GPU gets crazy hot when playing games and has weird clock patterns

[Tom Darshan]

First of all my specs:

• 
  GPU: HD 7970 Ghz
  CPU: Intel Core i5-2500k
  MB: AsRock Z68 Pro3 Gen3
  OS: Win7 64-bit Home Premium
  RAM: 12GB CL9 1333Mhz

I want to mention that I bought my computer parts on a website that puts them together and then delivers the Computer as a whole. I originally ordered it with a HD7970 but then had a problem and send the whole case back. Apparently the old GPU was broken somehow and because of a warranty I would now get a 7970 GHz edition because the prices had gotten lower.

Now the first weird thing is, it seems my GPU's max clock speeds are set at 950/1425 which doesn't
fit either of those 2 cards (1000~1050/1500 and 925/1375 respectively).

That aside my actual problems atm are:

My GPU becomes super hot* when playing games that aren't even all too demanding (Chivalry at Medium settings makes it go to a whoppin' 98 degrees celsius). *Up to over 100 degrees

Sometimes when I play games the clock speed just changes to idle (300/150) and stays there until I reboot my computer (Example: after ~10 minutes of gameplay of LoL at the highest settings ; seemingly related to the demand of the game)

In the past (~4 years) I could play these games without any problems at the highest settings

I tried googling the problem but only people that have overclocked their components seem to encounter such high temperatures and have already cleaned the whole case from dust which doesn't seem to have helped my problem in the slightest.

 

[Forgotten031511]

Your memory clock speeds appear to be quite high, as I see you did not overclock them, did the supplier of your PC do this? If so, maybe having them overclocked for so long has now caused heat damage...but I doubt that.

Can you run a Benchmark/Stresstest and post the results? (i.e. 3DMark or PassMark)

 

[Tom Darshan]

I did a quick PassMark Test and those are the results: http://www.passmark.com/baselines/V8/display.php?id=42675412857

Disregard the bad HD performance, I use a much better HD for my games.

Edit: Seems my old test was made when my gpu was "limited". Here's one where it seems to work fine (in all those tests >4 times the framrate): http://www.passmark.com/baselines/V8/display.php?id=42676164777

 

[Forgotten031511]

Please run the test a few times, if we do not get consistent results it may be your PSU not giving adequate power.

 

[Tom Darshan]

The last test was after the reboot. It's basically the problem that I decribed in the OP but I'll do some more tests just to see.

 

[Tom Darshan]

Please run the test a few times, if we do not get consistent results it may be your PSU not giving adequate power.

I repeated the test twice and the results were consistent with the last one.

 

[Forgotten031511]

The problem occurs after more usage, this points me to the direction of the PSU. What PSU do you have?

 

[Tom Darshan]

Should be a BeQuiet! Pure Power L8 600W

 

[Tom Darshan]

Okay a small update: I dismantled my case and booted up the computer and saw that 2 of my case fans weren't doing any work.
After some really long tinkering with various cables (which are all bundled up in ways that makes it close to impossible to keep track of them) I could get the two of them to work and already have experienced much better temperature (needs further investigation).

During my tinkering time I could also confirm that the GPU is a HD 7970 Sapphire and that the PSU is actually a BQ Pure Power L8 630W.

I'll now keep an eye on my benchmarks and look if the problems remain.

 

[Tom Darshan]

After some observation the temperature problems seem to be gone completely. The "throttled" speed problems however remain.

 

[Forgotten031511]

Well you have a Tier 3 PSU, which isn't horrible, and it has enough power for that system. The issue to me remains a heat issue. Now that the fans are semi-corrected, do you have the latest drivers?

Also, if the cables are that tangled, it could be blocking air-flow and adding to the heat problem. I suggest cable ties to organize your management better.

 

[Tom Darshan]

My cables are mostly tied together and into the case with cable ties. I had trouble getting to the cables because they are very tight under the cable tie. Also, my drivers seem to be the latest version (the first thing I checked naturally).

Might the # of fans be the problem? I only have one really big CPU-fan (HR-02 Macho IIRC) and 3 case fans (one on top of the case, one in the back, and one in front)

 

[Forgotten031511]

I'm still thinking the PSU is some how at fault, that should be enough fans to maintain a basic system.

My case has 2x120mm fans on the top, 2x120mm fans on the bottom, 1x120mm fan on the back, the CPU heatsink fan, and the GPU fans with their own exhaust channel out of the case...all with air filters...but I have a large case.

 

[Tom Darshan]

Is there any way to check for PSU problems like some program that can monitor jitters or something similar?

 

[Tom Darshan]

Reporting back: I tried installing a newer BIOS version, and set 1% in powertune because I heard it would help. Until now I haven't experienced any clock throttling any more but it seems the temperature problems are back. I can confirm that my fans are all spinning and all monitoring software indicates over 100 degrees celsius temperatures (after 10 minutes of playing; it continually climbs at a consistent rate from 40 degrees when playing MGS:Ground Zero)

 

[Tom Darshan]

This is such a mess: the GPU fans don't spin on the 300/150 mhz clock mode which seems fairly normal, however when the card is underload one of two event happen:

• 
  Only one fan spins
  
  No fan spins

Google search again leads to nothing useful.

Also interesting: MSI afterburner/TriXX OC Utility/CCC all show that the fans are running while they aren't.

 

[Forgotten031511]

Sorry for the delayed reply, took a technology free weekend out camping and didn't even have cell service.

Anyway, this looks to me like a faulty card at this point. If the software is showing one thing and the card is not responding correctly, something on the card is messed up.

As for testing the PSU, PassMark does a decent job of testing the overall system: http://www.passmark.com/download/pt_download.htm