• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

4090 Mystery Issue

J

jphil

n00b
Joined
Sep 30, 2018
Messages
33
Recently I got a used MSI 4090 Gaming Trio for my Threadripper 1950x system with a 1050W PSU, and I just went through the craziest 24 hours where the system had a complete meltdown and then recovered.

At first I wasn't able to boot into Windows. I couldn't even get into safe mode or run any recovery options, it would just hang and reboot. I did a Memtest overnight with no errors, and finally got it working after swapping PSUs.

I'm still not sure what caused the issue exactly. I ended up swapping the 1050W back and it's still working, so I just want to doublecheck that I have the right power cable setup for the GPU.

The PSU is a Cougar GX1050:
https://www.newegg.com/cougar-gx1050-1050-w-80-plus-gold-certified/p/N82E16817553008
https://cougargaming.com/products/discontinue_psus/gx/

There's one 12V line hardwired to the PSU that I'm using 2 connectors (yellow cables in the photo), and then an additional 12V cable coming from one of the 12V2 outputs (red port on the PSU in the photo). Does it matter 12V1 vs 12V2 on the PSU? I'm using the 3 8-pin to 12-pin adapter that came with the GPU (branded NVIDIA). I know there were changes to the power connector on the 4090, so I included a photo, don't know if it's relevant at all.
20241008_143114-1.jpg

20241008_133444-1.jpg

20241008_133350-1.jpg


I had previously been using the PSU with a 2080ti and 3080ti with the same cables no problems. As far as I understand from the specs, the PSU is ATX12V v2.31 compliant.

What started me down the dark path in the first place was, I was getting really bad benchmarks on Superposition with the 4090, around 13,000 compared to 30-40k online.

I thought I would try optimizing my CPU with Ryzen Master to see if that helped, but when I tried to install a newer version, that's when my system hung and went kaput.

I haven't even gotten to updating the drivers, so I don't know if that should help performance, but if there's any suggestions I'd appreciate it. In the end I don't know if it was a power issue, some random glitch, or if I got scammed, but I'm trying to give the seller the benefit of the doubt because they seemed legit and I don't know if I have any recourse at this point anyways.
 
Last edited:
Why does it appear that you're only using a single PCIe cable from the PSU to power a 4090? And that PSU isn't fantastic to begin with (C-tier). I'd use a full three cables, you have the ports to spare.

1728418374288.png
 
What you want to make sure to do here is have all 8 pin pci-e cables plugged into the red ports. Ideally 3 separate 8pin pci-e cables that go into the adapter.
 
sounds like you bumped a cable somewhere...
criccio said:
Why does it appear that you're only using a single PCIe cable from the PSU to power a 4090? And that PSU isn't fantastic to begin with (C-tier). I'd use a full three cables, you have the ports to spare.

View attachment 684369
Click to expand...
hes not. thats one of the old, and one of the new. i would perfer three separate cables, but that combo should be fine.
1728418979846.png
 
criccio said:
Why does it appear that you're only using a single PCIe cable from the PSU to power a 4090? And that PSU isn't fantastic to begin with (C-tier). I'd use a full three cables, you have the ports to spare.

View attachment 684369
Click to expand...

It's semi-modular, so there's one 12V 8pin hardwired out of the PSU. That's the yellow one that's daisychained to 2 connectors. I guess people are saying I should run a 3rd separate cable.
 
jphil said:
It's semi-modular, so there's one 12V 8pin hardwired out of the PSU. That's the yellow one that's daisychained to 2 connectors. I guess people are saying I should run a 3rd separate cable.
Click to expand...
yeah, if you can do one cable per plug that is the best way to spread the load, but it should be fine how you have it.
 
  • Like
Reactions: jphil
like this
So I added another cable to spread the load. I tried running a benchmark to see if it had any effect on performance but didn't see any difference. I appreciate the suggestions though, it gives me peace of mind I should be safe with the power draw.

I updated to the latest Nvidia driver, but when I run Superposition I'm still getting this really low score. Lower than the lowest score on the leaderboard which is a Xeon at 2.2GHz :dead:

Although I'm using it for 3D & AI rendering, in which case it rendered a sequence in 10min that previously took 24min with a 3080ti (if my records are correct), I'm not sure what to make of the benchmark. Is it only related to gaming, or am I missing a lot of juice?

Superposition_Benchmark_v1.0_12652_1728424158.png
241008-175841-chrome-1.jpg


I also tested with UserBenchmark and scored in the 7th percentile. Did I get a bum card?
241008-183032-chrome-1.jpg
 
Last edited:
jphil said:
So I added another cable to spread the load. I tried running a benchmark to see if it had any effect on performance but didn't see any difference. I appreciate the suggestions though, it gives me peace of mind I should be safe with the power draw.

I updated to the latest Nvidia driver, but when I run Superposition I'm still getting this really low score. Lower than the lowest score on the leaderboard which is a Xeon at 2.2GHz :dead:

Although I'm using it for 3D & AI rendering, in which case it rendered a sequence in 10min that previously took 24min with a 3080ti, I'm not sure what to make of the benchmark. Is it only related to gaming, or am I missing a lot of juice?

View attachment 684403View attachment 684409
Click to expand...
Bad news for you if you thought threadripper was for gaming… because it’s not at all. It’s always been lower performance for gaming.
 
jphil said:
So I added another cable to spread the load. I tried running a benchmark to see if it had any effect on performance but didn't see any difference. I appreciate the suggestions though, it gives me peace of mind I should be safe with the power draw.

I updated to the latest Nvidia driver, but when I run Superposition I'm still getting this really low score. Lower than the lowest score on the leaderboard which is a Xeon at 2.2GHz :dead:

Although I'm using it for 3D & AI rendering, in which case it rendered a sequence in 10min that previously took 24min with a 3080ti, I'm not sure what to make of the benchmark. Is it only related to gaming, or am I missing a lot of juice?

View attachment 684403View attachment 684409
Click to expand...
It's your CPU. It's a Zen 1 Threadripper so it's even slower than a regular one. The rendering doesn't depend on the cpu, so it's able to make use of the 4090.
 
Dreamerbydesign said:
Bad news for you if you thought threadripper was for gaming… because it’s not at all. It’s always been lower performance for gaming.
Click to expand...

I originally got it when CPU rendering was the only option in 2017, so the cores were necessary. Of course now it's all on the GPU. I don't do any gaming.
 
GoldenTiger said:
It's your CPU. It's a Zen 1 Threadripper so it's even slower than a regular one. The rendering doesn't depend on the cpu, so it's able to make use of the 4090.
Click to expand...
Ok, that's what I'm hoping. I just want to make sure it's not the card since I bought it used.
 
jphil said:
Ok, that's what I'm hoping. I just want to make sure it's not the card since I bought it used.
Click to expand...
Should be fine, but if you want peace of mind try a 3dmark test too and compare the ballpark graphics score (not the combined one). It should be relatively close as that graphics score isn't terribly cpu dependent :).
 
GoldenTiger said:
Should be fine, but if you want peace of mind try a 3dmark test too and compare the ballpark graphics score (not the combined one). It should be relatively close as that graphics score isn't terribly cpu dependent :).
Click to expand...

Thanks, I'll give it a shot
 
yeah based on the utilization, id say cpu too. could try bumping the res and retesting with more load on the gpu, but i think its fine.
GT's suggestion is good too.
 
  • Like
Reactions: jphil
like this
jphil said:
I originally got it when CPU rendering was the only option in 2017, so the cores were necessary. Of course now it's all on the GPU. I don't do any gaming.
Click to expand...
You did run the supposition benchmark tbf, and were comparing performance differences.
 
You must log in or register to reply here.
Back
Top