Message boards :
Number crunching :
7950 GPU core low frequency while crunching WUs
Message board moderation
Previous · 1 · 2 · 3 · Next
Author | Message |
---|---|
Send message Joined: 27 Jul 11 Posts: 21 Credit: 235,255,105 RAC: 0 ![]() ![]() |
2 Mumak 1.25v? - its too high voltage for 7950 core with degradation risk, my running on 1Ghz with just 1.143v now on 65C. And you always can manually adjust fan speed to stable 50-70% rpm for better cooling, i`m thinking about it. |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
Indeed, it seems pretty high - it's the default factory setting. 1.25 V is the VDDC reported, the GPU VRM reports 1.237 as the highest voltage which seems quite high to me too. I'm already using a more aggressive fan control scheme, so I guess I need to try to reduce the voltage. MSI AB doesn't seem to work with that, maybe TRiXX ? |
Send message Joined: 13 Dec 09 Posts: 10 Credit: 100,326,584 RAC: 0 ![]() ![]() |
I use Trixx for my GB Windforce 7950 and it works quote well. At 1.25v I can get my core up to 1250 without error or invalids, and if I use an agressive fan speed (80%) I can keep temps ariund 56-59C. |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
I use Trixx for my GB Windforce 7950 and it works quote well. At 1.25v I can get my core up to 1250 without error or invalids, and if I use an agressive fan speed (80%) I can keep temps ariund 56-59C. Is that with standard cooling, or do you use water cooling? I have 2x Club3D 7950 in the machine spaced by 1 free slot, so with default 1.25V the hotter one runs at ~80C, even using a higher fan speed won't cool it sufficient. The second GPU which has more space runs about 8C cooler. Even if I reduce VDDC to 1.16 V @ 1000MHz, the hotter one still runs at 75 C with default (auto) fan speed. That's still pretty high, I'm wondering how to improve that. |
Send message Joined: 27 Jul 11 Posts: 21 Credit: 235,255,105 RAC: 0 ![]() ![]() |
2 Mumak If your card cooling system looks like ![]() ![]() ![]() |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
No, mine is this: http://www.club-3d.com/index.php/products/reader.en/product/radeon-hd-7950-royalking.1314.html ![]() |
Send message Joined: 22 Aug 10 Posts: 32 Credit: 86,014,800 RAC: 0 ![]() ![]() |
Had this down-clocking issue initially after I put my Sapphire HD7950 in. I'm OC to 1200MHz in Catalyst Control but MW was only running around 850MHz. I switched projects for a day or so (to POEM, I think) which ran at the full 1200. I then tried switching back to MW and it ran at the full 1200. Never had a problem since then. I'm using the 13.4 drivers. MW units are finishing between 23 and 33 seconds. http://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=363470767 GPU load floats around 90%. Temps average mid 50s to low 60s with fans varying between 43 and 49% normally. Not sure what actually fixed the problem. I'm seeing a lot of people recommending to increase power by 20%. I'm still running on stock power setting with GPU and Memory clocks maxed out (1200 & 1575). |
![]() Send message Joined: 18 Jul 09 Posts: 300 Credit: 303,565,482 RAC: 78 ![]() ![]() ![]() |
There is no need to run the memory clock that fast for MW, all you are producing is heat. |
Send message Joined: 22 Aug 10 Posts: 32 Credit: 86,014,800 RAC: 0 ![]() ![]() |
Thanks. I use my card for a number of different projects (Milkyway, PrimeGrid, POEM, SETI, Collatz). What do you run on for memory, or do you know where I should target? |
![]() Send message Joined: 18 Jul 09 Posts: 300 Credit: 303,565,482 RAC: 78 ![]() ![]() ![]() |
I run my 5870 at 900MHz GPU clock, 300MHz memory clock. |
Send message Joined: 27 Jul 11 Posts: 21 Credit: 235,255,105 RAC: 0 ![]() ![]() |
2 Matt Very good OC results on your 7950, and yes, it can`t work on such extreme cloks with default voltage, or it was factory modified in BIOS. |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
Matt, what GPU voltage (VDDC) are you running ? |
Send message Joined: 22 Aug 10 Posts: 32 Credit: 86,014,800 RAC: 0 ![]() ![]() |
Sorry. Been away for a while. For anyone still monitoring this thread: My voltage is currently 1075, but I'm having the issue with MW running at 850MHz again, which is lower than the stock 925MHz and the 1200MHz I have it set to through CCC and used to get consistently. GPU load is always at or near 100% with one CPU core free. |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
Have you raised your GPU power limit to max ? I had the same issue and it was because the GPU was hitting the limit. DP tasks (like MW) load the GPU much more than others, which causes higher power/temperatures. Sorry. Been away for a while. For anyone still monitoring this thread: |
Send message Joined: 22 Aug 10 Posts: 32 Credit: 86,014,800 RAC: 0 ![]() ![]() |
Mumak Yes, since my last post I've increased the power limit 20% which has resulted in some improvement, with the card varying between 850MHz at 100% GPU load and 1200MHz for brief periods where the load drops to 60 - 65%. Temperature is around 60. I'm not sure what is causing the different behavior, but as I posted a while back, I used to be able stay at 1200MHz without ever dropping, and that was without adjusting the power in any way. When I first came back to this project a week or so ago, I was running POEM on my card at a constant 1200MHz and when that task finished and it started into my MW queue, the card stayed at a constant 1200 until Windows applied an update overnight and the computer rebooted without my realizing. I've been trying to test this scenario again, but POEM is not sending GPU WUs at the moment and other projects don't seem to produce the same result with this technique. POEM by default put very little load on the GPU, and as I stated above my card running MW only jumps to the full speed when my GPU load drops significantly, which it does not do when at default power target settings. Is this normal behavior? And I'm still curious as to why a card that is stock 925MHz from the manufacturer would EVER run at 850MHz under full load. |
![]() Send message Joined: 8 Apr 13 Posts: 89 Credit: 517,085,245 RAC: 0 ![]() ![]() ![]() |
Which tool have you used to check or set your GPU VDDC ? On such GPUs the VDDC is controlled using digital PWMs like CHL832x and not all tools report the correct VDDC. For me only the Sapphire TRiXX tool was able to change VDDC. You can also try the HWiNFO tool to monitor current status+VDDC ;) My 7950s had default VDDC set to 1.250V max, which was pretty high, so I reduced it to 1.150V and running stable @ 1100 MHz. |
Send message Joined: 22 Aug 10 Posts: 32 Credit: 86,014,800 RAC: 0 ![]() ![]() |
Using Catalyst Control Center to modify any GPU settings (Speed, power, fans) and GPU-Z to view usage information. |
![]() Send message Joined: 3 Mar 08 Posts: 10 Credit: 100,130,872 RAC: 349 ![]() ![]() |
Had similar problem. On my ASUS 7950 GPU freq started to jump ~ 650 MHz (only on Milkyway). I had CCC and MSI Afterburner. Both were causing problems. I solved problem by installing ASUS tweak utility. It is not so good as MSI, but it's stable. |
![]() Send message Joined: 3 Mar 08 Posts: 10 Credit: 100,130,872 RAC: 349 ![]() ![]() |
Update: ASUS tweak makes still same problem, but I noticed that freq drop is after 10 minutes...so I changed power setting for monitor to "never off". And no mo problems (for now). |
![]() Send message Joined: 3 Mar 08 Posts: 10 Credit: 100,130,872 RAC: 349 ![]() ![]() |
Update: It really helped. 5 hrs ago I have returned to MSI afterburner and it keeps all values. |
©2024 Astroinformatics Group