K404 Posted January 27, 2013 Share Posted January 27, 2013 As title. I've just had a Kepler card maxx out at 999MHz, but I could only tell by checking Thermspy & nVidiaInspector.... as well as noticing that my scores weren't scaling. Most (all?) >1GHz scores I see are verified in the screenshot with GPU-Z only and they're listing 1100,1200MHz etc.... but the card (very probably) isn't running at those speeds Quote Link to comment Share on other sites More sharing options...
Guest cowgut Posted January 28, 2013 Share Posted January 28, 2013 you can also just hit the render test in gpuz to get the true speed in the screen shot Quote Link to comment Share on other sites More sharing options...
K404 Posted January 28, 2013 Author Share Posted January 28, 2013 Here's a question Why is the render test a separate download? I've never downloaded that..... will it report 999Mhz under load? Quote Link to comment Share on other sites More sharing options...
xxbassplayerxx Posted January 29, 2013 Share Posted January 29, 2013 I'm pretty sure it's the new Nvidia drivers that limit the frequency. Roll back to ~2 months ago and you shouldn't get this limit. Quote Link to comment Share on other sites More sharing options...
K404 Posted January 29, 2013 Author Share Posted January 29, 2013 What are the first drivers that support Kepler? Quote Link to comment Share on other sites More sharing options...
Guest cowgut Posted January 29, 2013 Share Posted January 29, 2013 I think 301.40 or so But i still dont follow you,you should see the clocks boost during 3d with an osd. When it lists 1350 in gpuz for my 680 card/bios it will boost to 1450+ Quote Link to comment Share on other sites More sharing options...
K404 Posted January 29, 2013 Author Share Posted January 29, 2013 (edited) That's what I thought. 300-series is the only series of drivers for Kepler. Exactly..... with boost I ***suspect*** that nVidias method of implementing Boost has had...."consequences" for cards that don't. 620/630/640/650/Ti don't Boost. "We all know" that 300-series drivers with Fermi cards limit the REAL MHz to 999, regardless of what GPU-Z reports. GPU-Z reports what is set in nVidiaInspector/Afterburner/whatever, but nVidiaInspector shows "current clock" as 1000MHz. Thermspy says the same. Most importantly, performance says the same. Now, I am having the same problem with a non-boosting Kepler card. Here is an example using Fermi-5. Check the MHz Vs the score. Thermspy is missing from the second batch but still, the conclusion should be fairly obvious http://hwbot.org/forum/showthread.php?t=63538&page=2 Edited January 29, 2013 by K404 Quote Link to comment Share on other sites More sharing options...
K404 Posted January 30, 2013 Author Share Posted January 30, 2013 Wait a second.... I just read a forum post on a totally unrelated matter, then rechecked my GT630.... My GT630 (ALL GT630?) doesn't use Kepler silicon, it's a GF108 die, ie Fermi, which explains why i'm locked at 999MHz. Ok, that explains one problem, but the validation problem is still present because of the MHz cap not being shown in Riva. Quote Link to comment Share on other sites More sharing options...
K404 Posted March 11, 2013 Author Share Posted March 11, 2013 Bump for this. Ignore the GT630 discussion. Apart from GPU-Z MONITORING page (GPU MHz set to "Show Max") and nVidiaInspector (Est. max) fields, what ways are there to confirm Kepler MHz? Quote Link to comment Share on other sites More sharing options...
der8auer Posted March 11, 2013 Share Posted March 11, 2013 The question is whether the correct MHz is needed to have a valid result? I'm not realy a fan of specific rules for each hardware. This just makes the bot even more complicated and as long as the result is fine I don't think it's a must to have the correct MHz. Quote Link to comment Share on other sites More sharing options...
K404 Posted March 11, 2013 Author Share Posted March 11, 2013 Personally, I disagree I hate seeing the number of results going up with MHz info missing. How many cheats are caught because the MHz give away that something is wrong? If MHz are "muddy" then it would be very easy to run D-PP or Mip (if that even works on Kepler) and claim the MHz were much higher, they just aren't being shown in GPU-Z. IMO, it undermines the point of having GPU-Z verification..... if I have a 6.8GHz Ivy for 3D05 or AM3....... does it matter if i'm using a GTX580/680/Titan? Quote Link to comment Share on other sites More sharing options...
Bobnova Posted March 11, 2013 Share Posted March 11, 2013 I agree. If MHz aren't shown correctly GPUz is mostly pointless. Quote Link to comment Share on other sites More sharing options...
Oj0 Posted March 11, 2013 Share Posted March 11, 2013 I'm not sure I follow? My best 580 scales past 1,000 MHz, all the way to 1,050 MHz where it craps out on air cooling. Quote Link to comment Share on other sites More sharing options...
K404 Posted March 11, 2013 Author Share Posted March 11, 2013 There's two conversations goin on in this thread.... that's my fault OjO... what drivers? Are you getting better FPS with increased MHz? I believe i've shown beyond any doubt that Fermi + 3-series drivers crap out at 999MHz. GPU-Z can be ignored, it is not reflecting the truth. Quote Link to comment Share on other sites More sharing options...
Oj0 Posted March 11, 2013 Share Posted March 11, 2013 Ah I see, I thought you meant on all drivers. I was on either 285.xx or 296.xx, I was using AM3 as reference for performance relative to frequency. Quote Link to comment Share on other sites More sharing options...
K404 Posted March 11, 2013 Author Share Posted March 11, 2013 no no.... please don't confuse things- this is a 3-series driver problem More dedicated thread: http://hwbot.org/forum/showthread.php?t=63538&highlight=1075MHz Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.