![]() Because if it did, people with Quad Crossfire would rule the scoreboard (because the Nvidia OpenCL driver currently sucks). For this additional reason the score does not weigh much in the System Score. OpenCL scales perfectly with any number of GPUs and it doesn't care if those GPUs are CPU embedded or discreet, Nvidia, AMD or Intel. This is done because of another factor as well. While the difference int he OpenCL score will be great, the difference in your System Score will NOT. Run an Nvidia card and then an AMD card and compare. The OpenCL score weighs MUCH MUCH less in the system score. Guys you are missing a very important point that I have mentioned at least 3 times already. With that said, and if I am correct, I guess I can live with the Open CL test. So I suppose, as long as your video card puts what a GTX460 does at 800/1700, it is going to score the same in OpenCL as a more powerful video card will on the same system. What that tells me is that regardless of how powerful the video card is, it is only going to perform relative to the sub-system. With the GTX680 running stock in the i5-760 machine, OpenCL score is the same as the GTX460 with the 800/1700 overclock. Since the above post, I have swaped out the GTX460 for one of my GTX680s. The GTX680 runs in RealBench I was refering to above, were in my X79 system. Keep up the good work Nodens and the rest of the developers on RealBench. Learned a lot about my system and had a ton of fun doing it. Hell, I had a blast with V1.1 thru the summer. And I realize and appreciate the time, effort, and work put into creating it. That is too large of a disparity.ĭont get me wrong. With a stock GTX680, OpenCL finishes at 31131. At a slight overclock from 725/1800 to 800/1900, OpenCl score increases to 21077. At stock speed, the 460 yields an 18007 for OpenCL. Now it is basically just another benchmark.Īn example of how OpenCL plays into scoring is the GTX460 I am using in an i5-760 system. That is what made V1.1 stand above many other benchmarks for me. There are already numerous benchmarks out there that do this. Incorporating OpenCL now adds an even larger variable, because now the overclocking potential and power of VGA comes into play. With V1.1, the only real variables were the quality of sub-system components, and the user's ability to overclock them. I realize OpenCL can be omitted for a custom run, but then it isnt an 'official' score. ![]() With V2, VGA power now comes into play, adding another variable. Realbench 1.1 appealed to me because it was a good stress test and benchmark to determine the relative power and stability of a PC's sub-system compared to others. I would like to see the OpenCL element removed from 'official' score, and have it as an option for final V2XX release.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |