[SOLVED] Tdr failure with 3080ti

[Murillen]

Hello, this is my first time posting and I rarely post anywhere in general so sorry if I make a mistake.

I'm having a problem with my computer build which I've recently completed. Here is the spec sheet first:

Mobo: ASUS ROG STRIX X570-I
CPU: Ryzen 9 5950x
CPU Cooler: nzxt kraken z63
GPU: EVGA GeForce RTX 3080 Ti XC3 ULTRA 12G
RAM: Corsair Dominator 3600Mhz 2x16
PSU: silverstone sx700 pt sfx (700w)
Boot Drive: Intel 660p 1tb
Other Drives:
samsung 87p Evo 4tb sata ssd,
Toshiba ocz rd400 1 tb nvme

To preface I used my computer as is but with a 1070 while waiting for the 3080ti to come in for about a week with no issues. When I installed the 3080ti it would give me the vga debug light which was constantly on, so I made sure to re-seat and check connections, after which it would boot but bsod erratically either after bios splash, at desktop opening google or everytime when trying to install drivers. The vga led also occasionally comes on requiring me to keep attempting to restart the pc until it gets past the vga post.

At first I thought it was either a faulty GPU or PSU but when I put the gpu in other systems it worked perfectly fine and didn't crash and posted every time. For both systems which they had i5-3570k with 8g ddr3 cx650m psu and ryzen 5 3600 with 16gb ddr4 and same psu.

Putting the 1070 back into the system and it runs fine, and thinking that it may be that the power supply needs more wattage I put the cpu into eco mode (draws 65w) and still the same results of constant crashing and vga light on.

I swapped a cx 650m power supply into the system and same result, so I'm thinking it might be the motherboard or riser cable.

I wanted to see if anyone has any insight on this before I try RMAing anything as I'm at a loss on what to do

Thank you very much and if need any more info please ask

Edit: forgot to write the the bsod error is: VIDEO_TDR_FAILURE (nvlddmkm.sys) on windows 10 pro

 

[Phaaze88]

Go ahead and rule out the riser - plug the gpu directly into the motherboard.

 

[Murillen]

Go ahead and rule out the riser - plug the gpu directly into the motherboard.

When I did that the vga light was still on but now 2 of the 3 gpu fans spin at max speed while one doesn't spin at all. I did make it so that only the ram, gpu psu, cpu and cpu Cooler were only connected like a out of box test run since plugging the gpu into motherboard directly required me to take the motherboard out of the case

 

[Phaaze88]

When I did that the vga light was still on but now 2 of the 3 gpu fans spin at max speed while one doesn't spin at all.

You were using Precision X to control fans? I remember reading that EVGA's cards behave funky with the likes of Afterburner.


If you have to take the gpu out of the chassis, then do so - seat the motherboard on its own box. Get the cost free things out of the way first.
Neither the 5950X nor the X570 chipset have built in graphics, so you need to plug the monitor to the gpu or you'll get a VGA error.

The only other way to rule out the motherboard is with another motherboard.

 

[Murillen]

You were using Precision X to control fans? I remember reading that EVGA's cards behave funky with the likes of Afterburner.


If you have to take the gpu out of the chassis, then do so - seat the motherboard on its own box. Get the cost free things out of the way first.
Neither the 5950X nor the X570 chipset have built in graphics, so you need to plug the monitor to the gpu or you'll get a VGA error.

The only other way to rule out the motherboard is with another motherboard.

It's a fresh install of windows since I was wondering If that was the problem, so I'm not to sure about precision x as this behavior is the first for me since i got it, but doing as you said leads to the aforementioned problems; vga light on 2 fans spinning like crazy.

In terms of motherboard, I'm leaning towards that as it works fine with two other boards, but the cpus were different

I'm just in awe that a 1070 will work fine in this system perfectly but the 3080ti won't work at all

Though earlier I did boot it in safe mode with network connection and it worked fine for the duration of it (30 mins) but after turning it off the started the errors again.

Would you recommend I rma the motherboard? Unless there's possibly something else

 

[Phaaze88]

In terms of motherboard, I'm leaning towards that as it works fine with two other boards, but the cpus were different

The cpu shouldn't cause any trouble unless it's broken, or needs to be reseated - maybe there's some dirt in the socket, or you've over-tightened the cpu cooler.

I'm just in awe that a 1070 will work fine in this system perfectly but the 3080ti won't work at all

GTX 1070 is a PCIe Gen 3 device, 3080Ti comes with Gen 4 support, but riser devices have been very hit and miss with Gen 4 gpus - even Gen 4 risers!
That's why the first thing I suggested was plugging the card in the mobo directly.

Would you recommend I rma the motherboard? Unless there's possibly something else

If nothing changes with the riser removed, I would try taking it to a shop and see if they can get it to show display with their own X570 or B550 mobo.
If that's no good, RMA the mobo... but if they don't find anything wrong, you'll get the same board back.

 

[Phaaze88]

Continuing the above, RMA'ing either the motherboard or gpu will take time, with the wait time on the gpu being even worse - for obvious reasons - if you don't do that Advanced RMA(?) or whatever EVGA calls it, but that's not cheap either.

 

[helper800]

@Murillen I would get DDU on the computer with the 1070 and then uninstall the current graphics drivers. Then install the standalone Nvidia drivers for the 3080 ti and check that. I would also try resetting CMOS / flashing to the newest version of BIOS before RMAing anything. Also make sure that when you connect the PSU to the GPU to use a separate cable per 6 or 8 pin power connector on the GPU.

 

[Murillen]

Continuing the above, RMA'ing either the motherboard or gpu will take time, with the wait time on the gpu being even worse - for obvious reasons - if you don't do that Advanced RMA(?) or whatever EVGA calls it, but that's not cheap either.

It seemed like the the riser was the problem! I re seated everything while testing on the gpu box and its working perfectly now, I'm guessing the first time around I didn't sit everything right when testing without the riser. The Gen 4 support with the riser makes a lot of sense and points out a lot of the problems I was experiencing. Thank you I really appreciate it!

Now time to get a new case or riser 🥲

 

[Murillen]

@Murillen I would get DDU on the computer with the 1070 and then uninstall the current graphics drivers. Then install the standalone Nvidia drivers for the 3080 ti and check that. I would also try resetting CMOS / flashing to the newest version of BIOS before RMAing anything. Also make sure that when you connect the PSU to the GPU to use a separate cable per 6 or 8 pin power connector on the GPU.

I did try separate power cords for each 8 pin and updating bios but it turned out to be the darn riser, thank you though!

 

[boju]

If you need riser cable try force pcie to gen 3, that'll most likely work.