http://www.pcper.com/comments.php?nid=5679
Folding@Home NVIDIA GPU Client Confirmed
http://www.nordichardware.com/news,7777.html
GeForce GTX280 folds three times faster than Radeon HD 3870 – NordicHardware
話說要不是這篇,還不知道GeForce8系列已經賣70M個了….O_o)a
當然,平均100GFLOPS的話就是有把中低階算進去。
所以照這個數字的話,應該就是G84賣最多了。(32vs * 2FLOPS *1.35GHz = 86.4GFLOPS)

PS3 vs 3870 vs GeForce”NEXT”,這指的應該是GT200…每日500mol、高過PS3(100mol/day)和3870(170mol/day)不少。
3870在FAH執行上的C/P值(電晶體數量比例與效能)上不太妙,666M大了235M的CELL快2.8倍,記憶體頻寬的話3870也有2.25GHz 的GDDR3,72GB/s也差不多是CELL的三倍,實際性能只有1.7倍。
但是以這個觀點來說,GT200也只有和CELL同等的效率就是了….(電晶體成本和記憶體頻寬都差不多是CELL的五倍,和性能表現的比例相去不遠)
所以可以預期G84跑FAH的性能和CELL會差不多?
[EDIT]
http://www.legitreviews.com/article/713/1/
清晰照 + client screen shot
http://www.extremetech.com/article2/0,2845,2284132,00.asp
Folding@home GPU2 Client Examined
ExtremeTech: In the first GPU client, which worked through a DirectX interface, the types of work units that could be used were relatively limited. Now that DirectX is bypassed with the CAL driver for ATI cards, how have the work units been expanded? Is it comparable to the PS3 client in this regard?
Dr. Pande: Yes, going through CAL has helped a lot in several ways. The calculation is currently a superset of the PS3 1.31 client, but we hope to port some of the new code on GPU2 to the PS3 as well.
—-
http://pc.watch.impress.co.jp/docs/2008/0523/kaigai441.htm
Nehalemに見えるIntel CPUマイクロアーキテクチャの今後
本來Core MA的Loop Detector放在x86 Decoder前、現在Nehalem放在Decoder後。
雖然不是和trace cache一樣的發展方向,但是不同的出發點(Loop Detector)卻走到類似的方向,其實也蠻有趣的就是了。