New PowerPC 970FX, 24.5 Watts at 2GHz.

[minator]

According to IBM the new 90um die shrink of the PowerPC 970 (new revision is called 970FX) cuts the power consumption pretty severely. Previously the 1.8GHz 970 used 51 Watts whereas now the 2.0GHz 970FX uses only 24.5 Watts.

At 1.4GHz it's only 12.3 Watts, I can see a Powerbook upgrade pretty soon...

Another change in the 970FX the bus speed which has risen to 1.1GHz from 1.0GHz (giving 8.8 GigaBytes per second).

Intel on the other hand are having the opposite problem, their die shrink is sending power consumption UP.  The major difference seems to be IBMs use of SOI (Silicon On Insulator) which is preventing transistor current leakage, Intel are not using SOI so are suffering as a result, not that they appear to care.

According to rumours a while back the 970FX could go to 3GHz now if IBM wanted it to, the older 970 can go up to 2.5GHz but sucks something like 90W in the process.  I'd guess the new revision could do 3GHz but it's power useage will be something in the same region.

Future is looking good for the PowerPC.

...as is overclockability!

[downix]

Oh wow!

LAPTOP TIME!

[iamaboringperson]

Come to my new Pegasos, baby!!

I don't care what everyone else thinks, IBM definently have better technology than Intel.

Very good news :-)  :-) :-) :-) :-) :-)  :-)

[Argo]

Sounds like Intel will soon be selling 4GHz Ez-Bake Ovens. :-)

[JoannaK]

IIRC some of Intel 'soon' products were supposed to use about 150 Watts... So they don't only need totally new cooling, but new power supply and motherboard desing.

Looks good for PPC.. If we only get 970 supporting northbridge out sometime this year (summer I hope).

[SilvrDrgn]

Wow!  This sounds like a kick butt chip!!   :-)

[mikeymike]

It's good that two manufacturers are getting an idea of what the public want now.

AMD's Athlon 64 chip can clock down to 800MHz (from 3.2GHz) and not need a fan running... well, it's a start :-)

[Hammer]

IBM definently have better technology than Intel.

Note that Intel still has “Pentium M” as its 2nd backup plan. “Pentium M” is now clocked at 1.7 GHz using .13 micron process.

Also, the double pumped integer units effectively clocks some parts of Pentium VI twice the marketed rated speed.

[VidarL]

You can't compare those numbers with Intel and AMD numbers.. Those are typical power usage. Intel and AMD use TDP- Therman Design Power. The absolute maximum power they can draw.

I believe MicroDesignResources' estimation of a TDP close to 90W for a 2Ghz G5 (0.13 micron) processor is close to the truth. Which is why Apple delayed the introduction of a 1U Xserve-G5 until 90nm G5s were available.. AMD have had Opterons in 1U servers from day one.


Vidar

[Hammer]

IIRC some of Intel 'soon' products were supposed to use about 150 Watts... So they don't only need totally new cooling, but new power supply and motherboard desing.

I recall, “BTX” specs was required for some high speed Prescotts… Lower speed Prescotts (i.e. speed similar to current Northwood) would still use current Pentium IV infrastructure.

Also, SocketT has similar pin count as the AMD’s Athlon 64 processors.

[JoannaK]

According netrumours.. Prescott will be slower (on same ClockFrequency) than Northwood, so it'll need to be clocked higher to gain comparable speed..

[Hammer]

Why not try "http://www.pconline.com.cn/pchardware/tpylab/cpu/0401/298500_1.html"

http://www.pconline.com.cn/pchardware/tpylab/cpu/0401/298500_5.html
(It's in Chinese, sorry)

[Methuselas]

Why not try "http://www.pconline.com.cn/pchardware/tpylab/cpu/0401/298500_1.html"

Still a good read, for those that can. Looks quite interesting from what I could translate. :-)

[Hammer]

According netrumours.. Prescott will be slower (on same ClockFrequency) than Northwood, so it'll need to be clocked higher to gain comparable speed..

IF you are referring to ZDNET’s article on Prescott (1) then some its details is not quite right. ZDNET should stick to benchmarking the said processors…

References
1. http://zdnet.com.com/2100-1103_2-5144907.html
2. http://www.chip-architect.com/news/2003_09_21_Detailed_Architecture_of_AMDs_64bit_Core.html

PS; ZDNET(1) also got AMD's pipeline details wrong i.e. K8 should have a slightly longer pipeline than the K7, not the other way around (2).

[minator]

Those are typical power usage. Intel and AMD use TDP- Therman Design Power. The absolute maximum power they can draw.

No, TDP and absolute maximum are two diferent figures, the absolute maximum is usually quite a bit (up to 30 Watts) higher and is virtually impossible to find, at least from Intel.  Absolute maximum is a theoritical figure which may be impossible to reach in any case.

They all quote some form of average figure.