OK FOLKS SOME GOOD NEWS (for some people!)
BTW: This is not my work, I just compiled info and ported it here from the thread posted on Xtremesys a couple pages back
Disabling one core per module to avoid resource sharing improves single-threaded IPC ALOT - ACROSS THE BOARD! And surely fixes some of the stranger results we've seen!
Credit to DGLee @ XS who tested this, and to chew who stepped in to confirm having similar results when he was playing around with the chip, so here go the benches improvement when comparing 2M/4C vs 4M/4C:
Fritz Chess: 39,1% improvement
wPrime 32M: 31% improvement
Winrar: 9,5% improvement
3DMark06 CPU: 5.8% improvement
3DMark Vantage CPU: 22.1% improvement
3DMark11 Physics: 14.1% improvement
Cinebench R10: 21.4% improvement
Cinebench R11.5: 19.1% improvement
Blender: 21.7% improvement
TechARP x264 enc: 20% improvement
Daum PotEncoding H264 transcoding: 11,7% improvement
-------------- EDIT TO ELLABORATE ----------------
Now there isn't a direct comparison benchmark but word going around is that these gains across the board seem to leave Bulldozer with a bit more IPC than phenom II!
This means, for people willing to turn off 4 of their threads, that bulldozer will NOT SUCK AS HARD as originally intended in gaming and lightly threaded apps (although we are still waiting on gaming benches to confirm!).. In particular, gains by this are particulary big in some of the weirder results we had seen in original reviews.
IMO this makes BD at least a bit more viable as with 4 cores disabled, you will likely be seeing close to or actual 5Ghz on air and more on water with slightly higher IPC than Phenom! And all the while lowering power draw and heat output!
Ergo, not a 2500k killer by any stretch BUT AT LEAST a viable upgrade for those who already have a 990FX board..
ATTENTION, Currently it is not known if all boards support disabling individual cores, testing was done on a CHV, can't comment on others!
SOURCE:
You must be registered to see links
.