[Sascha Krissler] > > to fuel to competition, my GTX260 with 216 cores does 162.54 > > chains/s at 691mhz, but since it is headless i do not have to > > care for responsiveness. > > hehe.. It seems that I have trouble beating that, but I haven't > tried running in console mode yet... :)
since you have 24 more cores, with overclocking you should still surpass my gtx260 since user interface responsiveness does not limit performance very much. Updates to the screen are very fast and do not need any computing resources on the GPU, so with are working gui you loose maybe 5%, but gain 20% through overclocking. From my own experience i can say that nvidia GPUs do not let you increase the voltage over factory settings, so the increase in energy consumption does not skyrocket. With me being able to overclock 20%, there seems to be quite a margin. Be aware though that you loose your warranty. I suspect the cards blow a fuse when being overclocked, so that later it can be proven it happened. > > 353348 chains done, current rate: 153.70 chains/s > 368757 chains done, current rate: 153.78 chains/s > 384007 chains done, current rate: 153.91 chains/s > 399530 chains done, current rate: 154.08 chains/s > > Finding the max performance is fun and interessting, but I think > I have at least now found good parameters to use in different > situations. I find my computer usable at 150 chains/s... > > I would be happy to contribute data to this project, and I guess I > have not use for my GPU power more than maximum 5% of the time my > computer is powered on. of course not burning much energy also... > > So, some questions again... In your last reply you say this: > > [Terje Sannum] > >> My card is not overclocked, but can playing with these > >> performance also give me wrong values? > > [Sascha Krissler] > > yes. my gtx260 runs at 691 mhz. it is simple to compute test > > values. use --work increment to use the same start values on > > every run > > During all my testing the last days I have saved all the data > generated. I have only tweaked the --device cuda:... and the > --intermediate filter:runlength parameters. > > If my first test run with --chains 1024 is good, is the rest ok, or > should I start all over? After two days I have saved 224Mb of data... as long as --work, --condition, --roundfunc and --chainlength are the same, your chains are valid. it might still be a good idea to write to a new file. in case you used wrong options once you would have all the wrong chains isolated. ______________________________________________________ GRATIS für alle WEB.DE-Nutzer: Die maxdome Movie-FLAT! Jetzt freischalten unter http://movieflat.web.de _______________________________________________ A51 mailing list A51@lists.reflextor.com http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
