PC crashed several times when GPU was under load.

Toggle sidebar Toggle sidebar
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630
I overclocked my XFX 7870 Ghz Edition (Tahiti Le) because I just got a BenQ XL2720Z. I used Unigine Heaven to benchmark it and make sure it was stable but my PC has crashed several times when playing games. It isn't because of heat because I've never seen the GPU go past 75C. Should I increase the power limit or lower the clock?
PC specs:
CPU: Intel i5 4570
GPU: XFX 7870 Ghz Edition
PSU: Corsair CX600M
Mobo: Gigabyte GA H87 HD3
Case: Fractal Design Define R4
 
Solution
bgunner
Use OCCT's GPU test with error checking enabled. If it errors lower the clocks till it stops.

Power limit adjusts the point that the GPU throttles the frequency to prevent over heating. Just use this test to know if your GPU OC is stable.
Sort by date Sort by votes
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
Use OCCT's GPU test with error checking enabled. If it errors lower the clocks till it stops.

Power limit adjusts the point that the GPU throttles the frequency to prevent over heating. Just use this test to know if your GPU OC is stable.
 
Upvote 0 Downvote
Solution
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630


Thanks for the advice. My memory was to high at 1575 so it's now 1545. I saw I had no errors so I did the stupid thing and increased core clock, then my PC crashed and wouldn't even get to the BIOS. I think I'll stay away from overclocking for now.
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
It's not un heard of this happening when the GPU is OCed to much but I suggest you also stress test the other components to be sure the rest is stable. Most times if the GPU OC is unstable the driver will crash causing a black screen that recovers with an error message or you get artifacts.

test the ram first with memtest86+, put this on a disk or flash drive and run at boot for 11 passes.

then test the CPU with prime95 for 5 or more hours and next with IntelBurn Test set to maximum stress level and for 20 runs.

If these tests fail there is a part unstable or bad. Knowing all parts are stable is well worth the time it takes to stress test the system.
 
Upvote 0 Downvote
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630


When I OC the memory even by 10 or 20, I get errors but only at the start of the test. I assume it's because the GPU is getting up to speed, so can this be ignored?
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
No you can not ignore it at all If it errors the OC is unstable. Oh be sure not to use the PC while running this test as it can give a false positive. once you click start/run just leave it alone. you may have reached the max your memory can go to before it errors. errors can cause crashes and artifacts in game. I suggest you OC each clock one at a time. Example OC the memory by 50's and test each time. once you error, back it down by the 25 and retest, if it errors back it down by 10 and so on. If it errors and you back it down by 25 and it stops then bump ut by 5's till you error and then once you error drop it by 1's till you stop. this will give the max OC the memory can handle.

Next write this number down on a piece of paper and reset the memory to stock and then OC the GPU clock. you do this in the same way as the memory clock by 50's then 25's then 10's 5's and 1's. this will give you the max OC for the card. BUT watch that you are not causing the GPU to throttle ( lower the frequency on its own) or it will happen in game.

once you have both max clock enter and apply them both. now retest for stability. this has been my tried and true method for years and still works even for my card that has ECC memory on it. (ECC = Error Checking & Correcting)
 
Upvote 0 Downvote
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630


My card at stock is 1000/1250, I have got it to 1200/1250. I can boost the memory by 5-10 before I get errors, am I doing something wrong or is it a bad card?
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
But you are stable using the 13.9 I linked you to at stock settings? Try and use the older driver again to be sure it is a driver issue. Also make sure what ever program your using for OCing is allowing the Fan speed to fluctuate properly. Personally I like using MSI afterburner to OC and to set a personal fan curve to keep the card cool.

the heat thing may be a fan speed issue so check this first, next try the older driver again to be sure the heat and crashing is the driver and not a card issue. If the card has an issue it will show with both drivers.
 
Upvote 0 Downvote
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630
I installed 13.12 but still have the problem and I was using Afterburner with custom fan setting but it still crashed. Do you think the overclocking I did wrecked my card, if so how because I thought only voltage would affect the card but I can't change it, only the power.
Update: I've got the fan on 80% constantly and so far while playing sleeping dogs I've had no problems but it's still running way hotter at idle than it used to.
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
try the 13.9 driver and see if it helps as I have had issue with anything above with my setup. Ok any time you OC the chip you can hurt it. plane and simple fact of OCing. The power limit needs to be set in a way that does not allow the temps to reach above 85°c. this setting controls the throttle point of the GPU. If you hit to high of a temp it will hurt the Chip on the card.

try the 13.9 driver if this does not help at stock clocks then if you are still in warranty do an RMA and don't tell them you OCed it. IF you are out of warranty then you will need to purchase a new GPU as if it is unstable at stock clocks with different drivers the card is bad. Not all newer drivers are good. It depends on the equipment and the exact hardware.

things that hurt chips most often: Heat, voltage, to high of a frequency and liquid. these all can damage a chip. usually the high frequency is a long term issue but the other three can do damage very quickly.
 
Upvote 0 Downvote
T

TheLoneWolf989

Honorable
May 10, 2013
85
0
10,630


I will try 13.9 but I was playing Sleeping Dogs in window mode and it was fine but when I put it full-screen it crashed my PC again. It's strange, your thoughts?
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
my thoughts on that are try a different game to be sure it is not a driver/game compatibility issue. Together the two may have issues. This may take some time to be sure it is a compatibility issue. I mentioned before that the newer driver would give me issues But it was ONLY with one game, Sniper Ghost Warrior 2. The driver would cause the game to crash But only that game. So because it was the Beta driver, I sent in a trouble ticket to let them know it was the driver. I reverted back and had no issues at all. This proved it was a driver compatibility issue with my hardware and the game.

I understand it takes time to test and to figure out exactly what is happening and can understand the frustration it causes not being to able to know for sure right away. This is the nature of PC's and learning how to diagnose program and hardware compatibility issues. your well on your way in learning the steps to take to diagnose if it is hardware or software issues.
 
Upvote 0 Downvote
bgunner

bgunner

Judicious
Feb 13, 2009
5,144
14
32,515
there is a 1 year warranty from the company, not the store so yes they will take it and send you a new one or fix yours and send it back. you need to contact the manufacturer, XFX, with a copy of the purchase receipt to prove when it was purchased, and they will have you mail it to them and send you a replacement. DO NOT tell them it was OCed!! or they will void the warranty.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom