New GPU Crashing PC

[dippydoop]

Problem solved! Disabling Hardware Prefetch in the BIOS under CPU settings has put an end to my crash/reboots


Hi, I've been experiencing seemingly random crashes during gaming ever since I upgraded to an MSI GTX 970 GAMING 4G- after anywhere from half an hour to a few hours, my PC will just switch off and reboot without warning. I've not noticed any artifacting and my temperatures are in the lower 70's under load. Some games do not seem to cause the crashing- running War Thunder at 4k DSR for several hours was fine, but when I try the same with Skyrim, for example, I get the rebooting issue. This is on a fresh install of Windows. I'm currently on driver version 344.48, which has not changed anything with respect to the crashing.

This card is affected by the fan problems (one stuck at 100%, the other not spinning at all), however, I use MSI Afterburner to keep the fans spinning all the time, 25% at idle, gradually climbing as the temperatures rise. It seems to keep both fans working. I have also tried putting the power to 110%, but to no avail.

My previous card was a GTX 570, which ran without issue and since that card draws significantly more power than my new one, I doubt my PSU is to blame. I've run a MemtestG80 for 2048MB with 100 test iterations and came up with over 600k errors. I've done the same with the card downclocked by 500MHz on both the core and the memory and obtained just over 1 million errors.

I was hoping anyone could perhaps help ensure that the card is the problem here. My main worry is that if I send it back for an RMA, my retailer will not find the issue and send the card back to me, due to the inconsistent nature of the problem. Any advice would be greatly appreciated. Also, suggestions on benchmark programs/games to do extra tests would be welcome. I suspect the issue lies in the memory, so preferably programs that are memory hungry/intensive would be better.

My system specs are as follows:
OS: Win 7 x64
MOBO: Asus P6T Deluxe
CPU: Core I7 I920 @ ~3.7GHz
Mem: 3x2GB Corsair Dominator 1600MHz @ ~1400MHz
Storage: 240GB Sandisk Extreme Pro (boot) / 80GB Intel X-25M G2 / 1TB Samsung Spinpoint F1
PSU: 850W Thermaltake Toughpower
GPU: MSI GTX970 GAMING 4G
Sound: Asus Xonar Essence ST

I've got a similar thread on MSI's own site, but thought I may get more help here. I will keep both threads updated. Sorry if this was a bit of a dry read, I'm a bit disheartened by this problem.

Thanks for your time.

 

[Burpo]

Your board could be shutting down if VRM or chipset gets too hot (had this problem w/P6X58D). You're working that board & moving a lot more data with the new GPU.. I'd start monitoring temps.. It could also be the GPU tho..

 

[dippydoop]

Thanks for the reply, could you suggest some decent software for measuring such temps? I've been using HWiNFO64 but all I can get is some generic "Motherboard" readout, which seems to stay at 50C regardless of what I'm doing.

If it helps, I've also reset to stock clocks for the CPU (2.67GHz, and run the RAM at a very conservative 1066MHz) and recreated the issue.

 

[Burpo]

Alright.. seems MB temps not the issue.. Driver selection is pretty limited for the 970.. What all have you tried?
Move it to a dif slot?

 

[dippydoop]

I have tried different PCIe slots, as for the drivers, I've used the first available compliant driver (344.16, I think?) and am currently running 344.48.

The first time I encountered this issue, I hoped it was the driver, since it was a new architecture, but it doesn't seem to be. I also haven't seen many other people reporting a problem like this.

 

[Ketchup]

Usually a driver crash would bring you back to the desktop, not restart the PC.
A memory issue would normally show artifacts at some point.
A PC hard restart in the middle of a game usually indicates overheating or running out of power.

I did look around at that MemtestG80. Some found errors and found nothing, some found errors and we never heard from them again. So for me, the jury is still out on that one.

CPUID HW Monitor will show you the temps of your GPU. I would play a game for a minute or two and see what your max temp looks like.

Do you still have your old video card that you can use to make sure everything is still working as it should?

 

[dippydoop]

Ok, regarding the artifacting: I have noticed in Skyrim that very distant textures on the mountain tops are flashing, though only if the "camera" is panning or rotating. I've also noticed similar with War Thunder. It seems to be some kind of interaction with shadows? I actually thought this was a side-effect of running DSR, or again, immature drivers, so it never really twigged before. Still, I'm not getting anything really stark like green/purple lines/textures. It could just be that I'm looking for problems, though.

I used CPUID HW Monitor, and it reports the same temperatures as Afterburner for the GPU. Under high load, I'm hovering around 70C. All other temperatures are normal for what I've come to expect from this system.

A bit of an update regarding Skyrim: I've left my character standing in the middle of Whiterun (fairly busy place) for a couple of hours- no problem. Moving around, however does eventually cause it to restart my PC, I actually managed to get it to crash within 10 minutes of starting the game this way. To me, this suggests that the card can cope fine with high load, but has problems when changing what's in the memory. Just to confirm, I'm using a vanilla Skyrim installation, with no mods. Only the DLC and official texture packs are being used.

I do have my old GPU, and I'll put it in tomorrow. Do you have any tests in mind for this?

 

[Burpo]

You might want to try disabling Hardware Prefetch in bios under CPU settings, and see what happens..

 

[Ketchup]

You might want to try disabling Hardware Prefetch in bios under CPU settings, and see what happens..

Not a bad idea.

Your last update dippydoop does seem to support a memory problem, as it seems to crash when loading additional textures.

 

[dippydoop]

It's a bit early to tell, but disabling Hardware Prefetch does seem to have at least helped. I had a good run around Skyrim without crashing so far. Can you explain what made you suggest this in the first place? I'd like to have a better understanding of why it helped improve stability, as well as any consequences this may incur.

I managed to test my GTX 970 in a friend's PC earlier today, and didn't seem to have any issues, though I couldn't test for long. I'm reasonably confident that what I did would have caused a crash with my system (before disabling Hardware Prefetch).
For the sake of completeness, I also ran a MemtestG80 and pulled around 600k errors again. I know you mentioned that it isn't conclusive, but I was hoping for a result of 0, to perhaps highlight compatibility issues.

I'm a bit confused though, my friend's PC had Hardware Prefetch enabled, and apparently no problems. It was a Z97 based system though- maybe it's time I thought about moving on to a more modern core system.

 

[Burpo]

Just knowledge I gained playing with my overclocked X58. At the BCLK's I've been running, some programs error out to desktop. I found turning off the prefetch improved stability at higher BCLK.. Also, I could run BCLK so high that my HDD boot sector would fail or become corrupt. Disabling Prefetch (has to do with cache) cured this too.. X58 was a radical departure for intel, and very fast (in it's day), but many improvments have been made since.

Have you seen the "No Fix" errata list for this series?
http://download.intel.com/design/processor/specupdt/320836.pdf

Overclocking is the real culprit here..

 

[dippydoop]

That's a pretty scary list! But as you said, X58 incorporated a lot of new technologies for Intel. Strangely, for my system it doesn't seem as though overclocking is to blame, as I had the crashing problem even with stock clocks.

Since my last update I've been trying to make it crash, but nothing so far, so I'm reasonably confident that disabling Hardware Prefetch did the trick. Thanks a lot for helping me out- I'd never have thought of doing that!

 

[VirtualLarry]

Thanks for the tip, BURPO.

 

[Burpo]

You're welcome guys.. It's a privilege to be here on Anandtech with such great people.. Glad it worked out