PDA

View Full Version : Purely technical discussion: what is wrong with PPC?


ShadowHunter
Jun 7, 2005, 02:13 PM
Keeping the politics and personal feelings aside here, what is wrong with the PowerPC architecture? I know Motorola for years, and now IBM, has been having problems keeping yields up, why? The G5 was the world's most advanced processor and, what, 2 years later it's being practically abandoned in favor of a drastic architecture jump? IBM seemed so promising for future processors from the G5 and being able to delivery, what went wrong? I know that Intel, AMD, IBM and everyone else are all having the same issues....too much heat, power, etc.....what makes Intel the answer if they are having the same issues as IBM right now? The PowerPC, as I have understood, is generally considered superior, more robust, and better for many applications, especially scientific ones.....what has changed?

Sun Baked
Jun 7, 2005, 02:19 PM
Nothing, except that there were not enough customers for the next generation G3/G4 or the current PPC970.

That is usually death.

PowerPC is embedded and the embedded market works on different realities than the computer market.

Build a box and update it with drop in replacements over the next 6 years.

The PPC970's ElasticIO and G4-nextgen's RIO bus would require all new boxes, and as IBM found -- everybody looked at, and nobody built PPC970 widgets.

Too expensive compared to the slapping in a faster G3/G4 on the old 166MHz bus.

Dont Hurt Me
Jun 7, 2005, 02:21 PM
Bigest thing wrong with them they are slow, the second is they cant make enough of them. Third is most the software on the planet cant run on them. isnt that enough?

Le Big Mac
Jun 7, 2005, 02:37 PM
Bigest thing wrong with them they are slow, the second is they cant make enough of them. Third is most the software on the planet cant run on them. isnt that enough?

That doesn't really seem like the answer here. It's a fair question--what about the PowerPC architecture makes it harder to reduce heat and increase speed whereas Intel chips don't have the same problem.

When Jobs went with IBM for the G5 he knew that there would be production volume issues and software "issues" (which is really an OS question). But they went with them anyway. I don't really believe the problem now is that the volume pricing isn't good enough. Rather, Jobs has no confidence that IBM can scale up the speed as promised. What's the technical reason for that?

iN8
Jun 7, 2005, 02:50 PM
Bigest thing wrong with them they are slow, the second is they cant make enough of them. Third is most the software on the planet cant run on them. isnt that enough?

PowerPC's are only considered slow when comparing MHz. Intels P4, with its deep pipelines and Netburst was designed for high processor cycles at the expense of power consumption. However they did less work per cycle than their own Pentium M which runs at "slower" clock cycle, just like our beloved G5. This is why Intel has rethought their way of designing chips and has taken a cue from the M and the rest of the industy. They are moving thier future processors (Codenamed Merom, Conroe, Woodcrest) to a 64bit, multicore, lower power architecture with lower clock speeds and shorter pipelines. They plan to start at 2.5 GHz, "slower" than the latest G5.

So in essence Intel is building their next processors based on concepts that the PPC were based on. The PPC IS an excellent processor as far as I am concerned.

jamdr
Jun 7, 2005, 03:38 PM
Bigest thing wrong with them they are slow, the second is they cant make enough of them. Third is most the software on the planet cant run on them. isnt that enough?
These are pretty ridiculous, ignorant comments. For one thing, the G5 isn't slow. It consistently beat out Intel's top-of-the-line Xeon processors and was only slightly slower than the Opteron in a few tests, but faster in others. Speed was the not the issue here. The G5 is a very competitive processor in this respect.

They can't make enough of them? Neither can anyone else. The yield problems IBM was having with the G5 were also experienced by all other chip manufacturers. The past year or so has just not gone as well as IBM, Intel, or anyone else had hoped as the transition to 90 nm was more difficult than expected.

Your last comment is really the worst. Most software doesn't run on the G5? For one thing, how will a switch to Intel processors change this? It's not like Windows software will start working on Macs now. Also, this has never been an issue before. OS X as the best software on the planet, including Apple's own iApps, that fulfills the needs of just about every type of consumer. So why don't you quite the BS and say what you really mean: we all know you're obsessed with games so what you meant to say was that most games aren't available on the Mac. Whoa, what a huge revelation. No one really cares. Macs aren't gaming machines and we all knew that before we bought them.

jamdr
Jun 7, 2005, 03:43 PM
As far as I'm concerned the ONLY reason Apple is switching to Intel is because of the PowerBook G5. IBM couldn't deliver a low-power version of the G5 processor that could work in the PowerBooks and iBooks, and since laptops account for more than 50% of computers sales, this was the dealbreaker. If Apple lost more and more laptop sales because the iBook and PowerBook models became more and more obsolete (which they were), they wouldn't be able to survive on just iMac and PowerMac sales alone. It's really too bad. As much as I like the Pentium M processor, I also like the G5 processor a lot and wish Apple could continue to use it in its desktop line. I wonder how far away a PowerBook G5 really would have been. I wonder if they even had a prototype, or if it really was just too hot. Surely, a mobile G5 processor had to be somewhere on the horizon...

ajwitte
Jun 7, 2005, 05:37 PM
The PPC is a more modern, advanced architecture (RISC) with a clean design and real SIMD (AltiVec) and 64-bit support. The x86 architecture is old, even its 32-bit support is a bit kludgy (and Intel's 64-bit support barely exists), and its SIMD support consists of MMX and SSE2 (neither particularly good, though SSE2 is better). The x86 engineers, though, have managed to make the x86 run really fast, even faster than the (arguable better-designed) PPC, perhaps because of larger R&D budget.

Think of it like this: if the PPC is like a fighter jet, x86 is like a Piper Cub with a huge rocket engine bolted to the back.

Actually, I dislike the (apparent) switch from OpenFirmware to the BIOS even more than the switch from PPC to x86. I hope x86 Macs still have OpenFirmware though it appears from the developer PDF this will not be the case.

minimax
Jun 7, 2005, 06:30 PM
The lack of a G5 notebook isnt the real reason for this switch as well. Both Freescale and IBM have decent roadmaps for their respective processor lines Freescale: 8641 (667 MHZ FSB) / 8641D (dualcore) and the 64bit e700 core
IBM: 970 MP (dualcore) / 980 (POWER5 based?)

present lack of performance wasnt a convincing reason as well, as the G5 still holds up against both the Xeon and to a lesser extent the Opteron (but still acceptable by all means) although the G4 lags the Pentium M somewhat I have the feeling the pb's lack of performance is much exaggerated. Havent seen any benchmarks that support those claims. people tend to overestimate the importance of the system bus on overal system performance.

Price is not the reason as well since ppcchips are about half the price of comparable intel chips.

So what's the reason then?
Steve Jobs said it at the keynote: Apple and Intel have overlapping roadmaps, Apple needs both power and powerefficient solutions for both the notebook and the desktop and Intel, after the P4 debacle, is fully commited to this. That is a lot of R&D that will be poured into this effort.
Freescale offers the efficiency but lacks the power, and IBM offers the power but lacks the efficiency. Apple I think, is tired of the performance race against Intel and has chosen the path of least resistance.
As a sideeffect they hope to take away another barrier for people to consider the Apple, especially with the possibility to install Windows on the computer as well this will take away the necessity for a lot of people to buy a second pc. I can understand the strategy behind this move but I still consider it a defeat for Apple. They have lost their hardware independance.

petej
Jun 7, 2005, 06:34 PM
Going entirely with the guidance from Jobs, the big issue is heat. Apple are clearly planning new products where heat and power consumption are critical. That therefore implies small form factor and in some cases battery power. Think tablet, portable video, powerbook, media player as possible examples. The liquid cooling on the PM G5 is yet another costly addition that Apple would like to do without. The PPC chip itself is not fundamentally wrong it's just that for the speeds that Apple and consumers are demanding, the PPC in its current form is just too hot. IBM does not seemingly have the resources available to fund development of the chips Apple are demanding and no doubt Apple aren't prepared to fund development either. This leaves the only option which is a switch to Intel. Should IBM somehow manage to pick up the batton and achieve what currently seems impossible then I would not rule out a switch back. What Apple are now doing is showing that it has a mass market OS that is flexible about the underlying architecture. AMD could equally be added to the mix in the future. It also means that developers can easily supply optimised binaries for each architecture without the consumer having to do anything. Of course there is the small issue of Longhorn which Apple clearly have in their sights and I'm sure they would just love to challenge Longhorn on an equal basis and show a clear victory in terms of performance. Intel alos have a competitor in AMD that means they must keep on the ball and push technology hard.