Dual Core Opterons

[0iD]

On ZD Net yesterday:
AMD has revealed that its upcoming dual-core Opteron chip
will offer a performance boost of as much as 55 percent.
http://newsletters.zdnetuk.cneteu.ne...09667/14077/0/

Runs hot but what performance! Would make the ideal workstation chip but bound to be well pricey!

[Rave]

Today's single-core Opteron, built on a manufacturing process with 130-nanometer features, measures 194 square mm, Crume said. The dual-core model, built with a 90nm process, will be within 5 percent of that size, he said.

They shouldn't cost a great deal to produce, albeit that yields might be a bit lower with double the number of transistors. Really the price will depend on whether Intel come up with anything remotely competitive, or whether AMD get to charge whatever they like for them.....

As for the heat aspect, the Tech Report reckon that 90nm chips use a lot less power when compared clock for clock with 130nm chips, possibly 25% less. Given the 89W TDP of the top end 130nm chips, by my calculations a similar dual core chip on 90mm will end up somewhere around 130-145W. AMD plan on downclocking them by 600MHz-1GHz from today's top end models to bring them down to under 100W dissipation. A decent watercooling setup ought to be able to cope with 145W of heat, so there's no particular reason why these chips couldn't be clocked back up to the 2.2-2.4GHz range.

I really want one for running FAD.....

Rich :¬)

[0iD]

Ideal for the Vapochill/Mach2 guys I recon. Could be insanely fast.

[Rave]

Just think what sort of 3DMark scores a dual proc/ dual core board on Nforce4 running SLI'd 6800Us could lay down- especially if both processors and both GPUs were on phase change.

Anyone want to buy a kidney?

Rich :¬)

[0iD]

Take my wife & children meester, really cheap.

[Mulda]

Check the comment on the 3rd page of this article about heat

[Nomadd]

Hmm... I fail to be convinced. Sure, IF you have an application that contains inherent parallelism AND is specifically - well - coded for such a chip architecture, then yes, it COULD be fast(er). On today's apps., I wouldn't see much point for these chips. And it really makes me laugh when people like AMD come out with quotes of 130-160% faster (than your old washing powder), and everyone laps it up without question... EDIT: That performance quote was on The Inq. from AMD for next year's dual-core A64.

We run plenty of dual-core stuff at work, mainly IBM POWER series stuff in their AIX boxes. It's great for running application servers and databases, but anything else just doesn't see the benefit. I mean, how many of you are running dual-CPU configs at the moment? Probably very few, as it's extra cost for NO extra performance. Dual-core is just the same, except sat on one - bloody expensive, knowing AMD's recent marketing form - die.

I'll let real-world testing on real-world applications prove me wrong, late next year. Until then, it's all hype. Oh, and I'll wait at least another year for it to come out at sensible prices, in the unlikely event it manages to prove it's worth.

Cue flames from the (blind) AMD fanboi's! :-)

Nomadd (who's owned plenty of cheap, good AMD kit in the past.)