6900 xt won’t boot on PCIE1

Jeff G

Limp Gawd
Joined
Nov 2, 2017
Messages
234
Just put together a new build with a MSI z690 Tomahawk DDR4, 12700k, and 6900 xt. GPU won’t show anything when plugged into PCIE1, but works fine on PCIE3 (Gen 5 x16 vs Gen 3 x4). Swapped an old 980ti into PCIE1 and it works fine.
Any idea why the 6900 xt won’t work on PCIE1? Stress tested on PCIE3 and runs great other than being at x4.
Fresh Win10, tried various ram sticks, forced PCIE1 to Gen 4 and Gen 3 (neither worked), all drivers up to date, MB bios updated. I can’t figure this one out.
 
Slot/case alignment issue maybe. Remove the board from the case and bench test it in slot1 with the 6900xt to see if its shorting the slot to the case or misalignment issue.
 
Slot/case alignment issue maybe. Remove the board from the case and bench test it in slot1 with the 6900xt to see if its shorting the slot to the case or misalignment issue.
I’ll try this tomorrow evening, but with a dual custom loop I was really hoping to avoid this lol. Oddly, the old 980ti slots right in and boots up perfect in the PCIE1 slot so I wouldn’t think it would be one of those issues. Can’t hurt to try at this point though
 
May be related to this case of PCIE4 GPUs in Z690s....

https://community.intel.com/t5/Proc...sands-of-WHEA-17-errors-a-minute/td-p/1339941

Read a few sporadic cases in reddit forums as well. I don't have a conclusive answer on what it is though. I am not thoroughly up to date on Intel 12G CPUs other that they are otherwise great.
I can’t get to windows, I don’t even get a display to get to bios. I’ve pulled all ram and ssd drives to see if I could even get video to pop up telling me I don’t have ram installed, nothing at all. If I swap it over to PCIE3 (Gen 3 x4 slot) it runs good besides being x4. I tried an old Gen 3 980ti in PCIE1 slot and it worked fine too. So the GPU seems to work, and the PCIE1 lanes seems to work, but they don’t work together.
 
Hmm that is very interesting….is it a brand new 6900xt? Right now with the info you’ve given me it’s hard to pin point the problem. The only thing I can think of is thaw fact that pcie1 is a full 16x and pcie2 is 4x, perhaps there is a fault on the other 12 pcie lanes on the graphics card that prevents it from outputting an image on the full 16x slot? Again this is just from working with what you gave me, any other info would help make a better diagnosis
 
did you not test everything before building the loop? is the bios up to date?
Bios is up to date. The system booted with the 6900xt in the PCIE3 slot, I made sure motherboard bios was up to date, drivers were up to date, and all windows stuff was up to date.
I did not test before building the loop as I didn’t have an air cooler or a way to build the water loop outside of the case.
 
Hmm that is very interesting….is it a brand new 6900xt? Right now with the info you’ve given me it’s hard to pin point the problem. The only thing I can think of is thaw fact that pcie1 is a full 16x and pcie2 is 4x, perhaps there is a fault on the other 12 pcie lanes on the graphics card that prevents it from outputting an image on the full 16x slot? Again this is just from working with what you gave me, any other info would help make a better diagnosis
The 6900xt is second hand, everything else is new. What would cause PCI lanes on a gpu to fault? Is it possible to have a fault on x16 but not x4?
Any info you need to help diagnose, just let me know.
 
check that the pcie fingers are clean and theres no damage.
Update on everything tonight:
Pulled the whole thing apart and built a super basic loop with some old soft tubing on a cardboard box. No change. Dropped the GPU into an old z270/6700k build using the same ssd, same ram, same PSU as the z690. Posts and runs perfect at Gen 3 x16 speeds (which is what that board supports). Talked to PowerColor and they think it may be a conflict with the Gen 5 lanes. Oddly, I set bios to PCI_E1 @ Gen 3 and Gen 4 and couldn’t get it to post on either setting. They’re working with AMD to see if there’s an update to fix it. He said I’m the third person to have had this problem recently
 
Update on everything tonight:
Pulled the whole thing apart and built a super basic loop with some old soft tubing on a cardboard box. No change. Dropped the GPU into an old z270/6700k build using the same ssd, same ram, same PSU as the z690. Posts and runs perfect at Gen 3 x16 speeds (which is what that board supports). Talked to PowerColor and they think it may be a conflict with the Gen 5 lanes. Oddly, I set bios to PCI_E1 @ Gen 3 and Gen 4 and couldn’t get it to post on either setting. They’re working with AMD to see if there’s an update to fix it. He said I’m the third person to have had this problem recently
good. at least they are on it and youre not alone.
 
Update on everything tonight:
Pulled the whole thing apart and built a super basic loop with some old soft tubing on a cardboard box. No change. Dropped the GPU into an old z270/6700k build using the same ssd, same ram, same PSU as the z690. Posts and runs perfect at Gen 3 x16 speeds (which is what that board supports). Talked to PowerColor and they think it may be a conflict with the Gen 5 lanes. Oddly, I set bios to PCI_E1 @ Gen 3 and Gen 4 and couldn’t get it to post on either setting. They’re working with AMD to see if there’s an update to fix it. He said I’m the third person to have had this problem recently
Interesting, it could be the new pcie5 signal amplifiers that need to be used along the traces to the pcie slots, as the pcie5 protocol’s drawback to the higher throughput is increased signal degradation over travel distance, maybe they are causing the compatibility issues with pcie gen 4 speeds? Please keep us up to date on what powercolor says.

And just for FYI purposes, any physical damage to the signal traces on the gpu pcb could break signal continuity on one or multiple pcie lanes, or like pendragon1 said, the copper fingers on the gpu’s pcie connector are damaged or have insulting material buildup, etc. but since the 6900xt works at x16 throughput on another system you shouldn’t worry about these.
 
Interesting, it could be the new pcie5 signal amplifiers that need to be used along the traces to the pcie slots, as the pcie5 protocol’s drawback to the higher throughput is increased signal degradation over travel distance, maybe they are causing the compatibility issues with pcie gen 4 speeds? Please keep us up to date on what powercolor says.

And just for FYI purposes, any physical damage to the signal traces on the gpu pcb could break signal continuity on one or multiple pcie lanes, or like pendragon1 said, the copper fingers on the gpu’s pcie connector are damaged or have insulting material buildup, etc. but since the 6900xt works at x16 throughput on another system you shouldn’t worry about these.
I’ve found a few other people with the same issue, and they’ve all been with MSI z690 boards. I’ve reached out to the others and the only response I’ve gotten back is “I switched to Asus and the problem disappeared”. So right now, I’m leaning towards an issue with the MSI bios that affects this particular card. I’ve also reached out to MSI, and I’ll continue working with them and PowerColor, but if I can’t figure this out by the weekend then I’m returning the board before my return window is past.
 
[Problem Solved!]

I’ve been back and forth with PowerColor and AMD directly testing pretty much everything I can on my end, the consensus is it’s an issue on the MSI board. I’ve also been in touch with MSI, but haven’t got much past the “did you try unplugging it and plugging it back in?” type responses. My return window was coming to an end on the board, so I returned it and swapped over to Asus. Problem fixed, everything works as it should. Ran dozens of benchmarks to make sure and the card performs as it should. Hopefully this will help if anyone else with this issues stumbles in here.

*Side note: PowerColor tech support is hands down the best service I’ve received from any PC components company. They promptly answered emails and actually went through everything with me to try and get this running. I’ve dealt with pretty much all the major brands at some point over the last decade and they take the win for best Customer Service award.
 
Last edited:
Issues like this are the reason I avoid jumping into brand new platform equipment and tend to wait for a year or so to give vendors time to iron out the kinks. MSI will eventually debug the bios on that board but that doesn't resolve a non functional system overnight and leaves the end user with a bad taste towards the vendor.
 
Bios is up to date. The system booted with the 6900xt in the PCIE3 slot, I made sure motherboard bios was up to date, drivers were up to date, and all windows stuff was up to date.
I did not test before building the loop as I didn’t have an air cooler or a way to build the water loop outside of the case.

Unless you're using rigid tubing everywhere, the next time you redo your loop get two sets of quick disconnect connectors and place them on either side of your CPU so you can disconnect the GPU from the loop and reconnect the tubing for a CPU only loop.
 
[Problem Solved!]

I’ve been back and forth with PowerColor and AMD directly testing pretty much everything I can on my end, the consensus is it’s an issue on the MSI board. I’ve also been in touch with MSI, but haven’t got much past the “did you try unplugging it and plugging it back in?” type responses. My return window was coming to an end on the board, so I returned it and swapped over to Asus. Problem fixed, everything works as it should. Ran dozens of benchmarks to make sure and the card performs as it should. Hopefully this will help if anyone else with this issues stumbles in here.

*Side note: PowerColor tech support is hands down the best service I’ve received from any PC components company. They promptly answered emails and actually went through everything with me to try and get this running. I’ve dealt with pretty much all the major brands at some point over the last decade and they take the win for best Customer Service award.
I don't blame you since you were near the end of the return period. However, MSI support can be decent for BIOS issues. I've gotten beta BIOS's not released to public via email before from them after reporting issues.
 
I actually have the same issue with my MSI 6800xt Gaming X Trio and three different Intel boards.

MSI B660M Mortar - no output
Gigabyte B660M DS3H - no output
Asus Z690 TUF D4 - no output
Gigabyte B450 Aorus M - works fine

I just RMA'd the card hoping it's a vbios issue or something that can be worked out on their end.
 
I had the exact same setup as the OP. One 6900XT worked fine, another didn't. It was a PowerColor Red Devil. Explained behavior, got immediate RMA. RMA card worked fine. It was one of the first produced and it's VBios was not happy with PCIe 5.
 
I had the exact same setup as the OP. One 6900XT worked fine, another didn't. It was a PowerColor Red Devil. Explained behavior, got immediate RMA. RMA card worked fine. It was one of the first produced and it's VBios was not happy with PCIe 5.

Good to hear. I'm hoping it's something they can do a quick turnaround on for me. They didn't even ask any questions, they just approved the RMA, so it sounds like they are familiar with it.
 
  • Like
Reactions: Wat
like this
Good to hear. I'm hoping it's something they can do a quick turnaround on for me. They didn't even ask any questions, they just approved the RMA, so it sounds like they are familiar with it.
All those early 6000 cards were out before Intel PCIe 5 hit the market, so there was no qualification done on any wide scale, so while a PITA, not totally unexpected either.
 
The classic fix - moved from MSI to ASUS. :ROFLMAO:

I sadly have done the same, recently. Used to be a hardcore MSI guy (back when the MSI Lightning cards were out and their awesome X99 XPOWER boards)...but have slowly moved back to ASUS even though they pissed me off years ago. I'm sure I'll migrate back someday.
 
I just double-checked. The purchase date was 3/2021 so it was a pretty early card.
 
Back
Top