Question GPU (RTX 3090) is suddenly underperforming in main PCI slot after cleaning and reinstallation

[Broken 3090 Guy]

(Posted in motherboard forum because this appears to most likely be a motherboard issue.)

Hello there. My computer was due for a cleaning and I happened to remove my graphics card (an RTX 3090), dust it off with canned air, and reinstall it into the main PCI slot of my motherboard (MSI MPG X570 Gaming Plus). Suddenly I noticed a 15-30 fps drop across all games. Running the Blender 3D benchmark, my 3090 is benchmarking worse than a 3050 now.

Through tons of troubleshooting, the most notable things I found is the card basically behaves as expected when moved to the other PCI slot on my motherboard, but underperforms in the main slot. There appears to be low power consumption and low clock speed when it's in PCI_E1 instead of PCIE_E3. Whether one is causing the other, or something else is causing both is unknown. But we can reasonably determine that 1) the card itself is not faulty, and 2) the PSU and cables are not faulty.

I have never overclocked anything in my PC, nor messed with any power saving settings within windows or my BIOS.

I'm at wits end with this, any ideas what's causing the underperformance in PCI_E1?

See below for tech specs and troubleshooting.

Specs:​

• CPU: AMD Ryzen 9 3950X 3.5 GHz 16-Core Processor
• RAM: Corsair Vengeance RGB Pro 64 GB (2 x 32 GB) DDR4-3200 CL16 Memory
• Motherboard: MSI MPG X570 GAMING PLUS ATX AM4
• BIOS Version: E73C37AMS.AO0 (Build date: 07/15/2024)
• GPU: NVIDIA GeForce RTX 3090 by ASUS TUF Gaming
• GPU Driver: GeForce Game Ready 560.81
• PSU: RMx Series RM750x — 750 Watt 80 PLUS® Gold Certified Fully Modular PSU (Bought new from Corsair in Aug 2020)
• Operating System: Windows 11
• Boot Drive: Western Digital Blue SN550 NVMe M.2 2280 1TB PCI-Express 3.0 x4 3D (SSD)
• CPU Cooler: Noctua NH-D15

Troubleshooting:​

• Original graphics card driver (Studio 560.81) upgraded to Game Ready 560.81 via clean install. No change.
• Graphics driver uninstalled with Display Driver Uninstaller (DDU version 18.0.7.9), then Game Ready driver 560.81 reinstalled. No change.
• BIOS updated from E7C37AMS.A70 (Build Date 01/09/2020) to E73C37AMS.AO0 (Build date: 07/15/2024). No Change.
• Physical inspection of GPU: no damage observed. Dusted with canned air.
• Physical inspection of PCI slots: no damage observed. Dusted with canned air: nominal amount of dust dislodged.
• Moved GPU from PCI_E1 to PCI_E3: Card functions as expected in slot 3. See screenshots for data comparisons.
• Moved GPU back to PCI_E1 from PCI_E3: Card returned to same level of underperformance. See screenshots for data comparisons.

Blender 3D Benchmark​

I am primarily using Blender's benchmark to test the GPU, since that is my main use case for this card, other than gaming. I also get expected frame rates at 4k resolutions in games when the card is in slot 3 as opposed to slot 1.

 

[Lutfij]

Welcome to the forums, newcomer!

You forgot to mention the make and model of your PSU and it's age. As for your BIOS update, did you clear the CMOS after verifying your BIOS was flashed to the latest version?

 

[Broken 3090 Guy]

I have the RMx Series™ RM750x — 750 Watt 80 PLUS® Gold Certified Fully Modular PSU. I bought it new from Corsair in August 2020, so I'm not sure if that'd make it a 2019 or 2020 model. I'll update the original post with that info.

As for clearing the CMOS after verifying the BIOS.... I have no idea what that means, so chances are I probably didn't do it!

When I updated the BIOS, I restarted, went back into the BIOS, made sure it reflected the correct version, and then booted back to windows. I haven't changed or done anything in the BIOS since the update.

 

[Broken 3090 Guy]

Welcome to the forums, newcomer!

You forgot to mention the make and model of your PSU and it's age. As for your BIOS update, did you clear the CMOS after verifying your BIOS was flashed to the latest version?

Resetting the CMOS fixed it! (*Only for a time... see following message #5)

For anybody stumbling upon this thread with a similar issue: On this motherboard, I just had to short two pins for a few seconds and did so using a screwdriver. Rebooting took a minute or so longer, but now the GPU is getting 100% power in PCI_E1 and benchmarked where it should with Blender. My games are back up to their expected fps in high/ultra settings.

 

[Broken 3090 Guy]

Well crap. It worked for a night of gaming but now we're back to the same exact issue. What's worse: resetting the CMOS this time didn't fix it. It's still only getting 45% power.
Ugh...

 

[Broken 3090 Guy]

Fixed it, lol...

Somehow MSI Afterburner got set to tell the GPU to have the minimum power limit. Resetting this fixed it. (Imagine that, the limit-power slider limiting the power of the GPU!)

...In my defense, I never mess around with MSI Afterburner, so how this got changed I have no idea.