GPU freezing under load

[f3rmin]

Hello. I've bought a second hand 6600XT around 4 days ago and I'm having a problem.

 

I previously had a 1660 super which has, what, 30W less power draw compared to 6600xt and I never had a single problem with it.

 

Right after plugging the GPU, I started having a whole lot of problems - I did use DDU before swapping the cards - it'd freeze with minimal load, mostly when I was trying to change the refresh rate from 60 to 120. I thought weird, but maybe something was corrupted and I was eventually able to sort that out.

 

Then I tried gaming, and the same issue came back: The game would try to load and I'd get either a green screen (no text, no anything) or it'd just get stuck at the screen it was at and it'd become unresponsive - the only way was to shut it down manually.

 

Now, I've tried a lot of things: under and overvolting, trying different PCie cables, lowering the power limit, a fresh Windows install, different drivers, barebone drivers, turning off Re-Bar and Above 4G Decoding in BIOS. Temperatures are totally fine too.

 

Weirdly enough, OCCT VRAM test would freeze the PC the same way, after no more than 3 seconds of running which I'm guessing has to do something with the initial spike. It, however, doesn't freeze the PC anymore for some reason, even if games still do.

 

I doubt my PSU is the problem but maybe it is: it's able to run under maximum load though - if I get to load into games (because that sometimes happens), there's 0 further problems.

 

Event viewer also doesn't give me any errors around the time of the freeze.

 

Does anyone have any ideas? I AM fighting for a refund, but I'd like to troubleshoot more if there's anything left to troubleshoot, lol.

[Mumintroll]

Put the 1660 back in to test. Does it work?

Test the 6600XT in another pc. Maybe you can ask a friend to visit him for testing it on his pc?

[Amir F. Rad]

5 hours ago, f3rmin said:

Hello. I've bought a second hand 6600XT around 4 days ago and I'm having a problem.

 

I previously had a 1660 super which has, what, 30W less power draw compared to 6600xt and I never had a single problem with it.

 

Right after plugging the GPU, I started having a whole lot of problems - I did use DDU before swapping the cards - it'd freeze with minimal load, mostly when I was trying to change the refresh rate from 60 to 120. I thought weird, but maybe something was corrupted and I was eventually able to sort that out.

 

Then I tried gaming, and the same issue came back: The game would try to load and I'd get either a green screen (no text, no anything) or it'd just get stuck at the screen it was at and it'd become unresponsive - the only way was to shut it down manually.

 

Now, I've tried a lot of things: under and overvolting, trying different PCie cables, lowering the power limit, a fresh Windows install, different drivers, barebone drivers, turning off Re-Bar and Above 4G Decoding in BIOS. Temperatures are totally fine too.

 

Weirdly enough, OCCT VRAM test would freeze the PC the same way, after no more than 3 seconds of running which I'm guessing has to do something with the initial spike. It, however, doesn't freeze the PC anymore for some reason, even if games still do.

 

I doubt my PSU is the problem but maybe it is: it's able to run under maximum load though - if I get to load into games (because that sometimes happens), there's 0 further problems.

 

Event viewer also doesn't give me any errors around the time of the freeze.

 

Does anyone have any ideas? I AM fighting for a refund, but I'd like to troubleshoot more if there's anything left to troubleshoot, lol.

there are two sides of your problem, posible hardware problem or software problem, what you describing can happen in multiple scenarios, i am gonna try to explain one by one, 

1. modified bios with too tight timing or wrong bios for the memory type, solution is first change the bios to original, 

2. dying core on the gpu or broken trace under gpu core which when the gpu heats up, the broken solder joint expands and cuts the continouty of the circuit,

3. broken memory rail mosfet which crumble under load, if you can try a furmark test, try an older version like 1.33 which puts less stress on the memory, if it still crashes you can lower the chance of the mosfet,

for now you can do these; first change the bios to original, if you never touched this gpu bios do it anyway, because maybe previos owner might have done it,
second, try furmark lowest resolution and turn all the fans to max, i want you to put load on the core while keeping it cooler, and see if the time it takes to crash compared to higher resolution incrases or decreases, also please check the gpu specially tha back of it for any broken trace or scratch on the pcb or a broken part from the pcb, i lost count of how many times this was the main problem of a broken gpu

there are some other posible diagnosis which i am not qualified to explain or even talk about, so i wont waste your time by explaining what i dont understand in the first place

[f3rmin]

22 hours ago, f3rmin said:

Hello. I've bought a second hand 6600XT around 4 days ago and I'm having a problem.

 

I previously had a 1660 super which has, what, 30W less power draw compared to 6600xt and I never had a single problem with it.

 

Right after plugging the GPU, I started having a whole lot of problems - I did use DDU before swapping the cards - it'd freeze with minimal load, mostly when I was trying to change the refresh rate from 60 to 120. I thought weird, but maybe something was corrupted and I was eventually able to sort that out.

 

Then I tried gaming, and the same issue came back: The game would try to load and I'd get either a green screen (no text, no anything) or it'd just get stuck at the screen it was at and it'd become unresponsive - the only way was to shut it down manually.

 

Now, I've tried a lot of things: under and overvolting, trying different PCie cables, lowering the power limit, a fresh Windows install, different drivers, barebone drivers, turning off Re-Bar and Above 4G Decoding in BIOS. Temperatures are totally fine too.

 

Weirdly enough, OCCT VRAM test would freeze the PC the same way, after no more than 3 seconds of running which I'm guessing has to do something with the initial spike. It, however, doesn't freeze the PC anymore for some reason, even if games still do.

 

I doubt my PSU is the problem but maybe it is: it's able to run under maximum load though - if I get to load into games (because that sometimes happens), there's 0 further problems.

 

Event viewer also doesn't give me any errors around the time of the freeze.

 

Does anyone have any ideas? I AM fighting for a refund, but I'd like to troubleshoot more if there's anything left to troubleshoot, lol.

I found out that installing any kind of drivers is what causes the problem. I did a fresh install and I'm currently on the ones that Windows installs for you automatically and it's working just fine which confuses me even more.

 

Found one Reddit thread of a guy describing a very similar issue to mine, also ASRock 6600 but without the XT. Weirdly enough the guy who sold me this card didn't have this problem but obviously how would I know he's not lying.

[f3rmin]

20 hours ago, Mumintroll said:

Put the 1660 back in to test. Does it work?

Test the 6600XT in another pc. Maybe you can ask a friend to visit him for testing it on his pc?

Unfortunately sold the 1660 super and most of my friends are either on laptops or PCs with lower specs. 

 

I even tried asking the tech repair guys in my town if they could test it for me but none of them have good enough PSUs lmao

[f3rmin]

17 hours ago, Amir F. Rad said:

there are two sides of your problem, posible hardware problem or software problem, what you describing can happen in multiple scenarios, i am gonna try to explain one by one, 

1. modified bios with too tight timing or wrong bios for the memory type, solution is first change the bios to original, 

2. dying core on the gpu or broken trace under gpu core which when the gpu heats up, the broken solder joint expands and cuts the continouty of the circuit,

3. broken memory rail mosfet which crumble under load, if you can try a furmark test, try an older version like 1.33 which puts less stress on the memory, if it still crashes you can lower the chance of the mosfet,

for now you can do these; first change the bios to original, if you never touched this gpu bios do it anyway, because maybe previos owner might have done it,
second, try furmark lowest resolution and turn all the fans to max, i want you to put load on the core while keeping it cooler, and see if the time it takes to crash compared to higher resolution incrases or decreases, also please check the gpu specially tha back of it for any broken trace or scratch on the pcb or a broken part from the pcb, i lost count of how many times this was the main problem of a broken gpu

there are some other posible diagnosis which i am not qualified to explain or even talk about, so i wont waste your time by explaining what i dont understand in the first place

My first thought also was modified BIOS, but what's in GPU-z and techpowerup seems to match just fine.

 

I would also love to either open it up or play around with the BIOS but I'm currently pushing for a refund and the GPU's seal is still there + I'm a little worried to touch the BIOS because if there's a chance I get my money back, I don't want to risk borking it just yet lmao. PCB also seems totally fine on the back