Weird 7970M fan problem

Hi guys - I'm a bit concerned about an issue I've been having with my 7970M.
When I'm in a game everything seems OK for a couple of minutes and then the game begins to slow and speed up over and over. This is reproducible across numerous games.
When I minimize the game and maximize again the problem goes away but then starts up shortly after I start playing again.

I think I know what the issue is: When run EVGA Precision/MSI Afterburner, I am seeing frequent dips in the GPU utilization oscillating between 99% and 63%, which seems to correspond to the slowdown/speedup I'm seeing in game. The GPU utilization is dipping whenever the GPU temp exceeds 102C. The fan for the GPU comes on, but only very slightly compared with in the past where it would come on very loudly. So, I think the problem is with the fan not turning on appropriately.

Here's a screenshot from EVGA Precision. imgur: the simple image sharer

Any ideas?
Thanks!


Edit: Forgot to mention - I'm using a P170EM with the 04.06.05 BIOS update under Windows 8 Pro.

 

This is called thermal throttling. In simple terms, it means your video card is too hot. There is something wrong with your cooling sytem, there is no way it should be that hot. At 110C your CPU core will be cooked, so your system is protecting itself. Pretty sure people don't exceed 85C with the 7970M on stock. For comparison, my 680M (same 100W power draw) only hits 90C when overclocked and running burn testing.

 

Thanks for your reply! Do you know why the GPU fan might not be turning on properly?

 

Stop using your computer right now. Your video card idles at 79C from the screenshot in your other reply. Looks like your fan is dying, that's the only way it's gonna get that hot unless you made the heatsink lose contact.

If it's still under warranty, send it in and have it fixed.

 

No I think that's actually the CPU temperature you're looking at in the Afterburner screen: http://i.imgur.com/Ioxi2Jz.png

I opened up my computer and I can't really see anything wrong. I just wanted to make sure that this isn't a driver issue before I RMA.

 

Well, your CPU is thoroughly cooked at 102C. You should've triggered a shutdown at 100C for the CPU. Either way, if something overheats your machine will respond with throttling or shutdown. Since the CPU is the one that is overheating, I find it highly unlikely a driver made your fan go wimpy. Besides, 79C idle for a CPU is not safe.

 

Sorry, I think I may have confused things somewhere. The temperature of 79C that you saw in the Afterburner screenshot corresponds to the CPU, and the max temperature of the CPU is around 102C during gameplay, which yeah is really hot, but the fan is going like crazy. Right now the CPU is idling at 53C (which still seems a little warm to me).

The GPU is what was reaching 102C and was being throttled in response. The problem seems to be that the GPU fan isn't coming on appropriately, but the GPU isn't in use when I'm not gaming.

 

I think your best bet for now would be repasting both CPU and GPU and cleaning heatsinks out of dust.

 

But it seems as though the fan isn't even coming on, which you'd think it would if it was detecting that it was getting hot.

 

This. You need to open up your computer and see if something is blocking the fan. This really shouldn't be happening. If you don't know what you're doing, I suggest using your warranty.

 

Okay sounds good. I opened it up and there was nothing obviously obstructing the fan. I'll try a few more things and then RMA if need be. Thanks!

 

If the fan is not coming on, there are 2 possibilities. One, the fan is dead. You can remove it use an adapter to plug it into a desktop or a power supply (bench top power supply, 5V should do) and see if it spins at all. Two, the fan is not getting enough juice from the motherboard. Use a multimeter to measure the voltage of the fan connector (kinda tricky since it's so small). Either way, it's best to just send it in under warranty.

You can try to reflash the EC firmware and use the computer with a cooler to see how the fans behave. Maybe the EC firmware got corrupted, highly doubt it.

 

OK so good news and bad news. The good news is that I seem to have fixed it. When I do FurMark stress tests the temp won't go above 91C and in games it won't go abover 86C. Still a little higher than I would like, but I'll repaste with new thermal compound. Whoever did it originally dd a really bad job. Anyway, the fans come on to varying degrees now, which is nice. Strange though: I seem to recall the GPU fan being louder...

I opened up everything and turns out the heat sink was loaded with cat hair, so the bad news is that I'm a moron.

Does anyone know of a software that can be used to manually control the GPU fan speeds? Afterburner and Precision only give me temperatures and GPU utilizations.

Thanks for all your help

 

Those temps are really good for stress test programs, I wouldnt bother with redoing it, you can't really hope for much more.

 

Those temps look good now. To get more fan control and more oc potential, check out the vbios, bios, and ec from Prema on the tech infernal forums. You will need to join, and help in various threads before you can download though.

 

Pretty sure manual fan control requires modded EC firmware, you have a HM or EM?

 

After reading this thread in detail I have only one thing to say.

Get rid of the PUSSE....cat.

:thumbsup: ...

 

It's the EM.

 

EM doesn't have the manual override for fans, I do not what else

 

You can always flash Prema's modded BIOS and enjoy Fn+1 max fan (don't forget to put the headphones on though ). Also, the fan profiles in Prema's modded BIOS are a bit more aggressive than stock BIOS ones, at least in the P170EM.