Replies to post #2728 on AMD strictly moderated (AMD)

[chipguy]

And add about 10% for SLI instead of Crossfire, the former performing much better with AMD processors than the latter.

+7%.

Now add 4MB L3 for, say, 10%.

+17%.


As long as you are building silicon castles in the sky don't
forget to throw in a full bag of "PeterWatts" to help with
performance/power.

[wbmw]

Re: Right... like you said, 1 = 0. Brilliant.

You are the one saying 0 = 1. Quake had 0% improvement, as did the other apps besides F.E.A.R. And even iTunes is questionable, since a single second of difference in a measurement of 1 second resolution doesn't even necessary mean it's 1% better. There is a non-zero margin of error, you realize.

Re: sAM2 is >10% clock/clock right there. Add on another speedgrade, and that's 17%. Without 4MB of L3.

We'll revisit this when sAM2 is released, but I'm betting for the vast majority of apps, it'll be <<10% clock/clock. Secondly, if you are expecting another speed grade, that would be 3.0GHz. AMD may not even get to 3GHz without 65nm, and if they do, it will be a high wattage, boutique part against Intel's mainstream 65W part. Also, frequency doesn't scale perfectly in all apps, so that extra bin will contribute much less than 7% in most cases.

Re: And add about 10% for SLI instead of Crossfire

I don't buy it.

Re: Now add 4MB L3 for, say, 10%.

Bull. L3 cache is a ficticious item, and even if it were real, it wouldn't raise performance 10% across the board.

Looks like you're grasping at straws, Doug.