Message boards :
Graphics cards (GPUs) :
O/C Test Results
Message board moderation
| Author | Message |
|---|---|
ZydorSend message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]()
|
Always been wary of going too far with GPU O/C, life's too short to push the edges on this. I've never been really sure what I had was comfortably in the "safe" zone, everyones setup is slightly different, so there is a limit to "emulating" other setups. Anyway, came across a full O/C routine using a 275GTX using RivaTuner and Furmark. Hit the spot for me. It was at : http://www.guru3d.com/article/geforce-gtx-275-overclock-guide/ I started at 774/1927/2200 and immediately - once the card warmed up on the full stress test - had small, infrequent (one blip every 2 mins or so), but still there, artifacts. I went through the whole routine as recommended there, and ended up with a clean test screen, after a 15 min stress test to be sure, at 761/1907/2296, which interestingly showed a slight Furmark increase, and two degrees lower at full stress test running at 83degrees. I backed off 10Mhz on each once I reckon I had hit a clean test. Standard temp for me when GPUGRID crunching is 63 degrees (I have a mid tower and 4 disk RAID in there, so its going to be slightly higher than "expected"). Now a happy bunny - I think rofl:) Any other experiences and setup levels would be welcomed for comparison. Anyone who has not been to Guru3D, I recommend them without reservation, they are very good, and have an attached Forum populated by some serious hardware guys. I have depended on them for years. No doubt there are "better" ones around, usually are in these things, but you could do worse than keep them on the favourites list. Lets not turn this into a competition for highest card settings, its safe O/C I, and I suspect most people, am interested in. My card is a 9800GTX+ Regards Zy |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Usually I'd say a 15 mins test is not enough to guarantee stability, but FurMark is so much more demanding than anything else we throw at the cards, so if Furry is stable for 15 mins then it *should* be fine for GPU-Grid. And I have to add something: the clocks on NV GPUs are set in discrete steps and actually RivaTuner, developed / distributed by Guru3D, is one of the few tools which knows and shows this. So why does the guy not mention it in the guide? For your 9800GTX+ the shader clock should either be 1944 or 1890. Which clock is RivaTuner showing? MrS Scanning for our furry friends since Jan 2002 |
ZydorSend message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]()
|
He does mention the fact that they wrote Rivatuner and therefore recommend's it, but not known why he does not specifically mention the steps. I usually use it, however I found when I updated to the latest driver 185.85, Rivatuner claimed "No supported drivers detected for this display adapter" - presumably because it was so recent - and did not give any options to O/C, disabling the Driver settings section, only giving those for the display settings. I therefore used ASUS Smartdoctor to actually O/C. The settings I showed are a read from ASUS smartdoctor. I have found it will change the input value by a few Mhz presumably to some kind of "step" as you mentioned for Rivatuner, but it appears to do it in steps of 2-4 Mhz in terms of what you actually see. Regards Zy |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
What I want you to check: if you set the Asus tool for e.g. 1930 MHz shader clock, does RivaTuner show 1944 MHz or what else? GPU-Z and most other tools should show the 1930. MrS Scanning for our furry friends since Jan 2002 |
ZydorSend message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]()
|
Cant do it - Rivatuner does not give me access claiming it does not recognise the driver. I hear what you are saying, and its possible that ASUS smartdocter is telling "porky-pies", and in fact does only step the amounts you mentioned, not in 2-4Mhz steps. I have GPU-Z, I'll have a look and see what its claiming. If there is a way to "force" Rivatuner to see 185.85 I'm all ears. Not tried resinstalling it yet Regards Zy |
ZydorSend message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]()
|
Tried GPU-Z, it reports: GPU Clock: 761/1148/1907 (same as smartdoctor) Default Clock: 740/1100/1836 I also tried it on 1930 in ASUS Smartdoctor as you asked, GPU-Z instantly changed to 1930 without refresh. I Reverted back to the original 1907 setting I had. Regards Zy |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Ah, sorry.. I didn't consider the "too new driver" would be a problem, as it will be solved in the near future. Only problem: you'd need it now :D Well, actually you don't need the OC-option in RivaTuner. You only need the hardware monitor.. does that work? You could also downgrade your driver, but that would be asking for too much just to satisfy our curiousity. BTW: GPU-Z and the asus thingy both show only the clock speed which is requested by the system from the driver. The other tool besides RivaTuner, which is supposed to show the real clocks, is Everest. MrS Scanning for our furry friends since Jan 2002 |
ZydorSend message Joined: 8 Feb 09 Posts: 252 Credit: 1,309,451 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]()
|
No, cant get at it, probably the new driver. Its fair enough, would be kind of hairy operating on a driver they havent spec'd or tested I guess - we all want exciting lives from time to time, but thats probably too exciting for them rofl :) Regards Zy |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Ah, now I remember: when I went green last summer (since quite some time) and installed the current RivaTuner it gave me some warning, but the sensors worked anyway. But it was not about a too new driver, it was a too new card :D MrS Scanning for our furry friends since Jan 2002 |
|
Send message Joined: 15 Feb 09 Posts: 55 Credit: 3,542,733 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
EVGA Precision is a very nice OCing tool for members of the green team. I've been quite pleased with it. |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
And there's a *patch* available to make it work with any green non-EVGA card. MrS Scanning for our furry friends since Jan 2002 |
|
Send message Joined: 15 Feb 09 Posts: 55 Credit: 3,542,733 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
No patch needed for my PNY or BFG 260 216 cards. |
|
Send message Joined: 30 Jul 08 Posts: 17 Credit: 80,343,188 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]()
|
Thanks for all the information on the EVGA Precision alternative to RivaTuner, and the Furmark program for testing stability! I now feel much better that I can overclock my GTX260 a bit and not produce bad science results! Actually, does anyone know if GPUGrid is sensitive enough to detect too intense overclocking? Thanks. |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Actually, does anyone know if GPUGrid is sensitive enough to detect too intense overclocking? Definitely. You should be able to run games a higher clocks than GPU-Grid, as they tolerate more errors before they crash. However, it's not the most stressful benchmark.. I guess that's FurMark, which could almost be called a thermal virus. MrS Scanning for our furry friends since Jan 2002 |
|
Send message Joined: 18 Nov 08 Posts: 14 Credit: 30,687,791 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]()
|
You should be able to run games a higher clocks than GPU-Grid, as they tolerate more errors before they crash. If i forget to switch my OC settings back to normal when pausing GPUgrid and start a game it takes 5-25 min gaming and the graphic driver stops and reboots. Everytime. It´s not dramatic as it doesn´t crash anything, it just stops GPU traffic and after 1 min you can get back to windows and lower the clocks. And its with every game although older ones (hl2 engine) tolorate the higher clocks longer but they never run them longer than half an hour. Ergo: Even 5 year old games stress the GPU more than GPUgrid. If there is a way to "force" Rivatuner to see 185.85 I'm all ears Open Rivatuner, open tab "Power user", click the "Open database" button, open "RivaTuner.rtd". Click the + at "RivaTuner \ System" and scroll to "ForceDriverVersion". In the Value field write 18585. Be sure that the "Hexadecimal display" icon isn´t active when you type the number. RivaTuner now works normal again. End of June the new version is comming. |
|
Send message Joined: 17 Aug 08 Posts: 2705 Credit: 1,311,122,549 RAC: 0 Level ![]() Scientific publications ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Actually.. you're right. At the same temperautre I still think GPU-GRid is less tolerable to OCs than games. BUT they won't run at the same temperature, as the games use more functional units, make the card hotter and thus fail earlier. Thanks for the information. Sometimes hands-on experience still beats theory ;) MrS Scanning for our furry friends since Jan 2002 |
©2025 Universitat Pompeu Fabra