[SOLVED] Is it my GPU or PSU ?

[Zamatron]

Hello,
I am having a problem with my system which is running on an RTX 3090 where the screen goes black and the monitor loses signal or the computer just freezes followed by the black screen and no signal, all this However only happens when the gpu is in full use like during gaming or benchmarks, the gpu temperatures never goes above 80°C core temp and 100° maximum memory temp at stock and they lower down when i undervolt a bit. On undervolting the gpu does stay on for a little bit longer than stock but fails eventually.

Now coming to why I think this might be a psu issue is that the gpu does work for the most part when not in load or during peak load before crashing, no fps drops or glitching. And the psu has already once melted one of the cables that goes from the psu to the gpu on the psu end, i have however changed the cables and used a different port and that worked normally for a week or so, but now i can't seem to run any games on it or work my renders.

Note that the psu i am using is an Antec HCG 1000 extreme thats an 1000Watts 80+ gold rated psu that shouldn't have a problem running a 3090 and 5900x system should it?
Should i change my psu or rma my gpu?

Full spec:
MSI X570 UNIFY
TRIDENT Z 3600MHZ DDR4 32GB (2X16) CL16
INNO3D ICHILLX4 RTX 3090
R9 5900X
PSU ANTEC HCG 1000 EXTREME 1000W
2 NVME M.2 SSDS
1 HDD 7200RPM
O11 DYNAMIC CABINET

 

[RememberThe5th]

There are two things which are happneing, GPU overheating and that previous problem with PSU melting cables.
Also there is tier list which suggests that the pre 2018 versions are better than yours 2018 version.
From my opinion 150$ PSU for 1000W gold is a bit sketchy for me.

Also it might not be best option, but on hwinfo you have rail voltages, you can see if 12V,5V etc. drops/goes above its rated value of 5%
I would suggest you putting your GPU in another system with good enough GPU to run the game (benchmark/test) to see will it fail agian.
It would be good if you have another system, friends or work pc to test that theory before buying PSU and your GPU needs RMA.

 

[--SID--]

ANTEC HCG 1000 EXTREME

The symptoms don't surprise me with a Seasonic OEM PSU. It's a Focus Plus Gold under the hood, a platform know with ripple issues and shut downs with high-end GPUs.

 

[Zamatron]

The symptoms don't surprise me with a Seasonic OEM PSU.

Have you any experience with it before? It'd really help me with some insight to get to the bottom of the issue I'm facing.

 

[Zamatron]

There are two things which are happneing, GPU overheating and that previous problem with PSU melting cables.
Also there is tier list which suggests that the pre 2018 versions are better than yours 2018 version.
From my opinion 150$ PSU for 1000W gold is a bit sketchy for me.

Also it might not be best option, but on hwinfo you have rail voltages, you can see if 12V,5V etc. drops/goes above its rated value of 5%
I would suggest you putting your GPU in another system with good enough GPU to run the game (benchmark/test) to see will it fail agian.
It would be good if you have another system, friends or work pc to test that theory before buying PSU and your GPU needs RMA.

I've read somewhere that inefficient power delivery can over heat gpu? I don't know well though. However after cleaning the gpu heatsinks a bit the gpu works at around 77° max at high load 100% USAGE... which is about as high as it gets in the aftermarket card I'm using . I have tried powerlimiting the gpu to only use 75% power and not letting it go above 300w and that did solve the issue for quite a while but eventually it started failing as well. I still think the psu is degrading by the minute but it also might be the gpu. I'm so confused.

 

[--SID--]

Have you any experience with it before?

With the original, yes. Asus Strix GTX 970 black screened with Focus Plus Gold because of high ripple and the PSU shut down with a lot of high-end GPUs. 750w/850w shut down with 1080 Ti/2080 Ti and AMD Vega. No experience with RTX 3000 but the powerspikes are much worse than 1080 Ti/2080 Ti. You're issues lookes like a ripple issue though.

With how many seperate cables is de PSU connected to the GPU?

 

[Zamatron]

With the original, yes. Asus Strix GTX 970 black screened with Focus Plus Gold because of high ripple and the PSU shut down with a lot of high-end GPUs. 750w/850w shut down with 1080 Ti/2080 Ti and AMD Vega. No experience with RTX 3000 but the powerspikes are much worse than 1080 Ti/2080 Ti. You're issues lookes like a ripple issue though.

With how many seperate cables is de PSU connected to the GPU?

its a 2×8 pin gpu and i am using 2 separate cables to power it up. However the system was custom pre-built and the people who built it used the single daisy chained cable and I was dumb enough to trust them as i was running those cables for the first 3 months until the cable melted and then I realized what was up, i changed the cables into two separate ones and in different psu ports and it had been working like a charm until yesterday when i started facing this issue. Sorry if i sound dumb, I'm learning as i go and taking all this as a way to educate myself and any help/suggestions or insights help q lot, thank for that.

Do this has a high probability to be a psu issue right? Maybe it was a factory defective unit? And it started dying the day that cable burned possibly due to overheating?

 

[RememberThe5th]

The weird thing is that the PSU didnt react with OCP (over current protection) if it started to melt.

 

[Zamatron]

Here's the update.

So a few days back nvidia released a driver update and my issue started just 2 days after, i did however clean install gpu drivers and it didn't fix my problem, today while i was still troubleshooting the system, i ran heavens benchmark a few times until it crashes and i was monitoring it through hwinfo and i just realized that my fans were gpu fans were running at a mere 1600rpm and it hit me maybe the gpu was faulty and the fans aren't working properly, but i still put a manual fan curve on afterburner and ran the benchmark and to my surprise the pc almost crashed but quickly recovered and the gpu fans ramped up to around 3200 rpm at 90% then i put afterburner fan control to auto and it was still running at 3000+ rpms now and haven't crashed since.

My theory is that the nvidia drivers messed with my gpu bios default fan curves and hogged it to 1600rpm, lets see if the issue persists... I'm however still changing my psu because i don't wanna risk it damaging the rest of the system.