In this Microprocessor Report PDF excerpt, the PowerPC 970 is reviewed by outside industry experts who conclude that “IBM has created an impressive and affordable PowerPC chip for smaller servers, graphics workstations and desktop computers“. Also, ArsTechnica posted the second part of their PPC 970 series of articles, “Inside the PowerPC 970,
Part II: The Execution Core“.
IBM Trims Power4, Adds AltiVec; ArsTechnica Analysis
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I’d love to have a Power4 workstation to run my Windows apps and games on.
Keep drooling… The only version of Windows that might run on it would be Windows 3.51 in 32bit mode with no SIMD support.
Also, most Windows games and Application don’t come in native PPC format; one of the reasons Microsoft droped PPC support when they released Windows 2K.
I liked Windows 3.51; it ran better on the Alpha, PPC, Sparc, and MIPS processers then it did on the x86. But most peopled didn’t care and stayed with the x86 bucause that’s what Windows had always run on (the other reason for no future alternat CPU support).
With Intel decaring the x86 dead (P4 is the last chip in the series), people are going to have to find a new processor. Microsoft is betting AMD x86-64 and Intel ia64. When the x86 disappears in 2010, everyone will need to switch to a non-Intel chip or ia64. Have fun…
Me. I like Apple’s, so I’ll be using the POWER4 and later the POWER5 family. I’m sure IBM is working a new geration after the POWER5; after all the POWER5 is only the 8th generation (the P4 is the 10th generation).
I wasn’t implying that it could be done on a current system, I was just saying I’d like to be able to.
Once again, ArsTechnica proves that it is still a bastion of intelligence against the dumbing down of the computing industry. Way to go Mr. Stokes.
PS> For those interested, check out their PS2 vs PC article. It’s very enlightening.
Apples will have more IBM parts then a PC! Hahahah
I know I’m biased in this area but has anyone else noticed how poor the graphics device support is in this area (speaking of PPC). Of course on the apple side of the fence this is no big deal as apple customers have always been forced to use top of the line cards;) but for the x86 based PC user its going to be quite a shock not to have a plethora of HW available
how can x86 PC users be disappointed by a GPU support from a processor they can’t use?
IBM used to say bad things about altivec but now seems to realize they’d be better off having Apple using their cpu’s. And of course Apple is much happier relying on IBM than Motorola. Apple now has a real chance to cruise past Intel and AMD which should be interesting
Speaking to the comment made earlier in the thread.
With Intel decaring the x86 dead (P4 is the last chip in the series), people are going to have to find a new processor. Microsoft is betting AMD x86-64 and Intel ia64. When the x86 disappears in 2010, everyone will need to switch to a non-Intel chip or ia64.
at the point of looking for a new HW platform and potentially looking at the PPC the issue will quickley crop up.
Microsoft isn’t going to support PPC. There are only a handful of Nixes available on PPC. And there are virtually zero sources of third-party PPCs.
I think Window users would have more concern about whether their OS would run rather than if some obscure GPU has the proper driver.
Because some poster says that 7 years from now x86 will be abandoned for 64bit Intel/AMD solutions… doesn’t really have much bearing on reality does it? Besides, I beleive he was suggesting that IBM and Apple will have an advantage over x86 users during this ten year tradition… Not that everyone will be forced to run PPC-based machines.
Besides… if everyone switched to PPC, do you think there would be any lack of driver support?
If you don’t get it yet, I understand what you are saying but it is nonsensical because it is based on fantasy and unreality.
IBM never said they didn’t like Altivec, they said the G3 still had some legs under it and that servers didn’t need a SIMD extension.
As impressive as the 970 is or will be, I would also like to see a low end chip by IBM. I want to see one of those system-on a-chips that IBM has been rumored to develop. A G3 + Altivec, memory controller and chipset all integrated. Apple could push itself really low with that and really high with the 970 simultaneously.
nick, you are right, IBM certanily wouldn’t badmouth Altivec…. they helped develop it!
Hmm, next time I will be sure to complete my thought prior to posting – call it a senior moment.
Where I was going with my earlier post is this: If you take a look at the *existing* PPC desktop market and exclude apple (this should be a rather quick process) there are comparatively few developers working on OS’es, applications and drivers for this platform. If the next generation of these chips is to succeed in mass then it must first find a way to generate developer interest as well as HW vendor interest. HW vendors will need to port/develop drivers in mass for the platform to make the next generation PPC chip viable for the desktop. Again I may be skewing the importance of the device support issue here as declared in my original post. As we looked into porting SNAP to support the PPC, I was amazed at how poor the HW support was for this platform.
With out good device support end users will have no compelling reason to switch to this new chip (again Apple aside). Additionally one must also realize that the current level of device support available to the x86 platform has set an extremely high standard and will be expected in the future on any new platform. I believe this to be true regardless of the technology merits of the chip (this for the most part is transparent to all but a select few).
I think Window users would have more concern about whether their OS would run rather than if some obscure GPU has the proper driver.
ignoring the term obscure – with out a working driver how would you know if your OS was running. You can never under estimate the importance of good device support.
Besides… if everyone switched to PPC, do you think there would be any lack of driver support?
No, SNAP would have them covered;) Thanks for bringing me to task on this – cheers
Are you Tris or Andrew?
But anyway, this is still silly. This isn’t a new platform. Nor is anyone attempting to make it a competitive platform for some new OS or some OS which hasn’t used it before. The PPC is widely used in the embedded space. YDL exists. Mac OS exists. IBM’s systems exist.
And the PPC exists and is successful. No one said it’s dying because Windows and every flavor of Linux isn’t on it. No one said OEMs are clamoring to get their hands on PCs but can’t because their aren’t GPU drivers. No one is trying to compel anyone to switch to this platform–there are already several self-sufficient platforms available for this chip. Mac users do not feel a susbtantial need for more GPU drivers. In the embedded space, the necessary drivers are available. In the IBM server world this is taken care of and/or is unnecessary.
I don’t know why you ask how would you know if your OS was running before having a GPU driver–that’s retarded. What devices need support if there is no system, no software, nothing. Trust me–build the system first. Build some apps next. Then concern yourself with drivers.
Otherwise the PPC family is perfectly fine and the communities served by it are satisfied.
Have you guys looked at specmark ? FPU benchmarks ?
http://www.specbench.org/cpu2000/results/cfp2000.html
Intel rules. Getting excited over a chip that barely competes with top of the line intel chip is no big deal.
They probably cost a shit load too.
Wake me up when electro-optical chips with tunneling transistors chips come out or something as revolutionary.
This is another wana be chip and brings nothing new to the table except price for apples proprietary BULLSHIT.
Nice os though.
Intel rules. Getting excited over a chip that barely competes with top of the line intel chip is no big deal.
The excitement is mainly due to the widening performance gap between Apple’s G4 line and high-end x86 systems. PPC970 at least puts Apple back in the running.
You have to realize that this is the first in a line of processors IBM will be developing. It’s expected that a Power5-derived successor to PPC970 will be out next year, and will support SMT with 4 hardware threads.
The 970 is not a P4 or Itanium2. It doesn’t generate huge amounts of heat, yet it compares favorably at a lower clock speed with the P4 in raw performance that doesn’t take into account its superior SIMD implementation, and isn’t likely to be nearly as expensive as the Itanium2, allowing it to be used in desktop and laptop machines, and more practically in dense multi-CPU configurations.
Of course, much of this is speculative, given no one(that I know at least) actually has a 970, but I trust in the typical conservatism of IBM.
Hewlett-Packard Company AlphaServer GS1280 7/1150 1 1124 1482
Looks to me like its an Alpha 21364 that takes the gold as far as peak goes. Go alphas! And i dont see an PPC970s in there, so why are you comparing the two with that url? The Itanium is a mighty fine chip though, and i’d just rather see the x86 market convert to that than the AMD64.
I would rather see x86-64(amd64) die ASAP. Afterall it’s crappy architecture, and to the time we would need more than 2GB RAM(motherboard chipset limit on most cheap x86-ia-32 architectures) there will be better choices.
Much cleaner with no legacy at all, very elegant Intel Itanium. By that time I guess 2Ghz could seem realistic priced(considering it’s 64 bit) and with Hyperthreading added it gets the boost it needs to have(statical scheduling can’t use it’s power much on most applications(you can’t properly compile average application on 8 parallel instruction flow)), with HT it would achive more usage increasing the speed greatly.
Its always fun watching people with —.—.aol.com make posts about how they know whats best for computing
I can’t see any reason for AMD64 to die as the market is supporting it heavily.
in x86-64.org Linux kernel developers are working together with AMD architectures to develop an optimal kernel for AMD64.
Microsoft has also announced that it would prepare a version of its Windows (Longhorn???) for AMD64.
It would be possible for home users to buy AMD64/Desktop processors when they become available but Itanium is not designed with the home user in mind.
So I’m sure AMD will not die… as the market support will not let it to…
OK, I haven’t paid too much attention to Specmark in the past, so bear with me. I assume higher numbers mean faster speeds. In that case, it looks to me like the Dell Precision WorkStation 350 with a 3.06 GHz P4 is edged out by the IBM eServer pSeries 630 Model 6C4 with a 1450 MHz Power4+. Now these are very different kinds of computers, so the comparison may not mean a whole lot, but, all in all, I guess the megahertz myth really is that, eh?
So your point is, what? The PPC970 isn’t exactly like the Power4+, so we don’t know yet if it’s faster or slower clock-for-clock. Even if it’s slower, it’s going to be easier to ramp up the clock speed because of its design compared to the POWER4+. The reason everyone’s excited about the PPC970 is that it finally puts integer and floating-point operations on par with the P4 speed-wise, compared to the current G4 that can’t keep up except when using Altivec. But since the PPC970 also supports Altivec, the likelyhood is that the G4 is going to be toast when these new chips come out. Which is good news for anyone who isn’t already a rabid Apple hater.
Jared White
http://www.theideabasket.com
“Much cleaner with no legacy at all, very elegant Intel Itanium.”
I guess we have differing opinions on what an elegant architecture is, I tend to favor simple, efficient architectures. However the Itanium has such an “elegant” architecture that writing a semi decent compiler for it is nearly impossible. Cyndrome and Multiflow tried for a decade, and HP and then intel almost took another decade to get it to market. And guess what? it is still going nowhere, it does great on benchmarks because they can fine tune their compiler for those… but it real life apps… it blows serious monkey dung. You still have to fight HP to have them sell you a 21364 machine, because the Alpha beats the crap out of the Itanium(1 and 2)… and that is something HP doesn’t want you to know.
Oh, yeah… and as a poor person who actually had to write assmebler code for Itanium…. good luck to get code going, now you have to think vertically AND horizontally.
The problem was that VLIW (or EPIC or whatever Intel wants to rename the CLIW concept to erase the bad memories) was intended to trade hardware simplicity for software complexity (i.e simpler HW that is easier to manufacture, and spend most of the effort on the compiler). However there is certain gains from static analysis, and the traditional superscalar approaches with their far more complex HW can take advantage of dynamic scheduling. Intel found out the hard way, so they ended up increasing the complexity of the HW to make up for the shortcomings of the compiler limitations. Now they end up with an incredibly complex piece of silicon, that needs an incredibly complex piece of software to make it run properly. I do not think that is elegant, I rather adhere to the KISS principle.
The problem was that VLIW (or EPIC or whatever Intel wants to rename the CLIW concept to erase the bad memories) was intended to trade hardware simplicity for software complexity (i.e simpler HW that is easier to manufacture, and spend most of the effort on the compiler).
This might sound surprising to you I think this is the way to go. Why don’t have complex compiler and elegant CPUs? You don’t need to code in ASM, you code C++/java whatever, only one aplication is complex, and it’s name is compiler!
Besides many benefits, you don’t scrap CPU design by particular application needs(because they may change radically year after year), you have _more_ multipurpose processor design.
There are many strong software domes(from MS to IBM) that could be supported by HP to do compiler for you.
btw, KeepItSimpleStupid sucks =)
Um, that comparison has no meaning. The Power 4 has 128MB of cache! It’s got 25.6 GB/sec of bus bandwidth! The PPC 970 will not have anywhere near that amount. And I have to chastize anybody who brings up the “MHz Myth”. The marketing bullshit associated with the “MHz Myth” is so bad, that invoking “MHz Myth” is just as bad as saying “but it runs at 10 GHz!”
>This might sound surprising to you I think this is the way to go. Why don’t have complex compiler and elegant CPUs? You don’t need to code in ASM, you code C++/java whatever, only one aplication is complex, and it’s name is compiler!
Such a system can not dynamically adapt to the current workload and so must either be be compiled for ‘generic’ data (subptimal for almost all workloads) or optimized via profiling (optimal only for the profiled runs). And no one I have talked to about the IA64 do consider it a ‘elegant’ architecture, because it is not…
> Besides many benefits, you don’t scrap CPU design by particular application needs(because they may change radically year after year), you have _more_ multipurpose processor design.
Hmmm the reason processors are redesigned is because of the continuing advancements in manufacturing processes and research…
> btw, KeepItSimpleStupid sucks =)
No that is what all engineers should aim for.
IBM’s foray into desktop workstations (such as an Apple PowerMac) really has little bearing on the average Joe’s Dell running a Pentium 4. I consider an Apple a specializd piece of equipment, far better suited for graphics professionals than for the general public. Regardless of what some generic (and ultimately arbitrary) benchmark says, if the equipment does the job, that is all that matters… Unless you compare transaction speeds on database queries, all these numbers are pointless, needless and ultimately meaningless. Being a religious zealot is a waste of time and bandwidth.
You post a link to 2000 SPEC scores? Big freakin deal!
There are no “real” SPEC scores for the 970 yet as the benchmarks don’t just measure the chip but the whole machine–mobo, controllers, bandwidth, caches, memory…
We don’t know what is going into IBM’s or Apple’s design yet. IBM has released a theoretical SPEC score for a 1.4 GHz 970, and that’s it. It’s scores were very competitive.
Meanwhile, it appears the 970 is more likely to debut in volume at clock rates of 1.6, 1.8, and maybe 2.0.
i.e. it could easily be competitive or beat most Intel scores.
This is mostly contingent on bandwidth and memory used, I would say–but I expect that whatever 970 machines we see will actually be ahead of the game in terms of RapidIO, HyperTransport, or some other technology.
Your post was a joke; this is definitely a competitive processor. Will it win every benchmark out of the gate and forever more? No, but neither does Intel NOW–nevermind when this (and the soon to follow 980) appear on the scene.
> btw, KeepItSimpleStupid sucks =)
No that is what all engineers should aim for
I was refering to a KISS as a music bad, because Kiss simply sux musically.
and i still thing Itanium rox, half-year ago I was studying many matterials on IA-64 and I think it’s the future. I don’t find it that bad(nor any bad designed). Afterall it’s intell.
Can someone explain to me how a Radeon 9500 is a top of the line card? Can someone tell me when Apple started supporting ATI’s Fire GL boards? Or the NVidia Quatro Pro?
Support for both is planned for 10.3, at which point Apple will FINALLY have support for Real Production Video Cards.
Ever run Maya on a GeForce 2? omfg, the suck. Gaming boards that are outdated and twice as expensive as PC cards that are double the power are NOT, by any means, high end graphics hardware.
It’s a good thing all I use my myriad Macintoshes for is video editing, photoshop and file services. 😐
” I was studying many matterials on IA-64 and I think it’s the future. I don’t find it that bad(nor any bad designed). Afterall it’s intell.”
Could you be any more vage? There is a difference between “studying it” (whatever that means, I assumed you just glanced at Intel’s marketroid glossy panflets) and actually “using it.”
We had a few machines, and they were less than expectacular, specially if we compared their spec numbers to other older machines we were using, which seemed to run most of our code faster than the itaniums. And well, it seems to be the same experience shared by most people in industry. That is why Itaniums are selling like hotcakes… It may take them like 3 revisions to get it right, but from what it seems they will end up with another alpha at the end
.
Also just because it is intel, doesn’t mean scuat. They have great fab technology, and I believe their forte is in the process development… that is why they can push pieces of turd such as the PIV so fast. The architecture itself it is just plain horrible and ineficient. To me IA64 looks like another i860, remember it was “the future” back in 88-89.
I am afraid intel will continue pushing ia86 for the next decade or so, and Itanium will go the way of the dodo…