Found this article through Physxinfo http://www.geeks3d.com/20120502/physx-test-gtx-680-vs-gtx-580-vs-gtx-480-in-fluidmark/ 680 vs 580 vs 480 in fluidmark: 120000 SPH particles, 1920×1080 fullscreen
Considering it's double the performance, I'd like to see some proper CUDA based benchmarks between the 680 and 580. Don't think the OpenCL benchmarks show what the 680 can really do. Although GK110 will probably be a much less crippled version for GeForce Kepler.
the only thing I see wrong is the 480 is at stock, the 580 has a 25 mhz oc and the 680 is at 1097mhz where stock is 1006. all at stock or all similarily oc would be better
The 680's clock could have been from boost in the benchmark. Would be interesting to see how they rank clock for clock, although with the 680's score I'd still imagine it would be ahead.
Ah, now I get it, shouldn't add to the score since multithread is off. Wonder why he left it in the testbed though.
Well OpenCL is one thing, but directompute benchmark show the same.. in SP performance and fluidmark physx, the 680 do well, but when it come to DP, things is over. The DP rate of the 680 is lower of the 580... I will rather see too more physX and cuda benchmark for get the result.
Yeah after a few reviews comparing the 680's SP to the 580 in CUDA, it flies well ahead. The card is too crippled when it comes to DP though. PhysX works well but I'd like to see the uncrippled Tesla GK110 in benchmarks compared to the 20-series