[SOLVED] 3080 FTW3 Crashing (new build)

[Brobeans_]

I just finished my new build 2 days ago and am finally attempting to benchmark it, however I am running into some serious issues.

Mobo: ASUS Z490-A
GPU: EVGA 3080 FTW3
CPU: i7 10700k
RAM: 2x Corsair Vengeance 3200
AIO: H100i Corsair
PSU: Corsair RM850

Here are the ONLY things I have done to my PC:
Installed :
— windows 10 from usb (media creation tool)
— & updated drivers for GPU (install GeForce Experience)
— iCue
— ASUS Armoury Crate
— Logitech G-Hub
— Techpowerup
— HW Monitor
— Steam (Halo, 3D Mark, Satisfactory, Dota)
— Battle.net (Warzone)
— Discord

Enabled from BIOS in EZ Mode
— XMP for RAM
— Auto AI OC for CPU

I was very excited to finally test out my PC, however when launching Halo for the first time I was met with ~120 seconds of runtime before a crash to desktop. I thought that was weird and attempted to start it again, resulting in a quicker crash (main menu). I thought that was weird and restarted my PC and went into 3DMark to run a GPU bench and it lasted about 20 seconds into TimeSpy before error and crash to desktop. I did this about 8-10 more times, resulting in 3 BDOS total and lots of personal frustration.

After, I went back into BIOS and disabled XMP and turned my CPU OC off; at the same time doing a clean reinstall of NVIDIA drivers. I attempt to run another 3DMark bench and get to the second scene
/test before crashing to desktop again. I don’t understand what’s going on and am obviously very frustrated considering how hard it is to get 3000 series cards and am questioning what I’m doing incorrectly. I have a spare build from 5 years ago with a 1080 so I have another system to test on, but would really like to get this figured out.

One thing I could mention is that my GPU is 3x8 pin and my PSU only came with 2 pcie cables (daisy chains on each), so one cable is daisy chained to the third 8 pin. I really don’t know why this would cause any issues as it’s still providing (I would think) enough power.

I’m not the most proficient in troubleshooting software but definitely not a complete novice so I’m looking for any and all assistance possible. Thanks!

 

[Brobeans_]

Update:

I used the DDU tool to refresh my NVIDIA drivers. My next step is to run the PC with integrated graphics only, however I’m not sure if that directly implicates the GPU or not.

 

[sizzling]

You are not using a pcie riser cable?
You are not using any power extension cables?

As a test what happens if you undervolt the gpu by 100 or 200 MHz.

I will probably get moaned at again for saying this but I’ve seen several posts over the months with people having issues with RM PSU’s instead of RMx.

 

[Brobeans_]

You are not using a pcie riser cable?
You are not using any power extension cables?

As a test what happens if you undervolt the gpu by 100 or 200 MHz.

I will probably get moaned at again for saying this but I’ve seen several posts over the months with people having issues with RM PSU’s instead of RMx.

I am not using a GPU riser. I am using antec sleeves/extensions for the mobo, CPU, and GPU power cables. How would I be able to tell if this was causing an issue? I have looked and have been so far unsuccessful finding a dedicated third pcie cable from Corsair (out of stock).

Would I be able to undervolt the GPU in MSI afterburner? Would you mind giving me a brief walkthrough on that?

I guess the PSU could be a problem but it’s brand new and I am unsure how to single it out.

Thanks for the help, I will be able to test the undervolt and integrated graphics tonight.

 

[sizzling]

Can you just use the cables that came with the psu with no extensions. You don’t need tidy cable management to test this possibility. I have read that extensions often add sufficient additional resistance to cause problems with these very power hungry gpu’s. High quality replacement cables or the original cables should be the only ones used.

I used the wrong term, sorry. I meant a 100-200MHz underclock. I would not try undervolting while you have stability issues. There are lots of guides on how to do both of these things with MSIAfterburner, far better than what I can type here.

 

[Brobeans_]

Can you just use the cables that came with the psu with no extensions. You don’t need tidy cable management to test this possibility. I have read that extensions often add sufficient additional resistance to cause problems with these very power hungry gpu’s. High quality replacement cables or the original cables should be the only ones used.

I used the wrong term, sorry. I meant a 100-200MHz underclock. I would not try undervolting while you have stability issues. There are lots of guides on how to do both of these things with MSIAfterburner, far better than what I can type here.

Okay yes that actually was something I hadn’t strongly considered. I will remove the sleeves tonight before attempting to underclock. I will do some research prior. Thanks and I will report back with results!

 

[Phaaze88]

Would I be able to undervolt the GPU in MSI afterburner? Would you mind giving me a brief walkthrough on that?

Below is a pretty quick and easy way to do it, but you need to remove the sleeve extensions from the equation first as sizzling said.


Run EVGA Precision X in the background with the monitoring window enabled. Play a game for a few minutes.
Close the game completely and take a look at the highest core clock and Vcore recorded - memorize or write 'em down.
Click on Precision's Curve Editor. Take the max Vcore and subtract 0.05v from it. Find the new value - or the one closest to it - on the voltage curve and click on it.
Using the up key, raise the frequency to the max clock you saw, plus 10mhz. Lock it in with the L key and click apply. You can create and save profiles on Precision, which is recommended for quick access.
The only downside(maybe?) to this is that you have to run Precision whenever you want to play. Do not opt to apply the new settings at Windows startup.

If it crashes, instead of 0.05v, do 0.04/0.03/0.02/0.01v and test again.

 

[Brobeans_]

Can you just use the cables that came with the psu with no extensions. You don’t need tidy cable management to test this possibility. I have read that extensions often add sufficient additional resistance to cause problems with these very power hungry gpu’s. High quality replacement cables or the original cables should be the only ones used.

I used the wrong term, sorry. I meant a 100-200MHz underclock. I would not try undervolting while you have stability issues. There are lots of guides on how to do both of these things with MSIAfterburner, far better than what I can type here.

I removed the cable extenders and ran a Heaven benchmark, which actually ran 2 full cycles with no problems (hovering just under 2ghz clock steadily), then when I closed heaven and my clock speeds dropped to idle my discord and steam both stopped working and received errors such as “need to reinstall” (could be completely unrelated).

I restarted my PC and attempted to open 3DMark through steam and before even opening I BSOD with “kernel security check failure”.

I will make a few more attempts to 3DMark multiple cycles as this as at least farther than I’ve gotten with the extenders. If I keep crashing I will move onto underclock and then try integrated graphics to check that. Not sure if what I wrote above helps in any way but I wanted to document it.
Thanks

 

[Brobeans_]

Can you just use the cables that came with the psu with no extensions. You don’t need tidy cable management to test this possibility. I have read that extensions often add sufficient additional resistance to cause problems with these very power hungry gpu’s. High quality replacement cables or the original cables should be the only ones used.

I used the wrong term, sorry. I meant a 100-200MHz underclock. I would not try undervolting while you have stability issues. There are lots of guides on how to do both of these things with MSIAfterburner, far better than what I can type here.

Christmas Eve update:

I underclocked the card, first by 100 and then by 200. Both resulted in failures to finish the first stage of 3DMark - no BSOD but crash to desktop. Frustratingly, I messed again with the XMP and AI OC in the BIOS with no difference in results. I don’t understand how Heaven finished 2 full cycles but 3DMark could not finish even one.

Next, I pulled out my 3080 and put in my 1080 and was able to successfully run 3 full cycles of 3DMark. I figured this was an even better way to test than using the integrated graphics.

Is this a clear cut RMA of the 3080? Or am I still missing some potential causes such as not enough power or RAM issues? I’d be glad to share some logs if needed, although I would need some basic direction on that. I would also love to test the 3080 on my old build, but it’s only a EVGA 750 B2 PSU without 3x8 pins.

The last thing I’d want is EVGA to receive the 3080 and say “we don’t see anything wrong with it on our bench”

Looking forward to your response! Thanks again

 

[Brobeans_]

I really can’t be sure. If me I would probably go for a fresh start, so a complete reinstall of Windows and drivers. Also ensuring the BIOS is up to date.

Do you have another psu you can try or can you try the gpu in another system, maybe a friends?

I guess I would be confused as to how a fresh reinstall on a 3 day old computer would change anything.

I can try to flash my BIOS but definitely a little trepidatious on that front. The mobo was released in early 2020 I believe.

I do have a friend not far away that I could test but it might have to wait till after the holidays. Additionally I live next to a micro center so I am wondering if they have a bench or diagnostic center to run tests as well, unless that’s widely regarded as not the best idea?

 

[Kona45primo]

I'd see if your friend can run the 3080 assuming he's got a psu beefy enough to handle it.

It could be that psu just as easily as it could be the card. If your friend can run it, it's psu, if not gpu... Might also be able to have a decent PC repair shop check the card on their system.

 

[Brobeans_]

I'd see if your friend can run the 3080 assuming he's got a psu beefy enough to handle it.

It could be that psu just as easily as it could be the card. If your friend can run it, it's psu, if not gpu... Might also be able to have a decent PC repair shop check the card on their system.

Thanks for the response! I agree that I’d be hesitant to jump to conclusions on what is the sole culprit.

Not 100% sure if his PSU can handle it but I have a Microcenter less than a mile away, would you recommend using their diagnostic check?

 

[Kona45primo]

If it's not terribly expensive then microcenter is worth a shot. I'd talk to a tech there and see if they are capable of testing that card.



Once that system is working it's going to be awesome!

 

[Brobeans_]

If it's not terribly expensive then microcenter is worth a shot. I'd talk to a tech there and see if they are capable of testing that card.



Once that system is working it's going to be awesome!

Thanks I appreciate it! I called today on the off chance none of my friends had a compatible PSU and they said it was about $40 to do a full tower diagnostic. I’m a little skeptical of pc repair in general but after the holidays I will give it a shot as a last resort. I’ll be sure to follow up with everyone here on the results.

On a somewhat related note, I filed a tech support ticket with EVGA as a precaution to get a step ahead of the curve on an RMA. We’ll see what they say.

Thank you everyone for your input, I’ll post an update after the holidays. Enjoy time with your families!

 

[Brobeans_]

Update:

I reached out to EVGA to start a ticket and they suggested using a third dedicated pcie cable. I finally obtained one and tried it this morning with no better performance.

I’m snowed in today but I will take the tower to micro center tomorrow in order to run all the components against a test bench.

If anyone else has any other ideas I am all ears!

 

[rickypicky5]

One thing I could mention is that my GPU is 3x8 pin and my PSU only came with 2 pcie cables (daisy chains on each), so one cable is daisy chained to the third 8 pin. I really don’t know why this would cause any issues as it’s still providing (I would think) enough power.

I'm not sure this is a good idea. To be clear, are you saying that you are using only two connections on you PSU to power all three of the connections on your GPU? I may be wrong, but I don't think that's going to work very well.

 

[Brobeans_]

I'm not sure this is a good idea. To be clear, are you saying that you are using only two connections on you PSU to power all three of the connections on your GPU? I may be wrong, but I don't think that's going to work very well.

I was doing that, yes - it is all a Corsair RM850 comes with. I looked online and nobody else was having the same issue.

Regardless, as per my update this morning, it did not fix the issue.

 

[helper800]

I was doing that, yes - it is all a Corsair RM850 comes with. I looked online and nobody else was having the same issue.

Regardless, as per my update this morning, it did not fix the issue.

I have had no issues with the microcenter in California, quite the opposite actually. If the standard of care and professionalism are similar at the store near you I see no issue letting them give it a try.

 

[Brobeans_]

I just want to wrap this up - microcenter benched and the RAM was stated to be bad. Ran another memtest and my 2nd stick produced ~10k errors lol