GeForce vs. Stream #3 [dnetc on GPGPU]

引用部分をよく見ると、「HD2600XT AGP」となってますね。
ATI StreamはHD4000以上が対応だとどこかで読んだ記憶があるんですけど・・・
#追記
なんて勝手なことを思いこんでましたが、昨日のChunさんのコメントでご指摘いただいたここを見ると、2350から対応しているようですね。
　
HD4870の結果もあったので、引用。

Comment #12 From Rodrigues Nuno  2008-12-27 16:28 GMT  [reply]

Here the benchmark for HD4870(PCIe)

Dec 27 16:26:01 UTC] RC5-72: using core #0 (IL 4-pipe).
[Dec 27 16:26:13 UTC] RC5-72: Benchmark for core #0 (IL 4-pipe)
                      0.00:00:10.21 [428,757,587 keys/sec]
[Dec 27 16:26:13 UTC] RC5-72: using core #1 (IL 4-pipe cont).
[Dec 27 16:26:26 UTC] RC5-72: Benchmark for core #1 (IL 4-pipe cont)
                      0.00:00:09.93 [437,463,189 keys/sec]
[Dec 27 16:26:26 UTC] RC5-72: using core #2 (IL 4-pipe nand).
[Dec 27 16:26:38 UTC] RC5-72: Benchmark for core #2 (IL 4-pipe nand)
                      0.00:00:09.17 [481,974,451 keys/sec]

　
さらに、3850の結果も。
　

Comment #15 From MixMasta  2009-01-10 21:25 GMT  [reply]

3850 AGP with Catalyst 4.12 AGP fix drivers:

dnetc v2.9102-508-CPR-08112307 for Win32 (WindowsNT 5.1).
Please provide the *entire* version descriptor when submitting bug reports.
The distributed.net bug report pages are at http://bugs.distributed.net/

[Jan 10 21:13:25 UTC] RC5-72: using core #0 (IL 4-pipe).
[Jan 10 21:13:26 UTC] Using cached remote buffer
[Jan 10 21:13:45 UTC] RC5-72: Benchmark for core #0 (IL 4-pipe)
                      0.00:00:16.93 [98,002,114 keys/sec]
[Jan 10 21:13:45 UTC] RC5-72: using core #1 (IL 4-pipe cont).
[Jan 10 21:13:45 UTC] Using cached remote buffer
[Jan 10 21:14:03 UTC] RC5-72: Benchmark for core #1 (IL 4-pipe cont)
                      0.00:00:16.06 [101,965,395 keys/sec]
[Jan 10 21:14:03 UTC] RC5-72: using core #2 (IL 4-pipe nand).
[Jan 10 21:14:03 UTC] Using cached remote buffer
[Jan 10 21:14:22 UTC] RC5-72: Benchmark for core #2 (IL 4-pipe nand)
                      0.00:00:16.85 [103,746,333 keys/sec]

CPU usage is 100% and windows is very unresponsive while core is running. 
Please implement the CUDA type dynamic busy waiting.

　
以下にまとめると、

• 2600: 50M keys/sec
• 3850: 100M keys/sec
• 4870: 480M keys/sec

うーむ、なんか微妙〜
4000シリーズじゃなくても動くってのは明らかに朗報なんですけど、Streaming Processor当たりの搾乳量が少ないのはちょっと残念。
まぁ、それでも十二分に速いし、価格まで含めて考えるととGeForceと同等。
ある意味、つまらない、かな？