Need help with a GPU and/or PSU issue

[ComradeSader]

My first time back since 2013 and it's to ask for help, fun times This will be a bit long..

I've been having some fairly serious issues with a new (second-hand, zero issues beforehand) R9 280x GPU I got off a good mate over the weekend.
I was at his house for a LAN, not wanting to waste time I simply chucked it in (switched drivers from Nvidia to AMD) and started playing.
A few hours of CSGO, fine. A few hours more of Civ V, then I got a random "grey screen" freeze. Seems like the kinda screen you get when your overclock fails, right?

So I fired up OCCT, ran the GPU test and then around 40 seconds in, it recorded a few hundred errors in around 1 second flat. Around 50 seconds in my screen flashed black, 5-10 seconds after my system then completely rebooted.

Now, I had gotten yet another second hand card a couple of months earlier, a 580GTX. It also got errors while under OCCT stress, but after underclocking the Core somewhat, I had it stressing for 12 hours+ with zero issues. Never had any other problems. Neither cards were overclocked any higher than factory clocks/overclocks.
I tried the same with the 280x and it ran in OCCT for around 10mins before we started CSGO again. Not another freeze.

I can't remember exactly what I did the first night I got home(was still having issues of some sort), but the next night I put the card in the second PCI-E slot (thinking it might be an issue with the slot) and ran it overnight. It ran fine for about 3 hours before Windows decided to tell me my D: drive was apparently failing, minimizing the test. Restarted it and it ran fine for ~13 hours - I can't recall if I cancelled it or it was a * "Crysis2.exe has stopped responding".

I was sure it was just an issue with the PCI-E slot, thinking perhaps the top slot couldn't handle having so much power drain through it (GPUs do still pull power through the slot despite having external cords, right?) but then I realised that the second slot was only running at x8 compared to x16.
So now I thought that maybe it's a power issue, especially since it was taking the card 5~ minutes to reach 80C~ where as in the other slot it took <60 seconds.

I did some research and I read people saying that the Intel Turbo boost for their CPUs often drained a hell of a lot of power, despite the CPU barely being used in these tests, I decided to disable it and put the GPU back into the top slot - keeping it somewhat underclocked.
* My first stress lasted around 10 hours, but then OCCT "Crysis2.exe" started to crash - "become non-responsive". This is where I'm up to now, but just last night I had another "grey screen" freeze. A stress earlier yesterday resulted in a ~3min reboot with the GPU on factory clocks but CPU Turbo disabled.

I don't have temps of these tests as apparently by default OCCT doesn't record GPU temps, but I've never seen the temps go over 86C (I had RealTemp running earlier, but closed it after reading that another temp monitor maybe causing the .exe crash. Didn't help). I'm going to start another test after posting this with temps recording.
My last test last night lasted just short of 11 hours ended in the Crysis2.exe crash, after resetting to factory clocks but reducing the Power Limit (%) option in MSI Afterburner down to the minimum of -20%.

Right now I'm suspecting that it might be the PSU on it's way out (it is the oldest piece in my system, around 6 and a half years old now) since I had SOME issue with a previous, weaker GPU, and assuming that the 280x is OK as reported by my mate. I also had zero issues with a much older HD6950, which uses far less power than either of the other two.

Hopefully this actually makes sense, I've only just woken up and have to leave for a out of town daytrip shortly so I'm just getting this down as fast as possible.

I'm just out of ideas short of replacing individual parts, so I'm hoping somebody here can help point me in the right direction. Current specs are listed under my DP/in profile.

Thanks.



TLDR;
- Minor issue with 580GTX, underclock fix. Serious issue with 280x, unstable(system rebooting under stress) until reducing power usage of both GPU and CPU. Still getting rare (twice now) "grey screen" freezes afterwards.
- OCCT "Crysis2.exe" crashing after long stress tests, not sure if another form of GPU error or what. PSU is 6.5yo, thinking it might not be providing enough power for new(er) GPUs compared to my legacy HD6950 which had no issues.

 

[thebluebumblebee]

What's your fan setup in that case?

 

[ComradeSader]

It's an old case, so only two at the front and one exhaust + PSU up the top. No side fans due to the "silent" design. Everything's been air compressed out somewhat recently (mid-last month), GPU was mostly clean when I put it in.

Just ran another stress test for 52 minutes and the GPU didn't go any hotter than 69C. Idles at ~40C.

 

[hertz9753]

Did you take the side panel off to see if improves the temps? The 280x and 570 are heaters and need proper airflow.

 

[ComradeSader]

Did you take the side panel off to see if improves the temps? The 280x and 570 are heaters and need proper airflow.

Just tried it then. Power limit reverted to 0% (default) in Afterburner and the GPU hit 81C in ~100 seconds. Started getting errors at 79C, had over 200 errors by 81C then I cancelled it. Still idling at ~40 with the panel off.

I think around 83C is when it reboots, so maybe it has some form of auto shut-off in the low 80's? Though a 100%+ raise in temps during stress seems a bit much to me..
The previous test where it only went to 69C was at -20% power limited.

EDIT: Just ran it again, got to 78C and started getting errors (3k+). Got to 82C then both screens went black, came back but was still running (card had dropped to ~75C). Temp climbed back up to 82C, black screen again then I cancelled it.

EDIT2: Started again, climbed to 80C then black screen. Comes back a few seconds later at around ~75C (test still running), then it keeps repeating that process every time it hits 80C. Starts getting errors at around 78C (30k+ this time). I've got a video of it if I can upload it somewhere.

 

[Frick]

Borrow PSU from mate maybe?

 

[Ruyki]

You should not be getting any errors. If your are, expect graphic glitches/driver resets/crashes or possibly even system freezes. You can't expect your computer to run normally if your GPU makes so many calculation errors.

I would start by trying to eliminate the errors. Running your GPU at a lower frequency and or higher voltage may be able to help.

 

[hertz9753]

Have you opened that card up to clean it and replace the TIM? If it still has warranty you should think about an RMA.

Everything that you have posted about both cards tells me that your cards are hitting max temps and shutting down.

It is summer for the OP right now.

 

[ComradeSader]

Borrow PSU from mate maybe?

Managed to get a spare Corsair CX-600 from a mate today, gonna chuck it in shortly and see how we go.

You should not be getting any errors. If your are, expect graphic glitches/driver resets/crashes or possibly even system freezes. You can't expect your computer to run normally if your GPU makes so many calculation errors.

I would start by trying to eliminate the errors. Running your GPU at a lower frequency and or higher voltage may be able to help.

Though I don't really even know what the Power Limiter option in Afterburner does (doesn't explain it very well imo), lowering that to the minimum makes the GPU stable - albeit then OCCT starts crashing after X amount of hours, which I have no idea why. But that's just a temporary fix.

Have you opened that card up to clean it and replace the TIM? If it still has warranty you should think about an RMA.

Everything that you have posted about both cards tells me that your cards are hitting max temps and shutting down.

It is summer for the OP right now.

No, I haven't. If it comes to that I'll probably just give it back to my FRIEND who I got it from. He's already aware of what's going on and offered to take it back if I so choose. But still, he's said that he's had the card running up to 90C while playing Far Cry 3 when it was in his main rig, so I don't know why I'd be having issues just getting past 80C if he's had it 10C higher and running fine. This right here makes me think that it's gotta be the power supply or something else entirely, hopefully I'll find out tonight when I try this other PSU.

 

[hertz9753]

At least you called him the guy and not your friend. Another PSU will not solve your heat issues.

 

[ComradeSader]

At least you called him the guy and not your friend.

Not quite sure what you're trying to say here, just because I called him "the guy" doesn't mean he isn't my friend. Calling someone "the guy" isn't an insult. Either way, fixed.

And for clarification, Friend A is the one I lanned with and got the GPU off. The one I got the PSU off is Friend B, someone else entirely.

Another PSU will not solve your heat issues.

As I said in the last post, my FRIEND, while the card was in his main system, he was playing Far Cry 3 while the card was at around 90C with zero problems. Definitely a large implication that my problem may not actually be a heat problem. Hence why I'm trying another PSU.

 

[hertz9753]

Sorry I have only owned a few GPU's. Only about 30 of them were EVGA. I'm done in here.

 

[Ithanul]

Yeah, that sounds like heat issues for sure. I never let a GPU even break 65C. I know some radieons can handle fine at 70C for hours. But 80-90C, you are asking for a card to die on you.

Considering those are older cards and 2nd hand. The darn things probably still have their crap TIM from factory. I say take the air cooler off and smack some Gelid on them. I have a 980STRIX that would easily climb up to 70C, took the air cooler off and smacked some better TIM on since the factory loves to put to much of that crap TIM on they use.

 

[Frick]

Yeah, that sounds like heat issues for sure. I never let a GPU even break 65C. I know some radieons can handle fine at 70C for hours. But 80-90C, you are asking for a card to die on you.

OP's card can do over 90C. 95ish is the max afaik. If it breaks down at 80C something's wrong. Keeping them under 65 is ... zealous.

 

[hertz9753]

Come on Frick have you ever owned a 280x, a 7970 or a GTX 580? All of them are flame throwers and without proper cooling they will over heat and shut down.

 

[Filip Georgievski]

I got a 280x second hand, Vapor X model and it never goes above 62C at 50% fan.
You could get another 280x and test again to see if you get problems again.
If you do, change PSU. If not, get a different 280x.

 

[AsRock]

I think it sounds like a PSU issue too, i guess we will find out soon .

 

[Frick]

Come on Frick have you ever owned a 280x, a 7970 or a GTX 580? All of them are flame throwers and without proper cooling they will over heat and shut down.

They will, but not at 80C.

 

[Ithanul]

Still that some high temps. I had a 7970 that folded 24/7 with a small OC. Darn thing never broke past 70-73C. And that bugger had its stock air cooler. Plus it was stuck in my NZXT Apollo which is not the nicest air cooling case out there.

Plus not all cards are the same. Some can handle that heat, some won't. Like my 960 and 980. 960 can take the heat like a champ, but my 980 will scream bloody murder if I let it get to high in temps. My Titans on the other hand are boss at heat, use to fold air cool on one and it truck right up there at 73C. Though....the first one did kill itself. Reason I prefer to stay near 65C if possible 70-75C max for short bursts. At least keep the cards alive as long as possible.

Also had a GTX580 and GTX590 as my first cards. Those buggers are room heaters just by themselves.

 

[ComradeSader]

Apologies for the considerable delay, I've been suddenly chucked onto constant 12hr shifts and am usually suffering from the chronic "CBFs" in my down time.

First up, I plugged in the spare PSU I had; first time I turned it on it went *pop* *fizzle*. Thankfully nothing else was damaged.

Second option, was to put the card in my girlfriend's system which is almost identical to mine
(slightly older Mobo, Corsair HXi(I think?)750W PSU). She played on it for a while and it was crashing (coloured screen lockup and driver crashes), so she increased fan speed/underclocked it the same way I did. Had to roll-back to Catalyst drivers due to it "sticking" to 500mhz core clock (apparently this can cause the driver crashes?), afterwards it hasn't had any more issues apart from heat.
Running it through OCCT (stock clocks, 100% fan), the card easily gets past the ~80c mark that it went to on mine, but once it gets to 100c the core clock drops ~50% forcefully cooling it down. So yes, the card definitely has a heat issue and my mate is going to RMA it for me.

So my original problem seems to be the PSU, then? Is there any way to test it before sending it off?

 

[jaggerwild]

When you first switched cards, you went from an Nvidia to this AMD right? How did you remove the old drivers? could be a driver conflict. Try DDU to remove old drivers...........

 

[chuck216]

Apologies for the considerable delay, I've been suddenly chucked onto constant 12hr shifts and am usually suffering from the chronic "CBFs" in my down time.

First up, I plugged in the spare PSU I had; first time I turned it on it went *pop* *fizzle*. Thankfully nothing else was damaged.

Second option, was to put the card in my girlfriend's system which is almost identical to mine
(slightly older Mobo, Corsair HXi(I think?)750W PSU). She played on it for a while and it was crashing (coloured screen lockup and driver crashes), so she increased fan speed/underclocked it the same way I did. Had to roll-back to Catalyst drivers due to it "sticking" to 500mhz core clock (apparently this can cause the driver crashes?), afterwards it hasn't had any more issues apart from heat.
Running it through OCCT (stock clocks, 100% fan), the card easily gets past the ~80c mark that it went to on mine, but once it gets to 100c the core clock drops ~50% forcefully cooling it down. So yes, the card definitely has a heat issue and my mate is going to RMA it for me.

So my original problem seems to be the PSU, then? Is there any way to test it before sending it off?

OK that card is definitely having heating issues if it gets to 100c inside your GF's case especially @100% fan. My Asus R9 390 Strix DCU3 with fans set to 100% idles @ 30c and never goes above 65c even under a full load under furmark.

While poor ventilation inside the case may be a partial cause you're literally boiling your VRMs.

 

[ComradeSader]

When you first switched cards, you went from an Nvidia to this AMD right? How did you remove the old drivers? could be a driver conflict. Try DDU to remove old drivers...........

Would a driver conflict cause my GPU to seemingly shut down for 2 seconds during a stress test? I can definitely try that, just seems weird.

OK that card is definitely having heating issues if it gets to 100c inside your GF's case especially @100% fan. My Asus R9 390 Strix DCU3 with fans set to 100% idles @ 30c and never goes above 65c even under a full load under furmark.

While poor ventilation inside the case may be a partial cause you're literally boiling your VRMs.

Yep, just what I was thinking. Definitely getting my mate to RMA it, it's always been a hot card but 100c is ridiculous.

That's surely not what my original problem is though, right? It was only getting slightly past 80c before the GPU cut out for a couple seconds.