Replies to post #59885 on Advanced Micro Devices Inc (AMD)

[chipdesigner]

Alan, you're mixing up single core Opteron speeds (3GHz) with DC speeds, which would be the comparison point for Dempsey.

As for 5GHz Dempsey, I suggest you put down whatever it is you're smoking. :)

Just to be clear, my claim is: Dempsey will have a TDP of around 150W at 3.8GHz.

Just the other day the specs for reduced power Dempsey were mentioned at the Inq: 105W.

And it wouldn't make sense to bother with a separate class without a significant drop in TDP, a la 95W/68W for DC Opti regular/blade.

You might also look at Smithfield for clues.

Both the 830D (3GHz) and 840D (3.2GHz) are at 130W TDP *without* HT.

That is shooting up from 95W for the 2.8GHz part, so you know where it is on the power distribution curve.

A 3.8GHz Smithfield EE would likely exceed 200W TDP. Intel will be lucky to get the 65nm version below 150W.

[avatar]

Nice post, alan91

So, Intel will not reduce the gate oxide thickness in their shrink to 65nm? And, that will solve their leakage problems? This seems a simple solution.

Some questions:
Did Intel reduce the oxide thickness in their 90nm shrink?
If so, why?

Since maintaining the same thickness works so well why not increase it?

Is there a downside to a thicker gate oxide layer?

thanks