Replaced GPU, had Random Shut Down

[VesperMoirai]

Just replaced my EVGA 2080 XC with a EVGA 3080 TI FTW3 Ultra Gaming

 

Was running fine all of yesterday and then all of a sudden the PC shut off on me. GPU was hot but was low 80s. Apparently this GPU does that. So wasn't throttling. The PC Would not turn back on until I turned the PSU off and back on. Was playing Final Fantasy XIV (Main Monitor - Capped at 165FPS) had 2 Chrome Tabs open (YouTube and Facebook I think) and Discord on the 2nd monitor. That was it.

 

My PSU is older about 5 or 6 years old. However it is a 1000W Platinum rated PSU. (Rosewill Quark 1000W Platinum)  I ran a few benchmarks to get CPU/GPU wattage up but I was only seeing the PSU have a load of about 500-600w and no spikes. Is my PSU getting too old and frail for this new power sucker of a GPU?

 

 

Ryzen 7 5800x

Celsius S36 360mm AIO

32 GB 3200mhz CL 16 G.Skill Royals 2x16

Main monitor 1440p 165hz / Offhand Monitor 1080P UW

1000w Rosewill Quark PSU

ASUS STRIX-F X570 Mobo

500GB NVME SSD & 2TB HDD

Windows 11

 

[VesperMoirai]

If this needs moved to troubleshooting go ahead.

[RMTM]

There is a big chance that the available PCIe power from your PSU is not enough to power that beast of a 3080. Try plugging 3 pcs. PCIE power cables directly from PSU to your 3080 gpu. I suggest not to use daisy-chain with your PCIE power cables. Use one direct PCIE power cable from PSU to your gpu. So you should use a total of 3 PCIE power cables.

[VesperMoirai]

With it being 6 years old is it a better option to replace it and make it non daisy chain aswell?

8 hours ago, RMTM said:

There is a big chance that the available PCIe power from your PSU is not enough to power that beast of a 3080. Try plugging 3 pcs. PCIE power cables directly from PSU to your 3080 gpu. I suggest not to use daisy-chain with your PCIE power cables. Use one direct PCIE power cable from PSU to your gpu. So you should use a total of 3 PCIE power cables.

 

[TexasBulldog74]

I have the same GPU and I had to completely reinstall windows and start fresh.  No matter how many times i ran DDU, deleted manually, cleaned the registry I couldn't get away from random resets and I went down the rabbit hole of installing a new 1000w PSU, tried different ram as well as a different MOBO, Different modular cables with and without extensions.. the ONLY thing that stopped it from randomly restarting and the total lockup where i had to flip the switch on the PSU was a fresh OS install starting over.  

 

It happened at idle, in game, during benchmarks, 30* temps... I couldn't find a common situation to repeat it.  It drove me absolutely insane for over a week and a bit of money I didn't end up needing to spend.

[VesperMoirai]

I will most likely re-install windows too. But, still. Should I still get a new PSU for how old this is? I feel like 6 years is a bit long for a PSU. If not I'll just reinstall windows and hope it don't happen again.

 

Backing up FFXIV options and UI is a drag though...

[RMTM]

8 hours ago, VesperMoirai said:

With it being 6 years old is it a better option to replace it and make it non daisy chain aswell?

 

Yes. And "Clean Install" the latest NVidia driver.

[VesperMoirai]

8 hours ago, RMTM said:

Yes. And "Clean Install" the latest NVidia driver.

Seeing as though EVGA PSUs really don't have daisy tails on them it appears as though EVGA HIGHLY recommends that config. Just put a friend's 3080TI (same card) and new PSU in (EVGA brand G6 or somthing like that) and how that PSU is set is each rail is a 6+2 for PCIE. So guess it's time to buy a new PSU and do some backups. Thanks guys. 

[peacefulpolarbear]

Is there anything in event log? I was getting black screens after upgrading GPU, and it turns out there was a bug check code generated in event viewer.

 

I never figured it out, but reinstalling Windows fixed my issue

[VesperMoirai]

On 5/21/2022 at 9:10 PM, peacefulpolarbear said:

Is there anything in event log? I was getting black screens after upgrading GPU, and it turns out there was a bug check code generated in event viewer.

 

I never figured it out, but reinstalling Windows fixed my issue

Nope nothing the last error I had was 2 weeks ago from Armory Crate RGB crash.

[VesperMoirai]

Ok so. It might of actually been my RAM. I went to go turn my PC on today and I got a fast post then it went off and did a boot loop. Took sticks out did the every channel thing. My RGB Royals won't work in Slot 4 now even with XMP off. So guess it's new RAM instead of PSU. Did a PSU test and it's fine.

 

Guess I'll just have to look at Mobo Spec sheet and see what is certified working and go from there.

 

I'll keep you posted might order a kit tonight so we'll see when it gets here.

[VesperMoirai]

Also, I did use my brother's old RAM sticks and in slot 4 it booted.

[Aragorn-]

The fact you said it wouldn't power on at all untill you cycled the PSU at the mains suggests to me the PSU tripped one of it's protections. Certain supplies dislike the noise from these newer high power GPUs and can trip even though they have a much higher rating than should be required.

 

If it was bad ram, it might crash or reboot, but it wouldn't lose all power entirely. It might boot loop or sit there hard locked with all the fans spinning etc...

 

A simple test for potential ram issues is simply switch off xmp and run them at 2400mhz

[RMTM]

In your very first message, your computer shut down when you replaced your GPU. If you did not do anything to any other parts aside from your gpu. It can only be 3 things. Either a bad gpu, BIOS PCIE settings or cpu power cabling. And it unlikely for a newer 3080 to fail right away. Your motherboard should tell you anyway. Though I am not sure of how you connected your PCIE power cables to your gpu. As what I have suggested to you before not to daisy chain your power cables from PSU to your GPU. Can you show me a photo of how your power cables are connected to your gpu? In my 3080 I used one PCIE cable per gpu power socket.

 

 

[VesperMoirai]

I will have to double check but I do believe it is PSU related. Since it shuts down gard enough for windows to even know it crashed. I will try switching out my cables to the left overs I have from My wife's which is the exact same unit and see if it's stable after a few days.

[VesperMoirai]

I don't think I'm using any pigtails but I'll look when I get out of work.

[VesperMoirai]

OK, I mistook one of the pigtails for a standalone wire. All 3 PCIE-Power cables were plugged into the PSU but they were the 8x2 + 8 plugged in now 8,8,8 pin plugged in on standalone wires. We'll see if it does it again for a few days. It fell back into the PSU pit lol.

[VesperMoirai]

Well it just did it again.

[VesperMoirai]

[VesperMoirai]

[RMTM]

That's unfortunate. Before ruling-out the PSU, what is your motherboard (ASUS STRIX-F X570) BIOS version?

[VesperMoirai]

V.4204 I have Armory Crate keep that up to date and don't allow experimental.

[VesperMoirai]

Re-flashed the BIOS. Have ran Port Royal Extreme 3D Mark multiple times, Cernoblyte Ultra 1440p Benchmarks, Cyberpunk2077 Ultra Ray Traced Benchmark 1440p. No crashes yet. I may still get a new PSU cuz of how old this one is but it seems to be doing fine enough to not worry about it exploding. Sunday or Monday I'm just going to Reformat and see if it works better. It was about 75-100 average depending on game.

[RMTM]

2 hours ago, VesperMoirai said:

Re-flashed the BIOS. Have ran Port Royal Extreme 3D Mark multiple times, Cernoblyte Ultra 1440p Benchmarks, Cyberpunk2077 Ultra Ray Traced Benchmark 1440p. No crashes yet. I may still get a new PSU cuz of how old this one is but it seems to be doing fine enough to not worry about it exploding. Sunday or Monday I'm just going to Reformat and see if it works better. It was about 75-100 average depending on game.

Yes. I was about to ask you to flash your BIOS, after my leading question. That is a good choice. A new PSU for better power delivery is always better. 

2 hours ago, VesperMoirai said:

It was about 75-100 average depending on game.

If it is the temperature that you're talking about here. Below is my RTX 3080 fan curve. I do not reach 65degC playing CP2077 (Ultra Ray Tracing @1440p res). I understand that yours is the "Ti" version. But should it be really that far of a gap in terms of operating temps? How much FPS do you get when you play CP2077(with 1440p Ultra Ray Tracing settings)? See screenshot below as well.

 

 

[VesperMoirai]

I think I fixed it. No crashes since I did the BIOS flash. I usually get 80ish (Me and my Wife have about 1500w of heat getting thrown into our office so that don't Both PCs are pretty performant.) max on temps and about 88FPS in CP2077 at 1440p Ultra RT. Haven't adjusted fan curve yet just default Aggressive Fan curve. DLSS set to Auto. Also the firmware update with X1 seemed to help Temps as well. Usually about 415-430w Power draw if it's working hard.