GPU 부분만 단독으로 실행가능한 OCCT의 GPU 단독 버전입니다.
해상도, 에러체킹 및 셰이더 복잡도등을 설정하여 테스트 가능합니다.

***

GPU:OCCT v0.68 (Standalone beta) by Tetedeiench

For once, i'm going to release a beta version of the next version of OCCT's brand new GPU test, that's going to be integrated soon in OCCT's main project. The test is much more efficient, sports a new error detection algorithm... you'll see, it's pretty much a revolution!

This is the achievement of 3 months of hard work, and you'll find in this post the improvement list, what you can do to help me, and what is a meaningful test for me.

I'm asking for help with this beta, make sure you read the section "What you can do to help me" please !

변경사항 :
•A new parameter "Shader Complexity" has appeared, and twikering with it can definitly change the test efficiency. Read below for a glimpse of the best values for your cards ! (and post your findings here please !)
•Thanks to this parameter, the test is much more stressfull (tests reports shows up to 5°C more than the previous version, and up to 10°C more on ATI hardware)
•Error checking is MUCH more efficient, much more stable, and done on the GPU instead on the CPU (which is MUCH better)

최적의 셰이더 복잡도 설정값 :
•Nvidia GT2x0 : 7
•Nvidia GeForce 9 series : 8 (to be confirmed)
•ATI Radeon HD3XXX and HD4XXX : 3 (by far)
•Other cards : 1 seems to be a good candidate for the Default value

알아두어야 할 것 :
•Use OCCTGPUw.exe, it is a graphical frontend that'll help you greatly in launching the test. Do not launch OCCTGPU.exe directly.
•Error check mode is restricted to 512x512, windowed mode.
•The error counter is incremented every 3 seconds, for efficiency purposes

최적의 셰이더 복잡도를 찾기위한 방법 :
•Choose a test resolution (Windowed, fullscreen, doesn't matter). Do not use error checking here. Make sure you stick to it throughtout the test. I advise you to use windowed mode, it makes things easier.
•Launch your favorite monitoring program. Do not use OCCT for that purpose. This test won't run when OCCT is running. That's because i didn't remove the integration code, it still needs to be adapted. I'd suggest HWmonitor.
•Run the test for 2 minutes to make your card a bit "hot". That's just to get us started, and to make sure the card is not ice-cold for the first test.
•For all values of Shader complexity, run the test for 3 minutes, and note down the maximum temperature you reach. If you've got other value you can monitor (especially VrmA through RivaTUner), they're perfect indicators)
•Post your information here, and tell me which one reached the highest temp. That will indicate me which one is the best value for your architecture.

 

이 글에 남겨진 댓글은 0개 입니다.

*

*