgra...@pacbell.net> wrote:
> > >
> > > > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the
> > > > > Opteron.
> > > >
> > > > And it's not faster by enough to justify a wholesale platform change.
> > >
> > > It absolutely is, particularly for an x86-64 dual processor Powermac.
> > > Memory bandwidth by the truckload relative to anything else out there.
> >
> > Evidence?.....
>
> Are you really going to play that lame card again?
So in your view asking for evidence is lame?
Randy Howard
> In article <MPG.1d0d6f513...@news.verizon.net>,
> Randy Howard <randy...@FOOverizonBAR.net> wrote:
>
> > In article <Nowhere-21B783...@news.central.cox.net>,
> > Now...@spamfree.com says...
> > > In article <MPG.1d0d617e2...@news.verizon.net>,
> > > Randy Howard <randy...@FOOverizonBAR.net> wrote:
> > >
> > > > In article <Nowhere-4F0323...@news.central.cox.net>,
> > > > Now...@spamfree.com says...
> > > > > In article
> > > > > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > > > > George Graves <gmgra...@pacbell.net> wrote:
> > > >
> > > > > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > > > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the
> > > > > > Opteron.
> > > > >
> > > > > And it's not faster by enough to justify a wholesale platform change.
> > > >
> > > > It absolutely is, particularly for an x86-64 dual processor Powermac.
> > > > Memory bandwidth by the truckload relative to anything else out there.
> > >
> > > Evidence?.....
> >
> > Are you really going to play that lame card again?
>
> So in your view asking for evidence is lame?
It's fairly common knowledge, and has been discovered many places,
including CSMA for quite a while. If you want further information,
google for "hypertransport".
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
I'm quite aware of hypertransport. But since you seem to be pretty weak
on understanding performance issues, let me spell it out to you - bus
performance is only one part of the performance equation.
My statement was that the AMD processors don't have enough of a speed
advantage (if any) over the G5 to justify Apple making a wholesale
switch.
Basically, that's like my saying that your VW Rabbit won't beat a V8
Corvette and your responding "but look at the rpms of the Rabbit".
If you have any evidence that the overall performance of the Opteron has
enough of an advantage over the G5 to justify Apple dropping everything
to switch, please feel free to present it. And then explain why they're
talking to Intel if they want Opteron processors.
Your fantasies don't even make sense.
Chad Irby
Randy Howard <randy...@FOOverizonBAR.net> wrote:
But the real problem is that the Hypertransport tech just isn't
translating to huge real-world increases in speed, all by itself.
Randy Howard
Now...@spamfree.com says...
> > It's fairly common knowledge, and has been discovered many places,
> > including CSMA for quite a while. If you want further information,
> > google for "hypertransport".
>
> I'm quite aware of hypertransport.
Having heard of it before means very little in the grand scheme
of things.
> But since you seem to be pretty weak on understanding performance issues,
LOL
> let me spell it out to you - bus performance is only one part of the
> performance equation.
That's very true. The biggest pure performance problem with the Mac
platorm right now is file system-related, and no CPU change is likely
to be a bandaid for that.
> My statement was that the AMD processors don't have enough of a speed
> advantage (if any) over the G5 to justify Apple making a wholesale
> switch.
If they are going to switch, then Opteron is certainly a much better
reason than Xeon EM64t. If they are not going to switch, then it doesn't
matter at all.
> Basically, that's like my saying that your VW Rabbit won't beat a V8
> Corvette and your responding "but look at the rpms of the Rabbit".
Interesting in that your claim belief that a G5 is performance
competitive with an opteron system is ludicrious, where your pseudo-
analogy of claiming that a Rabbit can't beat a Vette is obvious, barring
nitrous or some other non-stock deviation.
> If you have any evidence that the overall performance of the Opteron has
> enough of an advantage over the G5 to justify Apple dropping everything
> to switch, please feel free to present it.
Ahh, different spin entirely. I have no believe that anything apart from
incredible anger between Apple and IBM would justify them dropping such
a crowbar in the mix right now.
> And then explain why they're talking to Intel if they want Opteron
> processors.
Rumors exist for talks between Apple and both companies, so it's no
big stretch to pretend like one leaked before the other.
> Your fantasies don't even make sense.
A) What fantasies?
B) In what universe would something making sense to the likes of you
be in any way interesting to others, other than perhaps shock value?
Randy Howard
ci...@cfl.rr.com says...
> > It's fairly common knowledge, and has been discovered many places,
> > including CSMA for quite a while. If you want further information,
> > google for "hypertransport".
>
> But the real problem is that the Hypertransport tech just isn't
> translating to huge real-world increases in speed, all by itself.
HyperTransport makes cache misses break-even to slightly faster on
uniprocessor, almost twice as fast on dual processor systems, and almost
4x as fast and quad CPU boxes. Cache misses are becoming increasingly
expensive as each CPU speed bump comes along.
David Wang
> In article <cirby-57B708....@news-server1.tampabay.rr.com>,
> ci...@cfl.rr.com says...
> >
> > translating to huge real-world increases in speed, all by itself.
> HyperTransport makes cache misses break-even to slightly faster on
> uniprocessor, almost twice as fast on dual processor systems, and almost
> 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> expensive as each CPU speed bump comes along.
You have that backwards.
ccHT (as in Opteron) shines best on uniprocessors. Its latency
advantage in 2P boxes vs Intel style shared FSB drops next to nil,
and cache-miss DRAM latency in glueless 4P Opteron box vs 4P shared
FSB is actually slightly worse.
The snoop-coherency latency predominates memory latency in 2P and
4P configs, and it gets worse as processor count increases.
--
davewang202(at)yahoo(dot)com
Chad Irby
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <cirby-57B708....@news-server1.tampabay.rr.com>,
> ci...@cfl.rr.com says...
> > > It's fairly common knowledge, and has been discovered many places,
> > > including CSMA for quite a while. If you want further information,
> > > google for "hypertransport".
> >
> > But the real problem is that the Hypertransport tech just isn't
> > translating to huge real-world increases in speed, all by itself.
>
> HyperTransport makes cache misses break-even to slightly faster on
> uniprocessor, almost twice as fast on dual processor systems, and almost
> 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> expensive as each CPU speed bump comes along.
...and are much more expensive on Intel and AMD-based chips, when
compared to the PPC line.
Randy Howard
> Randy Howard <randy...@fooverizonbar.net> wrote:
> > In article <cirby-57B708....@news-server1.tampabay.rr.com>,
> > ci...@cfl.rr.com says...
> > >
> > > But the real problem is that the Hypertransport tech just isn't
> > > translating to huge real-world increases in speed, all by itself.
>
> > HyperTransport makes cache misses break-even to slightly faster on
> > uniprocessor, almost twice as fast on dual processor systems, and almost
> > 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> > expensive as each CPU speed bump comes along.
>
> You have that backwards.
>
> ccHT (as in Opteron) shines best on uniprocessors.
Depends upon whether you are talking about bandwidth or latency.
> Its latency advantage in 2P boxes vs Intel style shared FSB drops next to
> nil,
And it's memory bandwidth is almost double that of Intel. nice tradeoff.
George Graves
TravelinMan <Now...@spamfree.com> wrote:
Well, precisely. But the speed of any processor family at any snapshot
in time is pretty meaningless. What's important is how well that family
can be developed in the future. A processor, for instance, that's twice
as fast as a PPC (if one existed) today would be useless to Apple to
migrate toward if it were at the end of its development potential.
George Graves
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-4F0323...@news.central.cox.net>,
> Now...@spamfree.com says...
> > In article
> > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > George Graves <gmgra...@pacbell.net> wrote:
>
> > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the Opteron.
> >
> > And it's not faster by enough to justify a wholesale platform change.
>
> It absolutely is, particularly for an x86-64 dual processor Powermac.
> Memory bandwidth by the truckload relative to anything else out there.
Today's performance is NOT important. What's its potential? If it cannot
be further developed to 5, 6, even 10 or 20 GHz over the next 5 years,
it's a dead end and not worth considering.
George Graves
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-21B783...@news.central.cox.net>,
> Now...@spamfree.com says...
> > In article <MPG.1d0d617e2...@news.verizon.net>,
> > Randy Howard <randy...@FOOverizonBAR.net> wrote:
> >
> > > In article <Nowhere-4F0323...@news.central.cox.net>,
> > > Now...@spamfree.com says...
> > > > In article
> > > > <gmgravesnos-126A...@newssvr13-ext.news.prodigy.com>,
> > > > George Graves <gmgra...@pacbell.net> wrote:
> > >
> > > > > Except that the only 64-bit "X-86/64" that's actually faster than the
> > > > > 2.7 GHz G5 isn't built by Intel, it's made by AMD and called the
> > > > > Opteron.
> > > >
> > > > And it's not faster by enough to justify a wholesale platform change.
> > >
> > > It absolutely is, particularly for an x86-64 dual processor Powermac.
> > > Memory bandwidth by the truckload relative to anything else out there.
> >
> > Evidence?.....
>
> Are you really going to play that lame card again?
I want to see evidence of a realistic development road-map. If the
architecture is not able to be developed 10 GHz and beyond, It won't buy
Apple or anybody else very much.
Randy Howard
ext.news.prodigy.com>, gmgra...@pacbell.net says...
I doubt very much that super-high clock speeds are in the cards, more
likely will be "wide" computers, multi-core, each core at lower clock
speeds, but lots more of them, and software that takes advantage of
threading to a much higher degree.
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
Notice that there's not even one line dedicated to your argument that
Opteron offers enough performance advantages over G5 to justify the
switch.
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <cirby-57B708....@news-server1.tampabay.rr.com>,
> ci...@cfl.rr.com says...
> > > It's fairly common knowledge, and has been discovered many places,
> > > including CSMA for quite a while. If you want further information,
> > > google for "hypertransport".
> >
> > But the real problem is that the Hypertransport tech just isn't
> > translating to huge real-world increases in speed, all by itself.
>
> HyperTransport makes cache misses break-even to slightly faster on
> uniprocessor, almost twice as fast on dual processor systems, and almost
> 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> expensive as each CPU speed bump comes along.
That's nice.
Now, where's the evidence that the Opteron CHIP (not just the bus
protocol) offers performance advantages great enough to justify a change
of this scale?
Randy Howard
Now...@spamfree.com says...
> > HyperTransport makes cache misses break-even to slightly faster on
> > uniprocessor, almost twice as fast on dual processor systems, and almost
> > 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> > expensive as each CPU speed bump comes along.
>
> That's nice.
>
> Now, where's the evidence that the Opteron CHIP (not just the bus
> protocol)
Guess where the "bus protocol" is located for Opteron? It's an
on-die memory controller clueless.
Furrfu
TravelinMan
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <Nowhere-704096...@news.central.cox.net>,
> Now...@spamfree.com says...
> > > HyperTransport makes cache misses break-even to slightly faster on
> > > uniprocessor, almost twice as fast on dual processor systems, and almost
> > > 4x as fast and quad CPU boxes. Cache misses are becoming increasingly
> > > expensive as each CPU speed bump comes along.
> >
> > That's nice.
> >
> > Now, where's the evidence that the Opteron CHIP (not just the bus
> > protocol)
>
> Guess where the "bus protocol" is located for Opteron? It's an
> on-die memory controller clueless.
>
> Furrfu
That's nice.
Now answer the question. Where's your evidence that OVERALL CHIP
PERFORMANCE for the opteron is fast enough to justify a wholesale change?
A fast bus is just one of the elements to consider.
George Graves
Randy Howard <randy...@FOOverizonBAR.net> wrote:
> In article <gmgravesnos-C89982.23344905062005@newssvr21-
> ext.news.prodigy.com>, gmgra...@pacbell.net says...
> > In article <MPG.1d0d6f513...@news.verizon.net>,
> > Randy Howard <randy...@FOOverizonBAR.net> wrote:
> > I want to see evidence of a realistic development road-map. If the
> > architecture is not able to be developed 10 GHz and beyond, It won't buy
> > Apple or anybody else very much.
>
> I doubt very much that super-high clock speeds are in the cards, more
> likely will be "wide" computers, multi-core, each core at lower clock
> speeds, but lots more of them, and software that takes advantage of
> threading to a much higher degree.
Oh, super-high clock-speeds are in the cards, just not with current
design rules or current architectures. Yes, wide computing,
multiprocessors, multicore processors are all in the offing too. But
scalability limits the efficacy of multi-cores and multi-processors, and
they will have to be used along with higher clock speeds, 'tricks' like
hyperthreading and vectoring co-processors to keep pushing the envelope
for applications which NEED that power.
Randy Howard
Now...@spamfree.com says...
Apparently Intel is enough to justify a wholesale change, so it's all
water under the bridge now.