Questions and Answers :
GPU applications :
Overclocking GPU - Does it have much affect with CUDA
Message board moderation
Author | Message |
---|---|
BigWaveSurfer Send message Joined: 29 Nov 01 Posts: 186 Credit: 36,311,381 RAC: 141 |
I have an EVGA Nvidia 9600GT KO 700 MHz Clock 1773 MHz Shader Clock 950 MHz Memory Clock running at 68 degrees under a Seti load with a fan speed of 56% and was curious if people thought I could push it any further or if EVGA has already done that to its max? Also, what benefits would I see on Seti if I pushed it further, would it really shorted the time for the WU's? The standard 9600 GT's specs are: 650 MHz Core Clock 1625 MHz Shader Clock 900 MHz Memory Clock Thanks! |
BigWaveSurfer Send message Joined: 29 Nov 01 Posts: 186 Credit: 36,311,381 RAC: 141 |
Anyone......? :) |
BigDaddyDave Send message Joined: 8 Oct 00 Posts: 67 Credit: 20,337,995 RAC: 15 |
Hi BigWaveSurfer, I have an "EVGA 260 216 core 55nm" and I'm currently overclocked at: Core Clock: 800 @ 56C/ 133F Shader Clock: 1725 (Linked to Core clock speed) Memory Clock: 1364 Fan Speed: 80 Please note: if you overclock you GPU to high and it's running hot, you can permanently fry your GPU. Please do so at your own risk. Here is a tip that I used, Google your graphics card to see what other websites have already reviewed and experimented with. Type the search term: review, EVGA Nvidia 9600 GT KO, overclock. I reviewed some of these websites and it look as if your GPU is pretty well overclocked already. Example review: http://www.fudzilla.com/index.php?option=com_content&task=view&id=6307&Itemid=40 So, they best way to know if your overclocked settings are working is to write down the CUDA unit completed "CPU time" at stock default speeds and then do the same at your overclocked settings speeds. It is located under the BIONC Manager "Task" tab, CPU time" Your fan setting will help in keeping your GPU running at a lower temperature which will make it more efficient but also protect it from possible permanent damage to the GPU. I started in small increments and then increased bit by bit until my screen frozen and looked terrible. I kept the fan speed at 100 the whole time. Once my GPU was running stable, I lowered the fan speed until I saw a temperature increase. Also, if you are getting BIONC "run time" errors, then you have overclocked to high. Good luck! Dave |
elgar Send message Joined: 21 May 99 Posts: 69 Credit: 2,687,478 RAC: 0 |
By definition evga has already overclocked that KO card as far as it is likely to go with stock cooling. Have you tested it to see how far the core/ram/shaders will go before locking up? If you don't mind voiding the warranty you can always put an aftermarket cooler on it, but there's probably not much oc headroom left on the card. As far was workunits go, don't know. I check the log and results from my (slightly) oc'd 260 and they seem to be accepted. Burns through a cuda unit in a little over 10 minutes as far as I can tell, don't know if its any faster than it was stock. |
BP1986 Send message Joined: 24 Mar 06 Posts: 18 Credit: 88,495 RAC: 0 |
If your cards fan setup is anything like my 8600GT cards, I would push the fan speed to %70. Anything higher just makes too much noise and offer little/no extra cooling. As for overclocking... when you do your overclocking, start up a graphically intensive game and look for video errors. If textures seem to change to different textures for some reason, it could be your mem clock is too high. Or if you have an outright crash right off, your core is too high. These were my experiences with my cards, but I also use SLI, so I dunno. :P Also, CUDA seems to not like temps over 70c. You may get alot of errors if the card gets up there in temp. |
BigWaveSurfer Send message Joined: 29 Nov 01 Posts: 186 Credit: 36,311,381 RAC: 141 |
Thanks for the info and that 'fudzilla' link was a great match for my card. It runs at 68 degrees under a full load with a fan speed of 54%. Almost exact with their results. I am thinking I might just leave it alone, I would like to push through the WU's a little faster but I am not sure it is worth the risk or errors in the WU's. It takes about 25 minutes for each one right now, if I could drop it under 20 minutes without errors I might think about it. In any case I am loving CUDA! How can you complain when it used to take hours to complete a WU to minutes?!?! It makes me want to get a new system with dual cards :) |
BigDaddyDave Send message Joined: 8 Oct 00 Posts: 67 Credit: 20,337,995 RAC: 15 |
In any case I am loving CUDA! How can you complain when it used to take hours to complete a WU to minutes?!?! It makes me want to get a new system with dual cards :) Yes it is amazing how much more efficient the GPUs are at processing SETI workloads, the SETI community is crunching away now. Minutes instead of hours! I'm curious as to how much faster the AP units will be once an AP CUDA version is released someday! Dave |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.