Original URL: http://www.theregister.co.uk/2004/05/26/ibm_ppc_975/
IBM's PowerPC 975 - verified or vapour?
Well, it's working on something...
Analysis So is IBM working on a processor called the PowerPC 975.
Take a look at the company's web site, and it appears that yes, IBM's first PowerPC processor derived from the Power 5 core will be called the PowerPC 975, company documents confirm.
Or do they? We found a reference to the CPU on IBM Taiwan's web site, but closer inspection reveals it to have been taken from a local publication, IT Home.
In essence, then, IBM is simply reprinting a rumour posted on a variety of Apple-watching web sites over the last few months.
They reckon that the 975 will be followed by a dual-core version, the 976, which is will be fabbed using a 65nm process. The 975 will be a 90nm CPU, the rumours suggest.
We're a little sceptical. IBM is having enough problems getting its 90nm 970FX out the door in volume, so it's hard to imagine it simultaneously starting to pump out chips based on an entirely new core using the same process but without the hitches. The problems is said to be experiencing with the 970FX sound like material problems rather than issues surrounding the chip design itself. That suggests the trouble lies with the 90nm process rather than the processor.
At this stage we can't rule out the existence of the 975 on IBM's roadmap, but it's interesting to note that much of the features attributed to the 975 were previously said to be components of the PowerPC 980. Nowadays, the 980 is alleged to be based on the Power 6 architecture. There's the PowerPC 990, a multi-core part based on the Power 8 coming after the 980. Previously, the 990 was a 65nm single-core part.
We've been supplied with accurate-looking roadmaps that purport to have come out of IBM, but the technology ramp is so fantastical, we remain doubtful as to their authenticity. At the very least, they're a guide to where IBM would like to be and not necessarily where it will be. Since so much can happen in just a short time in the life of a putative processor - just look at Intel's 'Tejas' cancellation - such long-term forecasts should be taken with a pinch of salt, even if they're genuine.
Does it matter? Inasmuch as IBM is almost certainly working on multi-core Power-derived PowerPC processors, no it doesn't. It's also working on a Power5-based powerPC. The fine point of what the products are actually called is irrelevant - particularly since Apple rarely acknowledges which specific chip it is using in any given system. Its admission that the Xserve G5 uses the 970FX was a rare slip-up.
Timing matters, of course. But that's as much because of the games played by x86 and PowerPC fanboys:
"My Apple's faster than your PC!"
"No! My PC's faster than your Mac!"
No! My Mac's... and so on.
Macatistas want Steve Jobs' promise to a 3GHz Power Mac for next summer to come true partly because they want to see the kit, but mostly because they don't want to Apple losing face by failing to make good its promise. Instead, they want to see Intel and AMD buffs eat crow/humble pie. Intel proponents and AMD fans want to see Apple fail to make good this promise for exactly the same reason.
This reporter would very much like to see a 3GHz Power Mac this summer, too. But like a lot of Mac users - Wintel users too, actually - he won't be rushing out to buy one. There's plenty of life left in his current system, a first-generation 12in PowerBook G4.
In any case, we look forward to Steve Jobs' Worldwide Developers Conference keynote next month with considerable interest. ®
Apple delays dualie Xserve G5 to April
IBM 90nm G5 chip to outrun Prescott, Athlon 64
IBM CPU sales rocket
IBM sells off low-end PowerPC chips
IBM fabs 90nm G5 using strained silicon
IBM dishes out Power tools for Linux
Motorola chip launch paves way for 1.5GHz PowerBook G4