After all the disapointing expos (part of the reason why I bought a PC, just nothing interesting coming out) I doubt anything will come of this. Weve been shouting G5 at every expo since the upgraded G4es.
I think one thing to be announced is the opening of the Apple Store in Soho-NYC. There may even be some MacWorld event tie-ins at the store. From what I have heard. this will be the biggest Apple Store yet. And something of a mega store. The fact that it is scheduled to open in July is too coincidental.
<strong>Apple is going to announce that despite 10 years of dedicated work on PowerPC by Apple, Motorola and IBM the whole thing is a dead end technologically speaking.]</strong><hr></blockquote>
4 core processor modules (with 2 CPUs per core) anyone?
[quote]For creation of multimodule systems it's possible to use up to 4 modules which allow for 16-processor SMP system or, taking into account 2 processors on each POWER4 chip, for a 32-processor one. Two unidirectional 64-bit buses are used to connect other modules;<hr></blockquote>
And no, I am NOT saying that Apple will be coming out with machines using these processors -- just noting that the PPC architecture is doing just fine...
<strong>moki: Talk about stirring the anthill with a stick! </strong><hr></blockquote>
This should help too
[code]CPU CPU MHz CINT2000 base/peak CFP2000 base/peak
AMD Athlon XP 1900+ 1600 677/701 588/634
HP PA-8700 750 568/604 526/576
IBM POWER4 (1CPU) 1300 790/814 1098/1169
Intel Itanium 800 314/314 645
Intel Pentium 4 2200 771/784 766/777
SUN UltraSPARC III-Cu 900 470/533 629/731
</pre><hr></blockquote>
The POWER4 consists of 2 identical processor cores which implement PowerPC AS instruction set, the die measures about 400 mm2, it's based on the 0.18 micron copper SOI IBM CMOS 8S2 technology with 7 metallization layers, works at 1.1 and 1.3 GHz, and is the fastest microprocessor for today.
Note: the G4 is turning in performance that's worse than it needs to be; the gcc compiler used to do these tests produces absolutely awful PPC code, and of course, AltiVec isn't factored into this test either. Nevertheless, it does show something about relative processor performance.
Don't be -- benchmarks are sort of like 0-60 times for cars. Sure, it's cool to have a car that'll go 0-60 in under 5 seconds, but 99% of the time, you're using your car to go to and from the grocery store.
Benchmarks show how good a processor/compiler combination is at a particular benchmark algorithm. It doesn't indicate how the computer system as a whole will perform in the real world, and it is very true that the compiler used to compile the benchmark is as crucial as the processor itself.
Still, *something* must be used to compare processors, so a set of contrived benchmarks is as good as anything else.
Regardless, the POWER4 is a very fast processor -- especially the floating point performance is impressive. Maybe I'll try to figure out how to shoehorn a 4 core package into my Mac so I can have an 8 processor system that does realtime PhotoShop filters
Don't be -- benchmarks are sort of like 0-60 times for cars. Sure, it's cool to have a car that'll go 0-60 in under 5 seconds, but 99% of the time, you're using your car to go to and from the grocery store.
Benchmarks show how good a processor/compiler combination is at a particular benchmark algorithm. It doesn't indicate how the computer system as a whole will perform in the real world, and it is very true that the compiler used to compile the benchmark is as crucial as the processor itself.
Still, *something* must be used to compare processors, so a set of contrived benchmarks is as good as anything else.
Regardless, the POWER4 is a very fast processor -- especially the floating point performance is impressive. Maybe I'll try to figure out how to shoehorn a 4 core package into my Mac so I can have an 8 processor system that does realtime PhotoShop filters </strong><hr></blockquote>
Moki - your recent posts seem to be concentrating on multi-core chips, know something your not telling or just hope for the future (escape clause)
Moki (and others). Sorry if this is really a stupid question, but is it at all possible to use a Power4 (or Power3 for that matter) in the Mac? Could there some day be a 'workstation' Mac? Obviously altivec wont work, but since they are in the same 'family', is it doable? Moki's comment about Photoshop on the Power4 caught my eye.. :cool:
It should also be noted that the POWER4 smokes the competition on a 0.18 micron process technology! They are set to move to a 0.13 micron technology in the 2nd half of '02.
The POWER4 in its current form wouldn't be directly applicable to a PowerMac. The 32 megabyte L3 cache is quite expensive, and the huge 170 million transistor cores would make the G4 look positively cheap. The loss of AltiVec would seriously impact Apple's code base. That doesn't mean there aren't some interesting possibilities here though...
The POWER4 has been in design since at least '99, and shipping since late last year. It uses very deep pipelines with the x86 style decoder into IOPs feeding an extremely super-scalar core with a huge number of internal registers. The caches are really big, and more importantly, have large line sizes which make the bursts to/from main memory more efficient. The design is a 2-way multicore chip.
The decoder and wide superscalar design should really lend itself well to implementing the AltiVec instruction set, I suspect. The AltiVec instructions would be internally broken into multiple IOPs, and fed to the many execution units. Add the cache control instructions and the vector registers and you should have an AltiVec-equivalent that uses the existing chip resources.
The bus interface IBM has implemented here doesn't look very applicable to a desktop machine (much less a portable) -- the motherboards would be much too expensive. Removing the L3 cache and replacing the bus with an on-chip memory controller (a la Motorola) and a HyperTransport link to the I/O chip and things should much better suited to an Apple machine. I'd sure as heck buy one... if they pushed this sucker out the door on 0.13 micron, it would be a real barn burner.
All purely speculation, of course, but this thing could be really impressive. At least doubling the G4's top performance numbers, even at low clock rates. Except for it being built by IBM and not Moto, it would also jive with all of the rumours from The Register, MOSR, AI, etc (including the timing of them last year)... which pretty much proves that its not true, right?
Moki - your recent posts seem to be concentrating on multi-core chips, know something your not telling or just hope for the future (escape clause)</strong><hr></blockquote>
Thought this snippet from The Silicon Grail (who acc. to an article quoted by Macrumors was just bought by Apple) Website:
[quote] More and more, workstations are sold with two (or more) processors, and RAYZ is able to take advantage of multiprocessor architectures much more effectively than its competitors. RAYZ running on 2 CPUs typically renders 180% faster than the same shot on a single CPU. Competing products typically experience only a 10% improvement in performance.
<hr></blockquote>
Wow Apple sure likes the video market and they seem to have a taste for acquiring products that are coded to work well with MP. I guess that could lead to more/better/faster of the current idea (multiple discrete processors) or something similar to the Power4 idea (multiple procs on a single piece of Silicon) <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />
Bob, thanks for the additional info. Although I'm tempted to rant in the gloom and doom mode from time to time (i.e. tempted to foresee a rotten future Apple), there's too much going on in Apple's acquisition world to suggest that they are going to lay down and smoke out.
And thank God for people like Programmer on these boards, the man who writes code and who brings a technical expertise that most of us (myself absolutely included) lack. I'm no technie--just a lazy hazy designer--and so all I can contribute on these boards is a twisted turn of phrase or a stab at a comprehensive overview now and then, but whether its an IBM Power4 chip or some imaginative multiprocessor configuration, the pieces suggest that Apple will find a way.
Quite simply Apple is too entrenched in the video and graphics realm to stop bringing on better hardware. Just the other day I read that Apple's software sales amount to only 18% of the company's total earnings. Hardware is the bread that their software butter melts into.
Going back to the title of this thread, the question of something big, hardware-wise, is not IF but WHEN. Unfortunately, it may be late fall or sometime next year before we see the harvest. Apple has shored up the consumer realm, at least for one quarter, with the iMac. The big hardware advances will come--we just don't know when.
________________________
I Haven't Bought the Dual Gig Machine Yet in Nashvegas ? Still Waiting
<strong>This post seemed more "emotional" than "factual". Finally, why would the poster describe his profession in such detail. How many "flame artists" (or whatever he called himself) could there possibly be, that would also happen to have a preproduction "Quad G5" system?</strong><hr></blockquote>
well, as far as flame artists go, i've personally worked with probably 30 of them. there are tons of them out there, so i don't think he gave too much away. of course, i don't know how many of them are important enough to get a quad g5 to test. maybe digital domain. anyway, flame artist is indeed a real job and it is exactly the field that shake is involved in.
well, as far as flame artists go, i've personally worked with probably 30 of them. there are tons of them out there, so i don't think he gave too much away. of course, i don't know how many of them are important enough to get a quad g5 to test. maybe digital domain. anyway, flame artist is indeed a real job and it is exactly the field that shake is involved in.</strong><hr></blockquote>
Comments
<strong>Maybe the NY update will be impressive with delayed shipping dates...</strong><hr></blockquote>
That must be the best idea I've seen. I hope it is. I don't want to go to a LAN party and be laughed at!
<strong>Apple is going to announce that despite 10 years of dedicated work on PowerPC by Apple, Motorola and IBM the whole thing is a dead end technologically speaking.]</strong><hr></blockquote>
IBM would be surprised to hear that; their <a href="http://www.digit-life.com/articles/ibmpower4/" target="_blank">POWER4</a> microprocessor family has been kicking ass and taking numbers...
4 core processor modules (with 2 CPUs per core) anyone?
[quote]For creation of multimodule systems it's possible to use up to 4 modules which allow for 16-processor SMP system or, taking into account 2 processors on each POWER4 chip, for a 32-processor one. Two unidirectional 64-bit buses are used to connect other modules;<hr></blockquote>
And no, I am NOT saying that Apple will be coming out with machines using these processors -- just noting that the PPC architecture is doing just fine...
[ 06-08-2002: Message edited by: moki ]</p>
<strong>moki: Talk about stirring the anthill with a stick! </strong><hr></blockquote>
This should help too
[code]CPU CPU MHz CINT2000 base/peak CFP2000 base/peak
AMD Athlon XP 1900+ 1600 677/701 588/634
HP PA-8700 750 568/604 526/576
IBM POWER4 (1CPU) 1300 790/814 1098/1169
Intel Itanium 800 314/314 645
Intel Pentium 4 2200 771/784 766/777
SUN UltraSPARC III-Cu 900 470/533 629/731
</pre><hr></blockquote>
The POWER4 consists of 2 identical processor cores which implement PowerPC AS instruction set, the die measures about 400 mm2, it's based on the 0.18 micron copper SOI IBM CMOS 8S2 technology with 7 metallization layers, works at 1.1 and 1.3 GHz, and is the fastest microprocessor for today.
or some others similar test.
It will be interesting.
<strong>Do we have the CINT2000 & CFP 2000 for the PowerPC 1000 Mhz ?
or some others similar test.
It will be interesting.</strong><hr></blockquote>
<a href="http://www.mtl.t.u-tokyo.ac.jp/~nminoru/memo/spec_cpu2000.html" target="_blank">http://www.mtl.t.u-tokyo.ac.jp/~nminoru/memo/spec_cpu2000.html</a>
Note: the G4 is turning in performance that's worse than it needs to be; the gcc compiler used to do these tests produces absolutely awful PPC code, and of course, AltiVec isn't factored into this test either. Nevertheless, it does show something about relative processor performance.
[ 06-08-2002: Message edited by: moki ]</p>
<strong>now i'm depressed </strong><hr></blockquote>
Don't be -- benchmarks are sort of like 0-60 times for cars. Sure, it's cool to have a car that'll go 0-60 in under 5 seconds, but 99% of the time, you're using your car to go to and from the grocery store.
Benchmarks show how good a processor/compiler combination is at a particular benchmark algorithm. It doesn't indicate how the computer system as a whole will perform in the real world, and it is very true that the compiler used to compile the benchmark is as crucial as the processor itself.
Still, *something* must be used to compare processors, so a set of contrived benchmarks is as good as anything else.
Regardless, the POWER4 is a very fast processor -- especially the floating point performance is impressive. Maybe I'll try to figure out how to shoehorn a 4 core package into my Mac so I can have an 8 processor system that does realtime PhotoShop filters
I love it when they mix 32 bit and 64 bit CPU's for testing!
<strong>LOL when I saw those benchmarks. <img src="graemlins/lol.gif" border="0" alt="[Laughing]" />
I love it when they mix 32 bit and 64 bit CPU's for testing!</strong><hr></blockquote>
Why shouldn't they?
<strong>
Don't be -- benchmarks are sort of like 0-60 times for cars. Sure, it's cool to have a car that'll go 0-60 in under 5 seconds, but 99% of the time, you're using your car to go to and from the grocery store.
Benchmarks show how good a processor/compiler combination is at a particular benchmark algorithm. It doesn't indicate how the computer system as a whole will perform in the real world, and it is very true that the compiler used to compile the benchmark is as crucial as the processor itself.
Still, *something* must be used to compare processors, so a set of contrived benchmarks is as good as anything else.
Regardless, the POWER4 is a very fast processor -- especially the floating point performance is impressive. Maybe I'll try to figure out how to shoehorn a 4 core package into my Mac so I can have an 8 processor system that does realtime PhotoShop filters </strong><hr></blockquote>
Moki - your recent posts seem to be concentrating on multi-core chips, know something your not telling or just hope for the future (escape clause)
[ 06-08-2002: Message edited by: Thai Moof ]</p>
The POWER4 in its current form wouldn't be directly applicable to a PowerMac. The 32 megabyte L3 cache is quite expensive, and the huge 170 million transistor cores would make the G4 look positively cheap. The loss of AltiVec would seriously impact Apple's code base. That doesn't mean there aren't some interesting possibilities here though...
The POWER4 has been in design since at least '99, and shipping since late last year. It uses very deep pipelines with the x86 style decoder into IOPs feeding an extremely super-scalar core with a huge number of internal registers. The caches are really big, and more importantly, have large line sizes which make the bursts to/from main memory more efficient. The design is a 2-way multicore chip.
The decoder and wide superscalar design should really lend itself well to implementing the AltiVec instruction set, I suspect. The AltiVec instructions would be internally broken into multiple IOPs, and fed to the many execution units. Add the cache control instructions and the vector registers and you should have an AltiVec-equivalent that uses the existing chip resources.
The bus interface IBM has implemented here doesn't look very applicable to a desktop machine (much less a portable) -- the motherboards would be much too expensive. Removing the L3 cache and replacing the bus with an on-chip memory controller (a la Motorola) and a HyperTransport link to the I/O chip and things should much better suited to an Apple machine. I'd sure as heck buy one... if they pushed this sucker out the door on 0.13 micron, it would be a real barn burner.
All purely speculation, of course, but this thing could be really impressive. At least doubling the G4's top performance numbers, even at low clock rates. Except for it being built by IBM and not Moto, it would also jive with all of the rumours from The Register, MOSR, AI, etc (including the timing of them last year)... which pretty much proves that its not true, right?
[ 06-08-2002: Message edited by: Programmer ]</p>
<strong>
Moki - your recent posts seem to be concentrating on multi-core chips, know something your not telling or just hope for the future (escape clause)</strong><hr></blockquote>
Thought this snippet from The Silicon Grail (who acc. to an article quoted by Macrumors was just bought by Apple) Website:
[quote] More and more, workstations are sold with two (or more) processors, and RAYZ is able to take advantage of multiprocessor architectures much more effectively than its competitors. RAYZ running on 2 CPUs typically renders 180% faster than the same shot on a single CPU. Competing products typically experience only a 10% improvement in performance.
<hr></blockquote>
Wow Apple sure likes the video market and they seem to have a taste for acquiring products that are coded to work well with MP. I guess that could lead to more/better/faster of the current idea (multiple discrete processors) or something similar to the Power4 idea (multiple procs on a single piece of Silicon) <img src="graemlins/hmmm.gif" border="0" alt="[Hmmm]" />
And thank God for people like Programmer on these boards, the man who writes code and who brings a technical expertise that most of us (myself absolutely included) lack. I'm no technie--just a lazy hazy designer--and so all I can contribute on these boards is a twisted turn of phrase or a stab at a comprehensive overview now and then, but whether its an IBM Power4 chip or some imaginative multiprocessor configuration, the pieces suggest that Apple will find a way.
Quite simply Apple is too entrenched in the video and graphics realm to stop bringing on better hardware. Just the other day I read that Apple's software sales amount to only 18% of the company's total earnings. Hardware is the bread that their software butter melts into.
Going back to the title of this thread, the question of something big, hardware-wise, is not IF but WHEN. Unfortunately, it may be late fall or sometime next year before we see the harvest. Apple has shored up the consumer realm, at least for one quarter, with the iMac. The big hardware advances will come--we just don't know when.
________________________
I Haven't Bought the Dual Gig Machine Yet in Nashvegas ? Still Waiting
[ 06-08-2002: Message edited by: Sybaritic ]
[ 06-08-2002: Message edited by: Sybaritic ]</p>
<strong>This post seemed more "emotional" than "factual". Finally, why would the poster describe his profession in such detail. How many "flame artists" (or whatever he called himself) could there possibly be, that would also happen to have a preproduction "Quad G5" system?</strong><hr></blockquote>
well, as far as flame artists go, i've personally worked with probably 30 of them. there are tons of them out there, so i don't think he gave too much away. of course, i don't know how many of them are important enough to get a quad g5 to test. maybe digital domain. anyway, flame artist is indeed a real job and it is exactly the field that shake is involved in.
<strong>
well, as far as flame artists go, i've personally worked with probably 30 of them. there are tons of them out there, so i don't think he gave too much away. of course, i don't know how many of them are important enough to get a quad g5 to test. maybe digital domain. anyway, flame artist is indeed a real job and it is exactly the field that shake is involved in.</strong><hr></blockquote>
What is a "Flame Artist"?