[stren]

>> Updated below, here's the original:

Hey everyone

I'm crashing while playing games, the symptoms vary from game to game. I doubt it's the cpu overclock as that's stable for bigadv with no problems

I tried reducing my gpu overclock to 850MHz and memory to 1900MHz in case the gpu wasn't stable but that didn't help, it merely prolonged how long it took to crash.

The worst offender is BFBC2 which locks up hard. Batman and DAO on the other hand will just crash out of the game and return to windows. The last crash temporarily stopped one of the gtx 460's from working, i.e. windows only recognized one card. After shutting down for ten minutes the problem went away.

I have two weak links in the system

1. PSU (Ocz 550W fatality is probably not able to handle overclocked 980x and 2 overclocked 460's)

2. the GPU VRM's - I have a waterblock that covers the gpu plus memory, but the VRM's have their original heatsink and they get pretty toasty. The airflow through that part of the case is weak and I don't have space to strap a fan on to the side. Here's a pic of the vrm heatsink:

I had been leaning towards diagnosing the PSU as the problem, but that last crash made me think it was probably over heating VRM's. Is there a thermal shutdown on the VRM's that could have caused it?

I could spend money to swap the cards for reference cards where the vrm's would be covered properly, or I could buy a new PSU, but probably can't afford both.

Any opinions on what to do?

>> Update

Putting the 20" box fan on the side of the case solved the crashing issues. I could also turn the overclock back up to 920 and dial AA back up too. I finally was able to get fps to drop. Guess I need to find a long term solution now whether it's trading in the 460s for reference ones or getting yet more molex splitters and modding the case further to add more airflow onto the gpu.

I remembered after I posted about folding that the gpu's were folding in linux and hence were back at their stock frequencies.

The gtx 460 2gb cards looked tempting (the evga one is still waterblock compatible) but at $200 each I'd rather save the money towards getting some 2nd hand 480/580's and a new psu.

 

[Forsaken_id]

I'd try whatever was needed to get a fan pointed directly at the cards just to try and rule it out. That way is cheap and may help you save some money.

I had a little different deal, but my system would crash all the time until I placed a fan to blow on my ram and then all better. Memtest found one stick bad, but the fan allowed me to get the info I needed to RMA.

 

[Chris Ihao]

Quote:

Originally Posted by Forsaken_id;13119408
I'd try whatever was needed to get a fan pointed directly at the cards just to try and rule it out. That way is cheap and may help you save some money.

I had a little different deal, but my system would crash all the time until I placed a fan to blow on my ram and then all better. Memtest found one stick bad, but the fan allowed me to get the info I needed to RMA.

Huh? In my experience cooling a broken ram chip wont help much (only way to "fix" it is to remove it).

You mean the chip simply got to hot, and hence was bad?

 

[Defoler]

Not all nvidia cards have VRM temps testing and threshold I'm afraid.
But too much heat might interfere when how they deliver the power to the GPU, and might cause you the problem.
If you can open the case and put a fan at the area and test, it might help.

And yes, a 550W PSU with a 980x at 4.3 and two OCed GPUs will defenetly might be a cause to the problem.
You can test that by resetting all OCing on the system (GPUs and CPU at full stock) and test it again.
Your system is about 600W+ if not more if you add the OC power requirments, watercooling power requirment and everything around. Most likely are you pushing the 650W+.
How much volt are you putting the system on atm (CPU and GPU)?

 

[Forsaken_id]

Quote:

Originally Posted by Chris Ihao;13119444
Huh? In my experience cooling a broken ram chip wont help much (only way to "fix" it is to remove it).

You mean the chip simply got to hot, and hence was bad?

I don't want to got OT, but it was bad, but keeping it cool to the touch would allow it enough stability to stay functional enough to load windows to get RMA info.

 

[stren]

Quote:

Originally Posted by Defoler;13119718
Not all nvidia cards have VRM temps testing and threshold I'm afraid.
But too much heat might interfere when how they deliver the power to the GPU, and might cause you the problem.
If you can open the case and put a fan at the area and test, it might help.

And yes, a 550W PSU with a 980x at 4.3 and two OCed GPUs will defenetly might be a cause to the problem.
You can test that by resetting all OCing on the system (GPUs and CPU at full stock) and test it again.
Your system is about 600W+ if not more if you add the OC power requirments, watercooling power requirment and everything around. Most likely are you pushing the 650W+.
How much volt are you putting the system on atm (CPU and GPU)?

1.0875v on the gpu and 1.38v on the cpu. I have folded with CPU at 1.5v and gpus though and it was stable back then.

I realized the easiest fan to hook up would be a twenty inch box fan, I'm out of molex connectors and this also won't load the psu any more

 

[Forsaken_id]

Max speed with ran right next to system, engage!

 

[Defoler]

Quote:

Originally Posted by stren;13119837
1.0875v on the gpu and 1.38v on the cpu. I have folded with CPU at 1.5v and gpus though and it was stable back then.

Well that is pretty much pushing it.
The OCZ might have been ok with the over-current for a bit, but it might have degraded over time, and might explain it a bit.

But yeah, add a fan and hopefully its not the PSU and just a bit of heat issue.
And still, just to make sure, run it under stock and check.

 

[stren]

Thanks everyone. Putting the 20" box fan on the side of the case solved the crashing issues. I could also turn the overclock back up to 920 and dial AA back up too. I finally was able to get fps to drop. Guess I need to find a long term solution now whether it's trading in the 460s for reference ones or getting yet more molex splitters and modding the case further to add more airflow onto the gpu.

I remembered after I posed about folding that the gpu's were folding in linux and hence were back at their stock frequencies.

The gtx 460 2gb cards looked tempting (the evga one is still waterblock compatible) but at $200 each I'd rather save the money towards getting some 2nd hand 480/580's and a new psu.