IBM PowerPC 970FX wins Microprocessor Report Analysts' Choice Award for Best Desktop Processor

[beachbreeze]

What do the CPU cognoscenti make of IBMs recent 970FX achievements?

IBM 90nm G5 chip to 'outrun' Prescott, Athlon 64

"IBM's technically as yet unannounced PowerPC 970FX has won the Microprocessor Report Analysts' Choice Award for Best Desktop Processor. And according to MR editor-in-chief Peter Glaskowsky, Intel and AMD had better watch out. The 970FX is the 90nm die-shrink of the original 130nm PowerPC 970, launched back in October 2002 and finally shipped last summer, most notably inside Apple's Power Mac G5.

IBM is expected to launch the 970FX formally next week at the IEEE Solid-State Circuits Conference (ISSCC), along with its PowerTune technology, which adds SpeedStep-style clock frequency scaling that can synchronise across multiple processors. The chip should be shipping very shortly. Apple's Xserve G5, which is based on the 970FX, is due to ship this month.

IBM documentation has already revealed the 970FX's impressive power characteristics: 24.5W at 2GHz. The 130nm 970, by contrast, eats up 51W at 1.8GHz."

This adds interest to a recent Anandtech thread on heat dissipation/power consumption (I quote):

G5 970FX 2GHz : 24.5W (normal load)

G4 1.33: 45 W (normal load)
G5 2GHz: 55 W (normal load)
Athlon XP 3200+: 77 W (Thermal Design Power)
Intel's 3.2GHz P4: 82W (Thermal Design Power)
Opteron 246 2.0: 84.7W to 89 W (different AMD sources) (Thermal Design Power)
Itanium 2 1.5 6M: 107 W (Thermal Design Power)
Xeon 3.2: 110 W (Thermal Design Power)


Outside the marketing hype what real world processing performance improvement are we likely to see?
Is G5 970FX 24.5W at 2GHz laptop feasible?
How does it compare with existing solutions?

 

[beachbreeze]

Another article at ElReg continues the theme:

IBM fabs 90nm G5 using strained silicon

"Transistors made from strained silicon are said to get a 35 per cent performance boost over regular transistors of the same size. The technique works by adding a layer of silicon over a layer of silicon germanium (SiGe). The top layer's silicon atoms align themselves with those in the SiGe layer's wider-spaced crystal lattice, pulling them apart. The trouble is - as Intel has found - that it's hard to integrate the addition of that SiGe layer into existing fabrication processes.

Now, where IBM differs from Intel is in the use of silicon-on-insulator (SOI), which conveniently aids the implementation of strained silicon. As it announced last September, IBM removes the SiGe layer before fabrication, after applying the strained silicon onto the insulator. The upshot: it gains benefits of strained silicon using what is essentially its standard SOI process. By removing the SiGe layer, it doesn't have to integrate that material into the chip fabrication process per se. It calls the new technique, Strained Silicon Directly on Insulator (SSDOI)."

So, is IBM opening up a lead here, or is it hype?

 

[JeremiahTheGreat]

Since there's no way i'm going to be using a G5 in the foreseeable future.. does this bode well for future AMD processors?

 

[InlineFive]

Originally posted by: JeremiahTheGreat
Since there's no way i'm going to be using a G5 in the foreseeable future.. does this bode well for future AMD processors?

 

[beachbreeze]

Originally posted by: JeremiahTheGreat
Since there's no way i'm going to be using a G5 in the foreseeable future.. does this bode well for future AMD processors?

AMDs short-term competitor is Intel... but if IBM keeps this up... with Sony & Microsoft using IBM chips for their games boxes... who knows.

I'd really like to know what CPU literate members think of these chips & their future.