Progressively worse scores in 3Dmark

[KpGaz]

System Specs:

 

• CPU: AMD Ryzen 7 9800X3D
• GPU: MSI RTX 5090 Suprim  SOC
• Motherboard: ASUS ROG Strix B650E-F
• RAM: 32GB DDR5 (EXPO enabled)
• Storage: 2TB NVMe Gen4 SSD
• PSU: 1200W Corsair 
• Cooling: High-end AIO-cooler 
• OS: Windows 11 , fully updated

Problem Description:

The issue began after I applied a light PBO undervolt on my CPU (all-core -10, +150 MHz boost), along with a UV 2800MHz @0.91mV, VRAM overclock of +2000 MHz via MSI Afterburner.

 

After a crash in RDR2 (black screen flickering, no WHEA errors), I started noticing progressively worse performance in 3DMark benchmarks, especially in Time Spy and Port Royal. Steel Nomad and CPU-focused tests still give acceptable results, but has gotten progressively worse. I also noticed very low GPU utilization in some DX12 titles (e.g. Hogwarts Legacy), despite normal FPS in other games (RDR2, KCD2, Cyberpunk 2077) albeit slightly less after reverting back all UV:s.

 

Troubleshooting Steps Taken:

• Cleared CMOS multiple times
• Reset BIOS to Optimized Defaults (EXPO on/off, PBO off)
• Updated BIOS to the latest version (irreversible)
• Ran DDU in Safe Mode before clean installing GPU drivers
• Reverted to previous NVIDIA driver before issues started
• Fully uninstalled MSI Afterburner and all profiles before DDU
• Disabled G-Sync
• Disabled all background apps including RivaTuner (before uninstalling) 
• Checked NVIDIA control panel settings (Prefer max performance, etc.)
• Ensured that power plan is set to performance in Win 11
• Tested multiple games and APIs (DX12/Vulkan)
• Monitored GPU/CPU utilization and thermals using HWInfo64 (thermals good but CPU effective clock suboptimal)
• Ran system file checks and DISM scans – repaired files and restart
• Tried both Silent and Performance BIOS on GPU
• Monitored VRAM stability (no OC now, everything stock)
• Verified proper fan curves and thermal contact
• Disabled iGPU in BIOS

 

Current Symptoms:

• Up to 2000 points lower than average in 3DMark Port Royal, Steel Nomad and Time Spy
• Low GPU & CPU utilization in select games (notably DX12-based ones like Hogwarts Legacy)
• No thermal throttling (GPU ~57-61°C, CPU ~55-57°C under load)
• System passes all stress tests (OCCT, cinabench R23 3DMark Stability, etc.)
• FPS drops and stutters in specific games (HL, Hogwarts) that didn’t exist before
• Occasionally yellow Q-LED on boot after certain stress tests. 

 

At this point, I’m unsure if the issue is due to:

 

• A corrupted BIOS state I can’t undo (latest is irreversible)
• Damaged VRAM or instability from that +2000 MHz OC (currently back to stock)
• A hidden CPU scheduling or driver conflict
• Some odd conflict in DX12 or Windows after that initial crash

 

Would really appreciate advice or fresh perspective – I’m at the end of the road here.

 

Thanks in advance!

[strange13930]

2 hours ago, KpGaz said:

System Specs:

 

• CPU: AMD Ryzen 7 9800X3D
• GPU: MSI RTX 5090 Suprim  SOC
• Motherboard: ASUS ROG Strix B650E-F
• RAM: 32GB DDR5 (EXPO enabled)
• Storage: 2TB NVMe Gen4 SSD
• PSU: 1200W Corsair 
• Cooling: High-end AIO-cooler 
• OS: Windows 11 , fully updated

Problem Description:

The issue began after I applied a light PBO undervolt on my CPU (all-core -10, +150 MHz boost), along with a UV 2800MHz @0.91mV, VRAM overclock of +2000 MHz via MSI Afterburner.

 

After a crash in RDR2 (black screen flickering, no WHEA errors), I started noticing progressively worse performance in 3DMark benchmarks, especially in Time Spy and Port Royal. Steel Nomad and CPU-focused tests still give acceptable results, but has gotten progressively worse. I also noticed very low GPU utilization in some DX12 titles (e.g. Hogwarts Legacy), despite normal FPS in other games (RDR2, KCD2, Cyberpunk 2077) albeit slightly less after reverting back all UV:s.

 

Troubleshooting Steps Taken:

• Cleared CMOS multiple times
• Reset BIOS to Optimized Defaults (EXPO on/off, PBO off)
• Updated BIOS to the latest version (irreversible)
• Ran DDU in Safe Mode before clean installing GPU drivers
• Reverted to previous NVIDIA driver before issues started
• Fully uninstalled MSI Afterburner and all profiles before DDU
• Disabled G-Sync
• Disabled all background apps including RivaTuner (before uninstalling) 
• Checked NVIDIA control panel settings (Prefer max performance, etc.)
• Ensured that power plan is set to performance in Win 11
• Tested multiple games and APIs (DX12/Vulkan)
• Monitored GPU/CPU utilization and thermals using HWInfo64 (thermals good but CPU effective clock suboptimal)
• Ran system file checks and DISM scans – repaired files and restart
• Tried both Silent and Performance BIOS on GPU
• Monitored VRAM stability (no OC now, everything stock)
• Verified proper fan curves and thermal contact
• Disabled iGPU in BIOS

 

Current Symptoms:

• Up to 2000 points lower than average in 3DMark Port Royal, Steel Nomad and Time Spy
• Low GPU & CPU utilization in select games (notably DX12-based ones like Hogwarts Legacy)
• No thermal throttling (GPU ~57-61°C, CPU ~55-57°C under load)
• System passes all stress tests (OCCT, cinabench R23 3DMark Stability, etc.)
• FPS drops and stutters in specific games (HL, Hogwarts) that didn’t exist before
• Occasionally yellow Q-LED on boot after certain stress tests. 

 

At this point, I’m unsure if the issue is due to:

 

• A corrupted BIOS state I can’t undo (latest is irreversible)
• Damaged VRAM or instability from that +2000 MHz OC (currently back to stock)
• A hidden CPU scheduling or driver conflict
• Some odd conflict in DX12 or Windows after that initial crash

 

Would really appreciate advice or fresh perspective – I’m at the end of the road here.

 

Thanks in advance!

you can try updating your chipset drivers from your motherboards website.

[OddOod]

Welcome to the forums!
 

2 hours ago, KpGaz said:

progressively worse performance

Can you quantify this? Like, it's under performing and *every* time you run it it gets worse? 

Honestly, at this point, I'd reset to 0 and start over. 
Reinstall windows, install a known good GPU driver, rollback the BIOS (there are instructions online, but it seems convoluted), reset CMOS. 
Once you're 100% clean, then you can start adding OCs and UVs etc. and see where you're stable again. 

[KpGaz]

7 minutes ago, strange13930 said:

you can try updating your chipset drivers from your motherboards website.

I’ve updated chipset drivers via AMDs own auto detect software for win 11. But I’ll check if the motherboard provides anything additional once I get home. 

[KpGaz]

11 minutes ago, OddOod said:

Welcome to the forums!
 

Can you quantify this? Like, it's under performing and *every* time you run it it gets worse? 

Honestly, at this point, I'd reset to 0 and start over. 
Reinstall windows, install a known good GPU driver, rollback the BIOS (there are instructions online, but it seems convoluted), reset CMOS. 
Once you're 100% clean, then you can start adding OCs and UVs etc. and see where you're stable again. 

When I bought and assembled the PC, I got roughly 14800-14900 in steel nomad , and around 36500-37200 in time spy. After applying the UV (both GPU and CPU) got around 15100-15200 in steel nomad. Then came the issue as described above.

 

Now I receive 14200 in SN, and if I keep rerunning the benchmark, I get worse scores every time (got a 12800 once). 
 

I have considered performing a clean re-install of windows 11, but wanted to check if there is anything else that I can do. If going down that route, should I perform a DDU and CMOS again before the clean uninstall? 
 

I cannot roll back the latest bios, as it was irreversible. So I might have screwed that up, don’t know. 

[CasualExtremist]

So is this performance decay only when  back-to-back? To me my first instinct would be to check everything about the thermals. Maybe bad batch of paste, or poorly applied paste or anything really could be allowing heat to build up which would fit the throttling you experience as you continue use. Maybe even a fan got unplugged or loos connections?

[KpGaz]

Just now, CasualExtremist said:

So is this performance decay only when  back-to-back? To me my first instinct would be to check everything about the thermals. Maybe bad batch of paste, or poorly applied paste or anything really could be allowing heat to build up which would fit the throttling you experience as you continue use. Maybe even a fan got unplugged or loos connections?

One thing I am considering doing is to go back in bios and ensure that fans are set to max and in PWM, reboot, and redo all my configs in FanControl. I read somewhere that cmos resets the fan controls in bios which might interfere with FanControl (software).

 

I will double check the plugs for the fan, however, HWinfo does not say that the CPU or GPU is thermal throttling. 

[KpGaz]

Also, wondering if it is worth reapplying all my UV profiles to see if that helps?
 

Sounds a bit stupid since stock does not seem to perform well. My line of thought is that something might still be active “silently” in both bios and in the GPU even after CMOS and DDU clean install. Redoing the setting could perhaps reset those hidden/silent/unknown settings? 

[OddOod]

1 hour ago, KpGaz said:

I cannot roll back the latest bios, as it was irreversible. So I might have screwed that up, don’t know. 

My apologies, apparently it's an exploit patch so they aren't letting people roll back which is annoying. 
 

1 hour ago, KpGaz said:

I get worse scores every time

That sounds kinda like it's getting thermally saturated. You're sure *all* the temps are good?
 

1 hour ago, KpGaz said:

I have considered performing a clean re-install of windows 11, but wanted to check if there is anything else that I can do

Given your specs, I assume you have enough cash on hand to buy a spare SSD, you could always pull your current and install clean on a new one and see if that does fix your problems. 
 

1 hour ago, KpGaz said:

If going down that route, should I perform a DDU and CMOS again before the clean uninstall? 

CMOS yes, DDU no. CMOS is gonna reset everything, DDU is software and will not survive a new install

[CasualExtremist]

57 minutes ago, KpGaz said:

One thing I am considering doing is to go back in bios and ensure that fans are set to max and in PWM, reboot, and redo all my configs in FanControl. I read somewhere that cmos resets the fan controls in bios which might interfere with FanControl (software).

 

I will double check the plugs for the fan, however, HWinfo does not say that the CPU or GPU is thermal throttling. 

So if for some reason you are not seeing all the thermals like OddOod suggested; I would also try to use HWinfo or whatever monitoring software you have to track if anything is noticeably performance throttling and by how much. I am by no means qualified to tell you where to look, but maybe even something like VRM on the motherboard, or the SSD you use for boot, or some other wild card that is getting hotter than it should might chain react to slow down the system as a whole. Being able to see what throttles first could point you towards what cooling system to inspect. All just food for thought. I could just be another keyboard wielding idiot.

 

Also just thought about the rare possibility that there is hot air from something like cpu or gpu that is blowing onto another thermally sensitive component. So perhaps all the heat is dissipating just fine, but it happens to be dumping out onto something that can't handle it. Again, just food for thought.

[KpGaz]

Thank you for sharing your experience and thoughts! I’ll try the fancontrol route first, and then perform a couple of additional benchmarks and look for other thermals that might be throttling. 
 

17 minutes ago, OddOod said:

Given your specs, I assume you have enough cash on hand to buy a spare SSD, you could always pull your current and install clean on a new one and see if that does fix your problems. 

Hehe, spent quite a bit and don’t really want to spend more . However, did not know I could just pull the drive with windows installed, swap it , and install a new partition onto the new one. I do not need to do anything else? I thought I would perform a clean reinstall through windows own application. 

[OddOod]

15 minutes ago, KpGaz said:

However, did not know I could just pull the drive with windows installed, swap it , and install a new partition onto the new one. I do not need to do anything else? I thought I would perform a clean reinstall through windows own application

I'd recommend making the install USB with the Media Creation Tool before pulling the drive, but yeah, otherwise you  should be set. Always recommended to pull ALL spare drives from a rig before installing Windows (or probably any OS) because if you don't sometimes the installer does weird things like putting the boot sector on a different drive from the OS. 

[KpGaz]

So quite interesting results.

 

Before leaving for work this morning, I completely powered down the system, unplugged the power cable, and drained any residual power. I left it like that until I got home. The only change I made in the BIOS was setting all fans to PWM mode and switching them to Turbo. After just one run of Steel Nomad, I got a score of 14 517 which is slightly above the average. I monitored everything in HWInfo during the run and didn't notice any thermal spikes or unusual sensor behavior. Nothing in the reds. One thing I did notice, though—when booting into Windows, the GPU was sitting at just 23°C

 

It got me thinking. My main M.2 SSD (with the Windows install) is installed in the slot closest to the CPU, per the motherboard's recommendation, but that slot is now partially buried under my massive 5090. Could this be a factor? I'm wondering if, during heavier GPU loads (like when running multiple benchmarks), the heat from the GPU is affecting the temperature of the M.2 drive and possibly causing some performance hiccups?

[Mark Kaine]

6 hours ago, KpGaz said:

I’ve updated chipset drivers via AMDs own auto detect software for win 11.

This is something you should thoroughly avoid (automatic updates/and or detection tools)and could easily cause such issues.

 

3 hours ago, KpGaz said:

the heat from the GPU is affecting the temperature of the M.2 drive and possibly causing some performance hiccups?

Unlikely but possible.

 

Download HWINFO64 and check max temps after the issues have occurred and post the numbers here.

 

 

[KpGaz]

18 minutes ago, Mark Kaine said:

This is something you should thoroughly avoid (automatic updates/and or detection tools)and could easily cause such issues.

How do I know which one to download and install? I get that I should probably download the latest one (release date), but the top one does not say chipset driver. When going to Device Manager --> System devices and AMD SMBus, I can only see that I have a chipset driver that is not visable on the motherboards web page (the second picture).

[Mark Kaine]

2 hours ago, KpGaz said:

How do I know which one to download and install? I get that I should probably download the latest one (release date), but the top one does not say chipset driver. When going to Device Manager --> System devices and AMD SMBus, I can only see that I have a chipset driver that is not visable on the motherboards web page (the second picture).

I would try the one from the mobo manufacturer website.

 

(chipset drivers)

[KpGaz]

13 hours ago, Mark Kaine said:

I would try the one from the mobo manufacturer website.

If I download it and start installing the chipset driver, will I get notified if I already have the latest driver or will it just overwrite it? 

[Mark Kaine]

1 hour ago, KpGaz said:

If I download it and start installing the chipset driver, will I get notified if I already have the latest driver or will it just overwrite it? 

Good question... In your case it would be better to uninstall first and then install the one from the website... It may not help,but that increases the chances of it helping a lot.

 

This might also be helpful.

 

https://www.amd.com/en/resources/support-articles/faqs/GPU-601.html