I heard some comments saying K8 architecture scales much better (non-linear, more than rate of 1, OMGWTFBBQ) after 2.4/2.5Ghz (On stuff like SuperPi) "flies" at 2.4+Ghz, but no solid references to prove so. (Nothing scales higher than 1 )

That should be the right assumption, but SuperPi as a first-run test is essentially retarded- shows no signs of how the cores work together in different situations.

This post has been edited by X.E.D: Aug 31 2007, 09:58 PM

So you care about SuperPi running on a chip on a mobo that runs HTT at 200mhz final (NOT base, bus running at 1/10 of intended speed), heavily crippled (ES CPU or bad BIOS), but not say, wait until Q4 2007 where reviewers without their own agenda (Cooaler is Intel heaven) tell you better on the chip that the consumer's buying?

Agena performance is only going to be better than Barcelona. Still, if you really want to brag (seriously, 75% above of general consensus on LYN) to everyone else that you can do single digit seconds in SuperPi and yet not across the board, there's the Core 2 Duos that you can buy now, clock to hell and back with Liquid Nitrogen, saves you the agony (if you really were aiming for SuperPi).


(And the author of the benches I think is acting dumb to whatever issues with the Barcelona benches, and proceeds to post Penryn 2.0Ghz ones instead- )

Mass volume Opteron (80+% of what the businesses will buy and compare) starts from B3.

BIOS is definitely not up to touch (even production grade rev.A boards are weeks away), and SuperPi was never as good on AMD as Intel.

It might not be fake, but there are so many variables in which the author can castrate the CPU and its performance, or he just had the "bad" B1s with discrete L2 cache disabled (only L1 and L3 shared) and/or HTT running at core link speed, not 2Ghz.

I don't think buffered RAM platforms do well with SuperPi anyway.