MSI MAG X670e Tomahawk Wifi - PCI Express stuck in x16.1.1

K

kwilque158702de

New member
Joined
Apr 22, 2023
Messages
5
Hello,

I'm having an issue with my motherboard, which sometimes gets stuck in PCIe 16.1.1 mode, and the graphics card remains locked at 450MHz for the GPU Clock, resulting in extremely poor performance.
3070.gif 2023-07-31_17h14_42.png 2023-07-31_17h16_43.png
This problem occurs about once every 6-8 startups, but when it starts up correctly in 16.4.0 mode, I don't encounter any issues afterward.

Initially, I suspected it might be a problem with my RTX 3080 graphics card, so I swapped it with a RTX 3070 from another computer, but the problem persisted (also stuck at 450MHz in 16.1.1).

Upon booting up the PC, I use GPU-Z to check the PCIe port's operating mode, and I see that it remains stuck in 16.1.1 mode, regardless of whether I launch a game or not.

I've already tried using DDU in safe mode to uninstall and reinstall the drivers, but it didn't make any difference. I also have a second Windows installation on my PC, and the problem occurs there as well.

In the BIOS, I attempted to change the PCIe port mode from "auto" to force it to 16.4.0, but it didn't change anything, and the issue persists.

I hope someone else has experienced similar problems so that I'm not alone, and together, we can find a solution.

thank you
 
Sort by date Sort by votes
I have a similar problem with my RTX 4080, rebooting the PC solves the problem regarding stuck GPU speed, the PCIs @ 1.1 seems a feature rather than a bug, run GPUz and then open a game in a window and the card will boost to PCIe 16x, at least that's with my system.
 
Upvote 0 Downvote
When I say that my PCIe port is locked in 16.1.1 mode and my graphics card is at 450MHz, they remain the same even when I launch a game.
Clearly, the graphics card can obviously switch to 16.1.1 in idle mode, but here it remains that way all the time, even when I use it, and I no longer have FPS in games.
 
Upvote 0 Downvote
kwilque158702de said:
When I say that my PCIe port is locked in 16.1.1 mode and my graphics card is at 450MHz, they remain the same even when I launch a game.
Clearly, the graphics card can obviously switch to 16.1.1 in idle mode, but here it remains that way all the time, even when I use it, and I no longer have FPS in games.
Click to expand...
The cause is unknown MSI has not been able to Replicate the Problem
Please see topic
 
Upvote 0 Downvote
Responding in summary that they have no solution is not acceptable to customers.
I will be more than happy to swap out my faulty motherboard with a new one so they can replicate the problem.
From what I've read, it seems to happen to a lot of people on different motherboards and chips.
I can add that my 2 Geforces 3070 /3080 are MSI brand so MSI can't accuse the GPU bios of another brand to be the issue.
I will continue to post on the other thread.
 
Upvote 0 Downvote
Hi kwilque158702de,
please check the "event viewer" of your OS (guess it is Win11?) - do you find numerous errors caused by "nvlddmkm" event code = 14?

I had this very same issue occur every time when this error showed up, tried several DDU uninstalls of the GPU drivers + Registry changes but nothing worked. In the end, best thing was to uninstall the OS and to run a clean re-install - I am currently on the BETA bios AGESA 1007a (137 for your board), so far every thing works after the re-install (7 days now with several cold boots, reboots etc.) + OC DDR5.

My suspicion: the drivers look like the main culprit, maybe not playing nice after Hardware changes w/o OS reinstall in case of AM5 + DDR5 mem OC (6000) + Win11/10 + certain Autostart configs/programs; my Nvidia MSI card / drivers worked flawlessly on the AM4 platform (MSI X570S Unify-X Max).

BTW: Since re-install, I did not enable GPU-Z to auto start (if this applies to you), just Hwinfo + Afterburner with applied OC.
 
Upvote 0 Downvote
felix.brueckmann said:
Hi kwilque158702de,
please check the "event viewer" of your OS (guess it is Win11?) - do you find numerous errors caused by "nvlddmkm" event code = 14?

I had this very same issue occur every time when this error showed up, tried several DDU uninstalls of the GPU drivers + Registry changes but nothing worked. In the end, best thing was to uninstall the OS and to run a clean re-install - I am currently on the BETA bios AGESA 1007a (137 for your board), so far every thing works after the re-install (7 days now with several cold boots, reboots etc.) + OC DDR5.

My suspicion: the drivers look like the main culprit, maybe not playing nice after Hardware changes w/o OS reinstall in case of AM5 + DDR5 mem OC (6000) + Win11/10 + certain Autostart configs/programs; my Nvidia MSI card / drivers worked flawlessly on the AM4 platform (MSI X570S Unify-X Max).

BTW: Since re-install, I did not enable GPU-Z to auto start (if this applies to you), just Hwinfo + Afterburner with applied OC.
Click to expand...

Yes, I have multiple "nvlddmkm" = 14" event code but they doesn't seem to coincide with the problem (or at least some are missing) I'll check next time it happens. I'm on windows 10.
For now I have reduced the Soc voltage to 1.25v to see if the problem returns. But you know how complicated it is to check if this solves the problem. I will think of a reinstallation if that can really be a solution, but my Windows installation is almost new and had no component changes except today to test another graphics card.
 
Upvote 0 Downvote
Yeah, with Win11, every time I got the messed up boot (i.e. driver not correctly loaded), event 14 happened with same time stamp (date, time within a minute). I was and still keep my 7590X operating at 1.21V, so - maybe some higher OC FCLK (above 2000/2033) could cause this problem regardless of SOC but...in my case, I can run Linpack 50x (10GB size), 3hrs Y-Cruncher VST and 2000% Memtest (that one is not so hard on the FCLK, granted) fully stable with 2067 or 2100.

I will post an update (here and in the other thread) once I got 10+ days w/o this issue to feel comfortable with the OS re-install suggestion; and maybe order of driver install matters as well -> my order: Chipset first, GPU second and NO install of any drivers via Windows Update = pause it or block internet connection, drivers can be downloaded on USB stick prior re-install
 
Upvote 0 Downvote
Update: New ACE BIOS (1.93 1007c) and here we go again with the "GPU driver bug" - it can be related to higher mem freq + fclk, w/ BIOS 1.84 I managed to cold boot 10x w/o issues @FCLK 2000 + 6000 - but still no guarantee it would not have occurred in the long-term; OC settings atm = 6200 MEMCLK / 2067 FCLK, stability tests passed = 2hrs VST, HNT, N64 + 50x Linpack + 1200% HCI Memtest. Currently trying to boot w/o Afterburner + RTSS (both uninstalled), in case this seems to avoid the bug, I would try out previous versions (from Guru3d: Afterburner 464 instead of 465 - that worked at least fine with my 4090 on AM4 and early AM5 BIOS versions, with no driver loading bugs).
 
Upvote 0 Downvote
hmj2k1158202e1 said:
Turn off expo profile and see if this helps.
Click to expand...
Thanks for the hint - I have been running EXPO disabled ever since switching to the AM5 platform; did a CMOS clear, not the "MSI Override" option for SOC appeared again (before that only via AMD OC available). Maybe PROC ODT is set too high on auto by the ACE = 60? Afterburner+RTSS is not the cause...bug still there after cold boots, but not like shutting down and just waiting 5min but only if the PC has been turned off for a longer time.
 
Upvote 0 Downvote
felix.brueckmann said:
Will try dialing down FCLK back to 2000
Click to expand...
*Update* seems way more consistent so far, 12hrs Mem/CPU stress test OCCT passed + no more sudden idle reboots + cold boot (after 6hrs turned off) no GPU driver loading error resulting in messed up initialization with PCIE stuck at x16.1.1 (2.5 GT/s) - will again post after a couple of days if this issue is gone with "standard" FCLK @2000
 
Last edited:
Upvote 0 Downvote
I guess so: Do you have EXPO enabled? For me, setting FCLK to 2000 (while running VSOC <1.25V) resolved the issue so far. My DDR5 timings below (Expo disabled):
1691498817968.png
 
Last edited:
Upvote 0 Downvote
felix.brueckmann said:
I guess so: Do you have EXPO enabled? For me, setting FCLK to 2000 (while running VSOC <1.25V) resolved the issue so far. My DDR5 timings below (Expo disabled): View attachment 176641
Click to expand...
Hi @felix.brueckmann ,
Did you solve the issue for this motherboard and EXPO settings? I am recommending this model to a friend, and i am worried about this.
Thanks,
Carlos
 
Upvote 0 Downvote
carlos.leon.villamayor said:
Hi @felix.brueckmann ,
Did you solve the issue for this motherboard and EXPO settings? I am recommending this model to a friend, and i am worried about this.
Thanks,
Carlos
Click to expand...
Not felix, but if I were to do my build again, I'd probably go with an ASUS board and an Intel CPU like I did last time. With an MSI board and AMD 7800X3D I've had way more issues than I've ever had going intel. Having to adjust my voltages to keep my system stable feels like something I should only have to deal with if I was overclocking...

But ya, after doing a ton of research, and tweaking a bunch of OC settings, I've been able to get my system mostly stable. It's still not perfect, and memtest86 doesn't detect any issues. Other people with the exact same hardware as me have reported no problems. So it seems like it might just be more of a quality control issue with some CPUs being better than others. But I'm not super impressed considering all the rave reviews I've seen online that convinced me to try AMD.
 
Upvote 0 Downvote
Back
Top