Just saw a link to the article on Slashdot. Good to see AI getting some notice.
This news excites me. I'm definitely anxious to see how the production intel boxes run and when they'll be released. I think my G5 is great, but having used it for two years now I don't feel that its as blazing as it used to be. Maybe its time for another gig of RAM!
BTW, I like your widget.
My other machine is a 1.2 Ghz iBook with 768 MB of RAM and an almost full 30 GB HD. So to me, my Dual G5 2.0 (1.5 GB RAM) still feels pretty blazing.
But the G5 doesn't have a "faster/wider bus". The P4's bus is a unidirection 64bit-wide bus running at 800MHz (6.4 GBps up or down). The G5's bus in a 2.0GHz model is a 1.00GHz bi-directional 32-bit bus (3.6 GBps up and/or 3.6 GBps down, 0.8GBps overhead).
I'm not at all understanding the statement "most computing tasks that are bandwidth limited are unidrectional". I can't imagine bandwidth sensitive applications where you simply suck vast amounts of data IN to a processor and have little to no output. I'd welcome education on that.
That's easy: its not true.
The MHz Myth is also not called into question by this report about the strength of a 3.6 GHz P4 compared to a 2 GHz G5. These systems have the latest Intel chipset, memories, hard disks, etc and what this developer or developers are reporting is how they handle day-to-day code which is almost all integer code... which Pentium4 and Athlon64 are much better optimized for than the G5. The G5 @ 2 GHz will stand up to a 3.6 GHz P4 (almost a 2x difference!) on properly optimized floating point code, and it will destroy the P4 on properly vectorized code (AltiVec vs. SSE3). A 2.7 GHz G5 will destroy the same P4 on the same code. Unfortunately most applications aren't properly optimized floating point or vector code, so the Pentium does very well in this comparison. Most users will be pleased with the performance of x86 Intel Macs because it is good at what most users do most of the time. Those few who care more about FPU or VPU performance are going to want to hang onto those PPC Macs for a little longer.
Nice try. A well-speced Yonah PB will only outrun a well-speced quad-core G5 as long as the application you're using does NOT make significant use of any of the following things:
Multithreading (i.e. written for multiple processors/cores)
Vector operations
Floating point operations
Memory bandwidth (the Yonah PBs are unlikely to be dual channel)
Disk access
Graphics card acceleration
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
Vector operations - I would expect the G5 to win for scientifically-targeted vector operations, and the Yonah to win for games-oriented vector operations. A lot of opportunities to vectorize games that exist in SSE3 aren't as easy to use (ie. slower) with Altivec with all the data massaging required. On the other hand, Altivec really does scream for certain imaging and scientific apps.
Floating point - G5 definitely wins from what we know so far. It's a super chip for floating point.
Memory bandwidth - Yonah, hands down. Intel is 667MHz DDR2 now, while Apple is stuck at 400MHz DDR. Bidirectional pipes to the CPU are nifty and all, but to then bottleneck them with slow RAM is silly.
Disk access - I don't see how either CPU has an advantage here. Disk speeds will be the bottleneck for virtually any bus or CPU. However, PCI-E is sure as heck better than Apple's northbridge/southbridge at shuffling the data between motherboard components here.
Graphics Card acceleration - Yonah win, easily. AGPx8 is a slow, slow bus compared to PCI-E.
okay, i am going to oversimplify here, but i think it posseses a grain of truth to it, so here goes:
if you think about it, the entire Mac OS X is an operating system that was hacked and ported to run on PowerPC chips since day one. NeXT was running on intel long ago, so why wouldn't it run sufficiently on intel hardware? it's like it's coming "home."
If you folks think the post about the internet being compiled for Windows was serious, there's something wrong with you in the head.
I don't know if he was serious or not but he's right. I just tried to print my Best Buy Reward Zone certificates and guess what? It can only be done with IE for Windows. The web page will not work with anything else. I've also noticed some pages where WMV files will not play, not because the Mac can't play the WMV files but because the page requires ActiveX. Microsoft likes to break internet standards and unfortunately idiot webmasters use Windows-only code without question.
okay, i am going to oversimplify here, but i think it posseses a grain of truth to it, so here goes:
if you think about it, the entire Mac OS X is an operating system that was hacked and ported to run on PowerPC chips since day one. NeXT was running on intel long ago, so why wouldn't it run sufficiently on intel hardware? it's like it's coming "home."
Interesting, I haven't thought of this. I don't know if this is true, but definitely sounds reasonable. Might make a good marketing spin.
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
Multi-core processors will beat Intel's (current) multithreading designs hands down in multithreading performance. A quad-core G5 or quad-processor G5 will beat Yonah system in multithreading apps pretty easy.
Quote:
Memory bandwidth - Yonah, hands down. Intel is 667MHz DDR2 now, while Apple is stuck at 400MHz DDR. Bidirectional pipes to the CPU are nifty and all, but to then bottleneck them with slow RAM is silly.
DDR2-667 performance is about the same as DDR-400 performance due to DDR2's latency problems. DDR2-667 runs on at quad-pumped 167 MHz bus to DDR-400's 200 MHz bus. If there is an improvement, there isn't much of one. There is probably a few streaming cases where DDR2-667 wins hand down, but not by much.
Intel does have an advantage in memory bandwidth and performance though, but it really doesn't have anything to do with memory specs and more to due with bus features.
Quote:
Graphics Card acceleration - Yonah win, easily. AGPx8 is a slow, slow bus compared to PCI-E.
The graphics card is more important than the bus. That is, same GPU on both systems will perform within 99% of each other regardless if its AGP 8x or PCIe. Maybe with edge conditions like SLI or 2006 graphics cards, but not now.
It might not be an "official" valid benchmark but most of us spend our days in a web browser. a 3.8ghz pc is signifigantly faster than any of my g5 dual 2.5 computers with any browser. i'll be glad when we finally have computers that have ghz that are faster. It makes HUGE noticable differences in things that most of us do daily like, email, web browsing, organizing files, and stuff that's day to day. I could care less if a g5 is 3 seconds faster at calculating some rendering...
I want a computer that FEELS faster on the front end. That responds to my clicks faster. I don't want some number on a chart to tell me it's faster...
But this has NOTHING to do with MHz. It has to do with how well the OS and the software is designed to display data. IE might feel faster than safari, but it could be just because IE starts showing the pages faster, while Safari waits until it gets more info to show data (except on my Dell box, where launching IE takes 10 seconds or more to show the home page, and you can't do anything with it while its working).
For example, Java performance on a Mac just sucks compared to a PC, but that has nothing to do with the PC having a better processor. Its because the Java VM just sucks for the mac, performance-wise. When Apple goes Intel, I don't see this changing much, as Apple will still be responsible for compiling up the VM for OS X (Sun compiles and optimizes Java for Sun and Windows and Linux [I think], but no one else).
Remember, OS X for Intel does not magically make any and all Windows or Intel compiled software magically run. I still hear people saying how this will bring us more video card options, like somehow all those video cards won't need special mac drivers or anything.
Multi-core processors will beat Intel's (current) multithreading designs hands down in multithreading performance. A quad-core G5 or quad-processor G5 will beat Yonah system in multithreading apps pretty easy.
Yonah is multi-core, not just symmetric multi-threaded, so I'm not sure what you're talking about here. A quad-core or even an octupal-core G5 will still starve waiting for data from RAM, and the northbridge required to maintain all those channels are going to run awfully hot.
Quote:
DDR2-667 performance is about the same as DDR-400 performance due to DDR2's latency problems. DDR2-667 runs on at quad-pumped 167 MHz bus to DDR-400's 200 MHz bus. If there is an improvement, there isn't much of one. There is probably a few streaming cases where DDR2-667 wins hand down, but not by much.
Intel does have an advantage in memory bandwidth and performance though, but it really doesn't have anything to do with memory specs and more to due with bus features.
The latency issue was a problem with some modules, not really the chipsets, early on in DDR2's life. It's pretty much been solved. And since DDR2 is lower voltage, it can scale faster and use less power.
You're right that the real-world difference isn't as much as the spec would indicate because of other bottlenecks, but it's still true that Apple is falling behind the motherboard curve in every category of motherboard design, and it adds up. The PowerMac G5's chipset hasn't changed significantly in almost 2 years now, while the move to Intel should provide a lot of innovation "for free".
Quote:
The graphics card is more important than the bus. That is, same GPU on both systems will perform within 99% of each other regardless if its AGP 8x or PCIe. Maybe with edge conditions like SLI or 2006 graphics cards, but not now.
That may be true for the year-old-or-more games that exist on the Macintosh today, but a modern game can nearly saturate PCI-E, which is a couple times faster than AGPx8.
It's really PCI-E that's the huge win over Apple motherboards here.
(Edited to remove extra bold in my replies to quoted text.)
Yonah is multi-core, not just symmetric multi-threaded, so I'm not sure what you're talking about here. A quad-core or even an octupal-core G5 will still starve waiting for data from RAM, and the northbridge required to maintain all those channels are going to run awfully hot.
Was replying to this:
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
You're saying that Yonah will perform better than a quad-core PPC in multithreading? That really doesn't make any sense, and I was surmising you were talking about SMT. It doesn't make any sense for Yonah to outperform a quad-core PPC (based on 970 arch.) or a quad-CPU Power Mac on multithreading performance at all. 4 cores are better than 2 (based on any reasonable assumptions for memory architectures) for multithreading.
The original comment was that a Yonah laptop would perform better than a quad-processor PPC PowerMac, which on the face of it, is ludicrus. That's quad-PPC, not quad-core. The likely scenario for such a machine is a dual 970mp machine, and no, Yonah will not outperform such a machine in multithreaded apps.
Quote:
The latency issue was a problem with some modules, not really the chipsets, early on in DDR2's life. It's pretty much been solved. And since DDR2 is lower voltage, it can scale faster and use less power.
...
You're right that the real-world difference isn't as much as the spec would indicate because of other bottlenecks, but it's still true that Apple is falling behind the motherboard curve in every category of motherboard design, and it adds up. The PowerMac G5's chipset hasn't changed significantly in almost 2 years now, while the move to Intel should provide a lot of innovation "for free".
I agree with you on Apple's lack of progress on Apple's boards, but DDR2-667, maybe even DDR2-600, with good timings will at long last solidly outperform DDR-400. Since DDR2 has come out, it hasn't been much benifit, and Apple staying with DDR-400 in PowerMacs is fine.
Now, I think that should have put in some support of DDR-500, not JEDEC approved, but it would be a real performance improvement.
Quote:
That may be true for the year-old-or-more games that exist on the Macintosh today, but a modern game can nearly saturate PCI-E, which is a couple times faster than AGPx8.
It's really PCI-E that's the huge win over Apple motherboards here.
There's a win, but it isn't huge. It's small. If a game is stressing the AGP 8x or PCIe bus, adding more local graphics memory will provide a higher performance boost. And the common practice is to increase GPU memory comiserate with the needs of the software as time moves on.
I do think that Apple should have PCIe and DDR2-667 in the next rev of PPC PowerMacs for the sake of compatibility and the performance gains they would have. That may not come to pass since they are making the switch.
I do think that Apple should have PCIe and DDR2-667 in the next rev of PPC PowerMacs for the sake of compatibility and the performance gains they would have. That may not come to pass since they are making the switch. [/B]
This has been frustrating. While I've bought G5 Powermacs for my business (since sold). I haven't as yet bought one for myself. I'm currently using two Digital Audio models which have been muchly upgraded over time.
I still have hope that Apple had designed an Express machine, and therefore will still come out with it. I don't agree that going to x86 will stop that by any means. The Powermacs may not be brought over until end of 2007. That's two and a half years from now!
If Apple expects us to continue to buy these hughly profitable machines, they MUST make them as current as possible. Two and a half years is about a typical life cycle for Apple's mobo's anyway. This is not a stretch by any means for them.
As Jobs has promised us "great new PPC products" to come, they better be a Hell of a lot more than more of the same.
Don't forget that the PCI-X bus is itself a dead end. As soon as Apple is off it (and they WILL be), there will be almost no PCI-X products available anymore. Few other manufacturers have adopted it, and those have been for low production products except for the server area. Servers rarely need hi end video cards, etc. The only reason they adopted PCI-X was because they needed a better bus than PCI, and the Express bus wasn't out yet.
Now that it is, it's been moving in pretty quickly, after a slow start. Apple had better get its act together on this or it will have almost no Powermac sales by the time the transition is complete.
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
You're saying that Yonah will perform better than a quad-core PPC in multithreading? That really doesn't make any sense, and I was surmising you were talking about SMT. It doesn't make any sense for Yonah to outperform a quad-core PPC (based on 970 arch.) or a quad-CPU Power Mac on multithreading performance at all. 4 cores are better than 2 (based on any reasonable assumptions for memory architectures) for multithreading.
The original comment was that a Yonah laptop would perform better than a quad-processor PPC PowerMac, which on the face of it, is ludicrus. That's quad-PPC, not quad-core. The likely scenario for such a machine is a dual 970mp machine, and no, Yonah will not outperform such a machine in multithreaded apps.
I think a dual-core Yonah will give a twin dual-core G5 a run for its money if you factor in motherboard chipsets, thermal limitations, etc. I still maintain, as I've said before, that four G5's will starve on the current motherboard designs, and that Apple historically has not done well designing northbridge/L3 cache designs. If four G5's starve, that means you're not going to be significantly better than the performance for two G5's. And Yonah is looking to be at least, if not more, efficient as the G5 per clock cycle and rapidly scaling up. Unless you're doing heavy floating point or Altivec, I don't see the hypothetical G5 system being all that stellar in real-world performance, even if you throw four of them at the problem. In short, I don't agree that this would be a runaway win for either side, and it's certainly not "ludicrous".
As for performance per Watt, and performance per square inch of case space, well, there's obviously no comparison there. Performance per dollar may be close, but if you factor in not having to design your own chipsets and ASICs, I think the Yonah will even win there.
I still have hope that Apple had designed an Express machine, and therefore will still come out with it. I don't agree that going to x86 will stop that by any means. The Powermacs may not be brought over until end of 2007. That's two and a half years from now!
If Apple expects us to continue to buy these hughly profitable machines, they MUST make them as current as possible. Two and a half years is about a typical life cycle for Apple's mobo's anyway. This is not a stretch by any means for them.
Exactly my thoughts. Only I don't expect Apple to release now (or in the next two-three months) new Power Macs with new motherboards. A reasonable guess for that would be MWSF 2006. But then again the scheduled end 2007 to finish the transition could be optimistic and postponed some time in 2008.
Exactly my thoughts. Only I don't expect Apple to release now (or in the next two-three months) new Power Macs with new motherboards. A reasonable guess for that would be MWSF 2006. But then again the scheduled end 2007 to finish the transition could be optimistic and postponed some time in 2008.
I'm figuring the Jan timeframe as well, but who knows?
I think a dual-core Yonah will give a twin dual-core G5 a run for its money if you factor in motherboard chipsets, thermal limitations, etc. I still maintain, as I've said before, that four G5's will starve on the current motherboard designs, and that Apple historically has not done well designing northbridge/L3 cache designs. If four G5's starve, that means you're not going to be significantly better than the performance for two G5's.
Have to violently disagree with you. If an app has 4 threads, a dual 970mp PowerMac using the current G5 motherboard will destroy a Yonah desktop in performance, all of the 970mp cores each have 1 MB on-die L2 cache afterall.
Can't really think of any edge cases where a Yonah system would handle 4 threads better than a dual 970mp system. It simply has a lot more execution resources than a Yonah system would have. Spaghetti integer code perhaps. Memory-bound applications, I don't think so since the L2 cache will alleviate that.
Quote:
And Yonah is looking to be at least, if not more, efficient as the G5 per clock cycle and rapidly scaling up. Unless you're doing heavy floating point or Altivec, I don't see the hypothetical G5 system being all that stellar in real-world performance, even if you throw four of them at the problem. In short, I don't agree that this would be a runaway win for either side, and it's certainly not "ludicrous".
The statement is for multithreaded performance. You can't get away from the fact the 4 cores are better than 2. It's a ludicrous statement to state otherwise. Really.
We all already know the relative weaknesses and strengths of the 2 processors. At the same clock rate, we would expect Yonah to be a little bit faster in integer (better OOOE and BPU), 970mp to be much faster in FPU, and 970mp faster in SIMD.
Quote:
As for performance per Watt, and performance per square inch of case space, well, there's obviously no comparison there. Performance per dollar may be close, but if you factor in not having to design your own chipsets and ASICs, I think the Yonah will even win there.
Yes of course, Yonah is great at performance/watt. It will be wonderful for Apple and it's industrial design, but a Yonah laptop ain't going to be outrunning a quad-PPC PowerMac. It's not really a prudent comparison. From the beginning:
Quote:
Quote:
Originally posted by Existence
Don't worry...buy a Yonah Powerbook instead! For most things, a dualcore Yonah PowerBook should outrun Apple's quad-PPC PowerMacs. It will be the first time in 7 years that PowerBooks are faster than PowerMacs.
Yonah laptops running faster than quad-PPC Macs? Let's not exaggerate too much. I can see a single 2.3 GHz Sossaman outrunning a single 2.5 GHz 970mp, but a dual-core laptop outrunning a quad-PPC, that's apples vs oranges.
We know that a 970mp is slated to run at 2.5 GHz, and that a Yonah will run at 2.1, maybe 2.2 GHz. I'm hard pressed to believe that a 2.2 GHz Yonah will be faster than a 2.5 GHz 970mp in anything with one exception: spaghetti integer code. Maybe that is "most" things, but a quad-PPC PowerMac would be intrinsically a specialist machine designed to run digital content creation apps and other things that would make use of the machine. And "most" things likely aren't memory-bound edge cases either.
We even should have doubts that a 2.2 GHz Yonah laptop would even outrun the dual 2 GHz PowerMac G5 currently shipping (faster in integer, slower in FPU, maybe mixed in SIMD). By the time a Mac/Intel Yonah laptop ships, I would certainly hope that Apple would be shipping a dual 2.3 GHz PowerMac as its low-end Mac/PowerPC pro desktop, making it quite debatable for a Yonah laptop to even outrun the low-end PowerMac Apple ships at the time, let alone a prospective high-end quad-PPC machine.
It's almost like saying a Yonah laptop will be faster than the dual 2.7 GHz PowerMac G5 Apple currently ships. That's pretty crazy when we know Yonah will only be at 2.2 GHz. It's ludicrous when we start saying a Yonah laptop will be faster than a quad-PPC like a dual 2.5 GHz 970mp system. One, it's not really an apt comparison. Two, a 970 is competitive to the P-M architecture at about the same clock rate and we already know that it clocks faster than P-M chips.
It's certainly nice that we can say a Mac/Intel Yonah laptop will be competitive to dual 2.0 to 2.3 GHz PowerMac G5 though. It'll run a full 30 fps for H.264/1080p video! Something I didn't think possible in a Mac laptop not 2 months ago.
Can't really think of any edge cases where a Yonah system would handle 4 threads better than a dual 970mp system. It simply has a lot more execution resources than a Yonah system would have. Spaghetti integer code perhaps. Memory-bound applications, I don't think so since the L2 cache will alleviate that.
We all already know the relative weaknesses and strengths of the 2 processors. At the same clock rate, we would expect Yonah to be a little bit faster in integer (better OOOE and BPU), 970mp to be much faster in FPU, and 970mp faster in SIMD.
Yes of course, Yonah is great at performance/watt. It will be wonderful for Apple and it's industrial design, but a Yonah laptop ain't going to be outrunning a quad-PPC PowerMac. It's not really a prudent comparison. From the beginning:
Yonah laptops running faster than quad-PPC Macs? Let's not exaggerate too much. I can see a single 2.3 GHz Sossaman outrunning a single 2.5 GHz 970mp, but a dual-core laptop outrunning a quad-PPC, that's apples vs oranges.
Quote:
We even should have doubts that a 2.2 GHz Yonah laptop would even outrun the dual 2 GHz PowerMac G5 currently shipping (faster in integer, slower in FPU, maybe mixed in SIMD). By the time a Mac/Intel Yonah laptop ships, I would certainly hope that Apple would be shipping a dual 2.3 GHz PowerMac as its low-end Mac/PowerPC pro desktop, making it quite debatable for a Yonah laptop to even outrun the low-end PowerMac Apple ships at the time, let alone a prospective high-end quad-PPC machine.
It's certainly nice that we can say a Mac/Intel Yonah laptop will be competitive to dual 2.0 to 2.3 GHz PowerMac G5 though. It'll run a full 30 fps for H.264/1080p video! Something I didn't think possible in a Mac laptop not 2 months ago.
I would like to purpose an observation that could be supported by the dev kits. We have for years been saying that the Mac is tuned for everyday life stuff but not for bench testing. Is it possible that we could look at the Intel processors as being somewhat the same in that they may be well balanced for desk top processing. They excell and running apps very well. They don't smoke at FPU they don't excel at SIMD, but they do appear to run apps very well and maybe great FPU performance is not what is needed for running desk top apps. What is needed is to deliver the perfect balance of integer to FPU to SIMD performance, and maybe Intel knows this. Maybe Apple can see that the Intel processors are better balanced. On a related note I would also expect that there are more optimizations in Unix for Intel than for PPC, it is like getting all of those optimizations for free, for PPC it was just about Apple and only Apple for the most part.
Comments
Originally posted by Xool
Just saw a link to the article on Slashdot. Good to see AI getting some notice.
This news excites me. I'm definitely anxious to see how the production intel boxes run and when they'll be released. I think my G5 is great, but having used it for two years now I don't feel that its as blazing as it used to be. Maybe its time for another gig of RAM!
BTW, I like your widget.
My other machine is a 1.2 Ghz iBook with 768 MB of RAM and an almost full 30 GB HD. So to me, my Dual G5 2.0 (1.5 GB RAM) still feels pretty blazing.
Originally posted by Gerardj
Quote:
Originally posted by Existence
But the G5 doesn't have a "faster/wider bus". The P4's bus is a unidirection 64bit-wide bus running at 800MHz (6.4 GBps up or down). The G5's bus in a 2.0GHz model is a 1.00GHz bi-directional 32-bit bus (3.6 GBps up and/or 3.6 GBps down, 0.8GBps overhead).
I'm not at all understanding the statement "most computing tasks that are bandwidth limited are unidrectional". I can't imagine bandwidth sensitive applications where you simply suck vast amounts of data IN to a processor and have little to no output. I'd welcome education on that.
That's easy: its not true.
The MHz Myth is also not called into question by this report about the strength of a 3.6 GHz P4 compared to a 2 GHz G5. These systems have the latest Intel chipset, memories, hard disks, etc and what this developer or developers are reporting is how they handle day-to-day code which is almost all integer code... which Pentium4 and Athlon64 are much better optimized for than the G5. The G5 @ 2 GHz will stand up to a 3.6 GHz P4 (almost a 2x difference!) on properly optimized floating point code, and it will destroy the P4 on properly vectorized code (AltiVec vs. SSE3). A 2.7 GHz G5 will destroy the same P4 on the same code. Unfortunately most applications aren't properly optimized floating point or vector code, so the Pentium does very well in this comparison. Most users will be pleased with the performance of x86 Intel Macs because it is good at what most users do most of the time. Those few who care more about FPU or VPU performance are going to want to hang onto those PPC Macs for a little longer.
Originally posted by bigmig
Nice try. A well-speced Yonah PB will only outrun a well-speced quad-core G5 as long as the application you're using does NOT make significant use of any of the following things:
Multithreading (i.e. written for multiple processors/cores)
Vector operations
Floating point operations
Memory bandwidth (the Yonah PBs are unlikely to be dual channel)
Disk access
Graphics card acceleration
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
Vector operations - I would expect the G5 to win for scientifically-targeted vector operations, and the Yonah to win for games-oriented vector operations. A lot of opportunities to vectorize games that exist in SSE3 aren't as easy to use (ie. slower) with Altivec with all the data massaging required. On the other hand, Altivec really does scream for certain imaging and scientific apps.
Floating point - G5 definitely wins from what we know so far. It's a super chip for floating point.
Memory bandwidth - Yonah, hands down. Intel is 667MHz DDR2 now, while Apple is stuck at 400MHz DDR. Bidirectional pipes to the CPU are nifty and all, but to then bottleneck them with slow RAM is silly.
Disk access - I don't see how either CPU has an advantage here. Disk speeds will be the bottleneck for virtually any bus or CPU. However, PCI-E is sure as heck better than Apple's northbridge/southbridge at shuffling the data between motherboard components here.
Graphics Card acceleration - Yonah win, easily. AGPx8 is a slow, slow bus compared to PCI-E.
I have yet to see it on my single processor intel devkit running OS X. How slow is iPhoto for you? Because iPhoto actually is fast on my testkit.
if you think about it, the entire Mac OS X is an operating system that was hacked and ported to run on PowerPC chips since day one. NeXT was running on intel long ago, so why wouldn't it run sufficiently on intel hardware? it's like it's coming "home."
Originally posted by BRussell
If you folks think the post about the internet being compiled for Windows was serious, there's something wrong with you in the head.
I don't know if he was serious or not but he's right. I just tried to print my Best Buy Reward Zone certificates and guess what? It can only be done with IE for Windows. The web page will not work with anything else. I've also noticed some pages where WMV files will not play, not because the Mac can't play the WMV files but because the page requires ActiveX. Microsoft likes to break internet standards and unfortunately idiot webmasters use Windows-only code without question.
Originally posted by rok
okay, i am going to oversimplify here, but i think it posseses a grain of truth to it, so here goes:
if you think about it, the entire Mac OS X is an operating system that was hacked and ported to run on PowerPC chips since day one. NeXT was running on intel long ago, so why wouldn't it run sufficiently on intel hardware? it's like it's coming "home."
Interesting, I haven't thought of this. I don't know if this is true, but definitely sounds reasonable. Might make a good marketing spin.
Originally posted by Booga
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
Multi-core processors will beat Intel's (current) multithreading designs hands down in multithreading performance. A quad-core G5 or quad-processor G5 will beat Yonah system in multithreading apps pretty easy.
Memory bandwidth - Yonah, hands down. Intel is 667MHz DDR2 now, while Apple is stuck at 400MHz DDR. Bidirectional pipes to the CPU are nifty and all, but to then bottleneck them with slow RAM is silly.
DDR2-667 performance is about the same as DDR-400 performance due to DDR2's latency problems. DDR2-667 runs on at quad-pumped 167 MHz bus to DDR-400's 200 MHz bus. If there is an improvement, there isn't much of one. There is probably a few streaming cases where DDR2-667 wins hand down, but not by much.
Intel does have an advantage in memory bandwidth and performance though, but it really doesn't have anything to do with memory specs and more to due with bus features.
Graphics Card acceleration - Yonah win, easily. AGPx8 is a slow, slow bus compared to PCI-E.
The graphics card is more important than the bus. That is, same GPU on both systems will perform within 99% of each other regardless if its AGP 8x or PCIe. Maybe with edge conditions like SLI or 2006 graphics cards, but not now.
Originally posted by webmail
It might not be an "official" valid benchmark but most of us spend our days in a web browser. a 3.8ghz pc is signifigantly faster than any of my g5 dual 2.5 computers with any browser. i'll be glad when we finally have computers that have ghz that are faster. It makes HUGE noticable differences in things that most of us do daily like, email, web browsing, organizing files, and stuff that's day to day. I could care less if a g5 is 3 seconds faster at calculating some rendering...
I want a computer that FEELS faster on the front end. That responds to my clicks faster. I don't want some number on a chart to tell me it's faster...
But this has NOTHING to do with MHz. It has to do with how well the OS and the software is designed to display data. IE might feel faster than safari, but it could be just because IE starts showing the pages faster, while Safari waits until it gets more info to show data (except on my Dell box, where launching IE takes 10 seconds or more to show the home page, and you can't do anything with it while its working).
For example, Java performance on a Mac just sucks compared to a PC, but that has nothing to do with the PC having a better processor. Its because the Java VM just sucks for the mac, performance-wise. When Apple goes Intel, I don't see this changing much, as Apple will still be responsible for compiling up the VM for OS X (Sun compiles and optimizes Java for Sun and Windows and Linux [I think], but no one else).
Remember, OS X for Intel does not magically make any and all Windows or Intel compiled software magically run. I still hear people saying how this will bring us more video card options, like somehow all those video cards won't need special mac drivers or anything.
Originally posted by THT
Multi-core processors will beat Intel's (current) multithreading designs hands down in multithreading performance. A quad-core G5 or quad-processor G5 will beat Yonah system in multithreading apps pretty easy.
Yonah is multi-core, not just symmetric multi-threaded, so I'm not sure what you're talking about here. A quad-core or even an octupal-core G5 will still starve waiting for data from RAM, and the northbridge required to maintain all those channels are going to run awfully hot.
DDR2-667 performance is about the same as DDR-400 performance due to DDR2's latency problems. DDR2-667 runs on at quad-pumped 167 MHz bus to DDR-400's 200 MHz bus. If there is an improvement, there isn't much of one. There is probably a few streaming cases where DDR2-667 wins hand down, but not by much.
Intel does have an advantage in memory bandwidth and performance though, but it really doesn't have anything to do with memory specs and more to due with bus features.
The latency issue was a problem with some modules, not really the chipsets, early on in DDR2's life. It's pretty much been solved. And since DDR2 is lower voltage, it can scale faster and use less power.
http://www.xbitlabs.com/articles/mem.../ddr2-ddr.html
You're right that the real-world difference isn't as much as the spec would indicate because of other bottlenecks, but it's still true that Apple is falling behind the motherboard curve in every category of motherboard design, and it adds up. The PowerMac G5's chipset hasn't changed significantly in almost 2 years now, while the move to Intel should provide a lot of innovation "for free".
The graphics card is more important than the bus. That is, same GPU on both systems will perform within 99% of each other regardless if its AGP 8x or PCIe. Maybe with edge conditions like SLI or 2006 graphics cards, but not now.
That may be true for the year-old-or-more games that exist on the Macintosh today, but a modern game can nearly saturate PCI-E, which is a couple times faster than AGPx8.
It's really PCI-E that's the huge win over Apple motherboards here.
(Edited to remove extra bold in my replies to quoted text.)
Originally posted by BRussell
If you folks think the post about the internet being compiled for Windows was serious, there's something wrong with you in the head.
I'd like to see you point out every time when someone is serious. There is so much posted that filtering it out is almost impossible.
Originally posted by Booga
Yonah is multi-core, not just symmetric multi-threaded, so I'm not sure what you're talking about here. A quad-core or even an octupal-core G5 will still starve waiting for data from RAM, and the northbridge required to maintain all those channels are going to run awfully hot.
Was replying to this:
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
You're saying that Yonah will perform better than a quad-core PPC in multithreading? That really doesn't make any sense, and I was surmising you were talking about SMT. It doesn't make any sense for Yonah to outperform a quad-core PPC (based on 970 arch.) or a quad-CPU Power Mac on multithreading performance at all. 4 cores are better than 2 (based on any reasonable assumptions for memory architectures) for multithreading.
The original comment was that a Yonah laptop would perform better than a quad-processor PPC PowerMac, which on the face of it, is ludicrus. That's quad-PPC, not quad-core. The likely scenario for such a machine is a dual 970mp machine, and no, Yonah will not outperform such a machine in multithreaded apps.
The latency issue was a problem with some modules, not really the chipsets, early on in DDR2's life. It's pretty much been solved. And since DDR2 is lower voltage, it can scale faster and use less power.
...
You're right that the real-world difference isn't as much as the spec would indicate because of other bottlenecks, but it's still true that Apple is falling behind the motherboard curve in every category of motherboard design, and it adds up. The PowerMac G5's chipset hasn't changed significantly in almost 2 years now, while the move to Intel should provide a lot of innovation "for free".
I agree with you on Apple's lack of progress on Apple's boards, but DDR2-667, maybe even DDR2-600, with good timings will at long last solidly outperform DDR-400. Since DDR2 has come out, it hasn't been much benifit, and Apple staying with DDR-400 in PowerMacs is fine.
Now, I think that should have put in some support of DDR-500, not JEDEC approved, but it would be a real performance improvement.
That may be true for the year-old-or-more games that exist on the Macintosh today, but a modern game can nearly saturate PCI-E, which is a couple times faster than AGPx8.
It's really PCI-E that's the huge win over Apple motherboards here.
There's a win, but it isn't huge. It's small. If a game is stressing the AGP 8x or PCIe bus, adding more local graphics memory will provide a higher performance boost. And the common practice is to increase GPU memory comiserate with the needs of the software as time moves on.
I do think that Apple should have PCIe and DDR2-667 in the next rev of PPC PowerMacs for the sake of compatibility and the performance gains they would have. That may not come to pass since they are making the switch.
Originally posted by THT
[B
I do think that Apple should have PCIe and DDR2-667 in the next rev of PPC PowerMacs for the sake of compatibility and the performance gains they would have. That may not come to pass since they are making the switch. [/B]
This has been frustrating. While I've bought G5 Powermacs for my business (since sold). I haven't as yet bought one for myself. I'm currently using two Digital Audio models which have been muchly upgraded over time.
I still have hope that Apple had designed an Express machine, and therefore will still come out with it. I don't agree that going to x86 will stop that by any means. The Powermacs may not be brought over until end of 2007. That's two and a half years from now!
If Apple expects us to continue to buy these hughly profitable machines, they MUST make them as current as possible. Two and a half years is about a typical life cycle for Apple's mobo's anyway. This is not a stretch by any means for them.
As Jobs has promised us "great new PPC products" to come, they better be a Hell of a lot more than more of the same.
Don't forget that the PCI-X bus is itself a dead end. As soon as Apple is off it (and they WILL be), there will be almost no PCI-X products available anymore. Few other manufacturers have adopted it, and those have been for low production products except for the server area. Servers rarely need hi end video cards, etc. The only reason they adopted PCI-X was because they needed a better bus than PCI, and the Express bus wasn't out yet.
Now that it is, it's been moving in pretty quickly, after a slow start. Apple had better get its act together on this or it will have almost no Powermac sales by the time the transition is complete.
Originally posted by THT
Was replying to this:
Multithreading - I would expect the Yonah to win. Intel now has way more experience with desktop dual-core than anyone else, except perhaps AMD. The quad-core G5 is a nice thought, but it would very much depend on the cache architecture to overcome the anemic RAM configuration, and Apple has shown no evidence to be clueful in this regard.
You're saying that Yonah will perform better than a quad-core PPC in multithreading? That really doesn't make any sense, and I was surmising you were talking about SMT. It doesn't make any sense for Yonah to outperform a quad-core PPC (based on 970 arch.) or a quad-CPU Power Mac on multithreading performance at all. 4 cores are better than 2 (based on any reasonable assumptions for memory architectures) for multithreading.
The original comment was that a Yonah laptop would perform better than a quad-processor PPC PowerMac, which on the face of it, is ludicrus. That's quad-PPC, not quad-core. The likely scenario for such a machine is a dual 970mp machine, and no, Yonah will not outperform such a machine in multithreaded apps.
I think a dual-core Yonah will give a twin dual-core G5 a run for its money if you factor in motherboard chipsets, thermal limitations, etc. I still maintain, as I've said before, that four G5's will starve on the current motherboard designs, and that Apple historically has not done well designing northbridge/L3 cache designs. If four G5's starve, that means you're not going to be significantly better than the performance for two G5's. And Yonah is looking to be at least, if not more, efficient as the G5 per clock cycle and rapidly scaling up. Unless you're doing heavy floating point or Altivec, I don't see the hypothetical G5 system being all that stellar in real-world performance, even if you throw four of them at the problem. In short, I don't agree that this would be a runaway win for either side, and it's certainly not "ludicrous".
As for performance per Watt, and performance per square inch of case space, well, there's obviously no comparison there. Performance per dollar may be close, but if you factor in not having to design your own chipsets and ASICs, I think the Yonah will even win there.
Originally posted by webmail
let me put it this way. When is the last time you saw the beachball on your dual 2.5 ghz machine? I see it at least twice a day...
I have yet to see it on my single processor intel devkit running OS X. How slow is iPhoto for you? Because iPhoto actually is fast on my testkit.
= Nice!
Originally posted by melgross
I still have hope that Apple had designed an Express machine, and therefore will still come out with it. I don't agree that going to x86 will stop that by any means. The Powermacs may not be brought over until end of 2007. That's two and a half years from now!
If Apple expects us to continue to buy these hughly profitable machines, they MUST make them as current as possible. Two and a half years is about a typical life cycle for Apple's mobo's anyway. This is not a stretch by any means for them.
Exactly my thoughts. Only I don't expect Apple to release now (or in the next two-three months) new Power Macs with new motherboards. A reasonable guess for that would be MWSF 2006. But then again the scheduled end 2007 to finish the transition could be optimistic and postponed some time in 2008.
Originally posted by PB
Exactly my thoughts. Only I don't expect Apple to release now (or in the next two-three months) new Power Macs with new motherboards. A reasonable guess for that would be MWSF 2006. But then again the scheduled end 2007 to finish the transition could be optimistic and postponed some time in 2008.
I'm figuring the Jan timeframe as well, but who knows?
Originally posted by Booga
I think a dual-core Yonah will give a twin dual-core G5 a run for its money if you factor in motherboard chipsets, thermal limitations, etc. I still maintain, as I've said before, that four G5's will starve on the current motherboard designs, and that Apple historically has not done well designing northbridge/L3 cache designs. If four G5's starve, that means you're not going to be significantly better than the performance for two G5's.
Have to violently disagree with you. If an app has 4 threads, a dual 970mp PowerMac using the current G5 motherboard will destroy a Yonah desktop in performance, all of the 970mp cores each have 1 MB on-die L2 cache afterall.
Can't really think of any edge cases where a Yonah system would handle 4 threads better than a dual 970mp system. It simply has a lot more execution resources than a Yonah system would have. Spaghetti integer code perhaps. Memory-bound applications, I don't think so since the L2 cache will alleviate that.
And Yonah is looking to be at least, if not more, efficient as the G5 per clock cycle and rapidly scaling up. Unless you're doing heavy floating point or Altivec, I don't see the hypothetical G5 system being all that stellar in real-world performance, even if you throw four of them at the problem. In short, I don't agree that this would be a runaway win for either side, and it's certainly not "ludicrous".
The statement is for multithreaded performance. You can't get away from the fact the 4 cores are better than 2. It's a ludicrous statement to state otherwise. Really.
We all already know the relative weaknesses and strengths of the 2 processors. At the same clock rate, we would expect Yonah to be a little bit faster in integer (better OOOE and BPU), 970mp to be much faster in FPU, and 970mp faster in SIMD.
As for performance per Watt, and performance per square inch of case space, well, there's obviously no comparison there. Performance per dollar may be close, but if you factor in not having to design your own chipsets and ASICs, I think the Yonah will even win there.
Yes of course, Yonah is great at performance/watt. It will be wonderful for Apple and it's industrial design, but a Yonah laptop ain't going to be outrunning a quad-PPC PowerMac. It's not really a prudent comparison. From the beginning:
Originally posted by Existence
Don't worry...buy a Yonah Powerbook instead! For most things, a dualcore Yonah PowerBook should outrun Apple's quad-PPC PowerMacs. It will be the first time in 7 years that PowerBooks are faster than PowerMacs.
Yonah laptops running faster than quad-PPC Macs? Let's not exaggerate too much. I can see a single 2.3 GHz Sossaman outrunning a single 2.5 GHz 970mp, but a dual-core laptop outrunning a quad-PPC, that's apples vs oranges.
We know that a 970mp is slated to run at 2.5 GHz, and that a Yonah will run at 2.1, maybe 2.2 GHz. I'm hard pressed to believe that a 2.2 GHz Yonah will be faster than a 2.5 GHz 970mp in anything with one exception: spaghetti integer code. Maybe that is "most" things, but a quad-PPC PowerMac would be intrinsically a specialist machine designed to run digital content creation apps and other things that would make use of the machine. And "most" things likely aren't memory-bound edge cases either.
We even should have doubts that a 2.2 GHz Yonah laptop would even outrun the dual 2 GHz PowerMac G5 currently shipping (faster in integer, slower in FPU, maybe mixed in SIMD). By the time a Mac/Intel Yonah laptop ships, I would certainly hope that Apple would be shipping a dual 2.3 GHz PowerMac as its low-end Mac/PowerPC pro desktop, making it quite debatable for a Yonah laptop to even outrun the low-end PowerMac Apple ships at the time, let alone a prospective high-end quad-PPC machine.
It's almost like saying a Yonah laptop will be faster than the dual 2.7 GHz PowerMac G5 Apple currently ships. That's pretty crazy when we know Yonah will only be at 2.2 GHz. It's ludicrous when we start saying a Yonah laptop will be faster than a quad-PPC like a dual 2.5 GHz 970mp system. One, it's not really an apt comparison. Two, a 970 is competitive to the P-M architecture at about the same clock rate and we already know that it clocks faster than P-M chips.
It's certainly nice that we can say a Mac/Intel Yonah laptop will be competitive to dual 2.0 to 2.3 GHz PowerMac G5 though. It'll run a full 30 fps for H.264/1080p video! Something I didn't think possible in a Mac laptop not 2 months ago.
Originally posted by THT
Have to violently disagree with you.
Had to keep this
Quote:
Can't really think of any edge cases where a Yonah system would handle 4 threads better than a dual 970mp system. It simply has a lot more execution resources than a Yonah system would have. Spaghetti integer code perhaps. Memory-bound applications, I don't think so since the L2 cache will alleviate that.
We all already know the relative weaknesses and strengths of the 2 processors. At the same clock rate, we would expect Yonah to be a little bit faster in integer (better OOOE and BPU), 970mp to be much faster in FPU, and 970mp faster in SIMD.
Yes of course, Yonah is great at performance/watt. It will be wonderful for Apple and it's industrial design, but a Yonah laptop ain't going to be outrunning a quad-PPC PowerMac. It's not really a prudent comparison. From the beginning:
Yonah laptops running faster than quad-PPC Macs? Let's not exaggerate too much. I can see a single 2.3 GHz Sossaman outrunning a single 2.5 GHz 970mp, but a dual-core laptop outrunning a quad-PPC, that's apples vs oranges.
Quote:
We even should have doubts that a 2.2 GHz Yonah laptop would even outrun the dual 2 GHz PowerMac G5 currently shipping (faster in integer, slower in FPU, maybe mixed in SIMD). By the time a Mac/Intel Yonah laptop ships, I would certainly hope that Apple would be shipping a dual 2.3 GHz PowerMac as its low-end Mac/PowerPC pro desktop, making it quite debatable for a Yonah laptop to even outrun the low-end PowerMac Apple ships at the time, let alone a prospective high-end quad-PPC machine.
It's certainly nice that we can say a Mac/Intel Yonah laptop will be competitive to dual 2.0 to 2.3 GHz PowerMac G5 though. It'll run a full 30 fps for H.264/1080p video! Something I didn't think possible in a Mac laptop not 2 months ago.
I would like to purpose an observation that could be supported by the dev kits. We have for years been saying that the Mac is tuned for everyday life stuff but not for bench testing. Is it possible that we could look at the Intel processors as being somewhat the same in that they may be well balanced for desk top processing. They excell and running apps very well. They don't smoke at FPU they don't excel at SIMD, but they do appear to run apps very well and maybe great FPU performance is not what is needed for running desk top apps. What is needed is to deliver the perfect balance of integer to FPU to SIMD performance, and maybe Intel knows this. Maybe Apple can see that the Intel processors are better balanced. On a related note I would also expect that there are more optimizations in Unix for Intel than for PPC, it is like getting all of those optimizations for free, for PPC it was just about Apple and only Apple for the most part.
Originally posted by T'hain Esh Kelch
I was just about to quote all those guys who actually saw the ironi and write "They got it!"
Damn newbs.
Irony and sarcasm rear their head in a ton of posts here people, and TEK just expressed his. Time to move on...