Question FPS Dropping to Zero

[qsold]

tldr: FPS intermittently dropping to zero in online games. Drop lasts from less than a second to a few seconds. Happens every 5-10 minutes give or take. Possible random stuttering in normal use (freezing while typing in forum). Root cause still undetermined.

System Specs:
CPU: Ryzen 9 5900X
GPU: EVGA Nvidia RTX 3070 Ti
Motherboard: MSI X570 Gaming Plus
RAM: Corsair Vengeance Pro 3600 18-22-22-42 2x16GB (plus the 2x16GB recently bought)
SSD: Intel 670P
PSU: Corsair RM750X

Preceding all of this:

Sunday: Installed some new RAM (2x16GB - same model as original 2x16GB). Ran Memtest. RAM failed hammer test; also had difficulty booting at 3600 (Motherboard loaded it at 2133mhz). Removed new RAM and ran computer.
Rebooted: Windows said it installed some updates (probably hadn't rebooted in a month or so).

Monday: First instance of the 0 FPS drop. Happened at least once during 1 game. Thought it might be server fluke.

Tuesday: Multiple instances of FPS drop. Began investigating.

Wednesday: Thought there might be an issue with my old RAM (perhaps it was loose after installing new RAM). Moved old RAM from slots 2/4 to 1/3. Same issue. Tried slower RAM from old PC (wouldn't boot... might have been a BIOS thing as XMP was still enabled and this RAM was much slower. No luck. Ran MEMTEST on old RAM. No failures.

Tried swapping out old RAM and put new RAM in 2/4 (testing seating issue hypothesis). Ran MEMTEST 8 passes no failure.

Thursday: Tried running with with new RAM in 2/4 after MEMTEST passed. Still issue.

Stuff I've checked/tried.

1. Checked temps of CPU/GPU. No troubling spikes.
2. Ran Prime95 to test stability. No issues.
3. Ran Furmark to test GPU stability. No issues (no drop to 0 FPS either... can't seem to reproduce in synthetic tests).
4. Ran 3DMark benchmark and stress test. Benchmark results were consistent with my original benchmark results from when I built the PC a year ago. Stress test score had 98.7% frame rate stability. No drops to 0.
5. Rolled back Windows installation. Did not fix.
6. Ran DDU. Updated Nvidia drivers. Did not fix.
7. Ran sfc /scannow. Said it found corrupted files but did not fix the issue.
8. Checked PSU cables. Appear to be attached.
9. Check Task Manager during drops. The only processes I see (briefly spike) were "Service Host: Local Service (Network Restricted)" and "Service Host: State Repository Service". If I wasn't paying close attention, I wouldn't even have been able to notice that they spiked so I'm guessing it's not the issue.
10. Will add more that I have forgotten later....
11. Tested in Single player game. Framerate would randomly drop from 180-200 to 30 fps. All in same area of game.
12. I also pinged google.com during the drops during online game session. No ping spikes.
13. In game ping monitor was consistent during the fps drop.
14. Updated BIOS from 7C37vAH to 7C37vAI.
15. Put GTX 1070 in. Same issue.
16. Reset CMOS.
17. Checked CPU pins and reinstalled.
18. Tested a different PSU.
19. Manually set timings on RAM. Can get 4x16GB to boot. Still intermittent FPS drops.
20. Both SSD/HDD are showing good on SMART and getting reasonable speeds.

Before moving to more drastic measures any suggestions?

At this point I'm thinking of swapping Video Card (I have an old GTX 1070 I can run) and swapping PSU to see if either of those makes a difference but I hate rerunning PSU wires.

I'm still suspicious of the RAM at this point but given I've tried both pairs separately and tried the different slots, I'm not convinced it's the issue either. Especially since the old RAM was perfectly stable. Has there ever been an instance of old RAM becoming corrupted by new RAM (RAM STDs?)

 

[Lutfij]

What BIOS version are you on for your motherboard at this moment of time? As for your GPU driver reinstall, did you use DDU prior to reinstalling your GPU drivers with the latest version?

 

[qsold]

What BIOS version are you on for your motherboard at this moment of time? As for your GPU driver reinstall, did you use DDU prior to reinstalling your GPU drivers with the latest version?

BIOS version is 7C37vAI. BIOS was 7C37vAH when I first encountered the problem so I tried writing the updated BIOS. Problem persisted. I'll add to the list.

Using DDU & reinstalling drivers didn't fix.

 

[qsold]

Newest observation:
2x16 GB RAM will boot at 3600
4x16 GB RAM, will not boot at 3600MHz. MSI EZ Debug light for CPU (not RAM) lights up. Not sure if that's normal or not. Reboots several time until it loads OS @ 2133MHz.

At this point I may take apart and reinstalling the CPU.

 

[Lutfij]

(plus the 2x16GB recently bought)
Are they both the same make and model, i.e, From Corsair + Vengeance? If so and they were bought at different times, please look up the version/revision number on the sticks from each kit.

I'd try and reflash the BIOS and then clear the CMOS after the BIOS has been successfully updated with the latest revision. It could be a board/BIOS issue or the memory controller on the processor being the issue. You might want to try and take your ram and processor over to another board and see if they work without a hiccup. If they work without an issue, it's possible that the board/BIOS is the issue. If they also exhibit issues like your platform, then the processor might be the culprit or different revision rams might be the anomaly.

 

[qsold]

(plus the 2x16GB recently bought)
Are they both the same make and model, i.e, From Corsair + Vengeance? If so and they were bought at different times, please look up the version/revision number on the sticks from each kit.

I'd try and reflash the BIOS and then clear the CMOS after the BIOS has been successfully updated with the latest revision. It could be a board/BIOS issue or the memory controller on the processor being the issue. You might want to try and take your ram and processor over to another board and see if they work without a hiccup. If they work without an issue, it's possible that the board/BIOS is the issue. If they also exhibit issues like your platform, then the processor might be the culprit or different revision rams might be the anomaly.

I did a CMOS clear before I tried the CPU reinstall. Also tested a PSU swap w/o success (adding to main list).

Regarding the RAM, yes, same made/model. The versions are different. I've confirmed I can get them both to run separately at 3600 or together at 2133 (haven't tested anything between.) But FPS drops 0 to occur in all scenarios now.

So I didn't have this issue prior to the RAM 4x16 install so is it possible that could have messed up memory controller? I'm definitely leaning on it being mobo/cpu at this point but don't have any other AM4 CPU/mobos I can swap with. Maybe I'll use this as an excuse to upgrade my home server and do a ryzen build for that so I have some parts to swap. I think that thing is still running 1150 socket.

 

[Lutfij]

Regarding the RAM, yes, same made/model. The versions are different. I've confirmed I can get them both to run separately at 3600 or together at 2133 (haven't tested anything between.) But FPS drops 0 to occur in all scenarios now.

So I didn't have this issue prior to the RAM 4x16 install so is it possible that could have messed up memory controller?
You should include the versions of the rams. The fact that the platform works with either of the ram kits without hiccups would indicate that the platform doesn't like you running mismatched ram kits. I say mismatched since the version numbers designate what IC's are used to make the sticks/kits of ram. Often times IC's are swapped in their manufacturing lines for favored prices or in other cases, sourcing materials faster/easier.

Further reading.

Using one kit of ram, manually input timings, frequency and voltage(maybe slightly higher) as what's shown on the stickered side of the ram kit in use.

 

[qsold]

Regarding the RAM, yes, same made/model. The versions are different. I've confirmed I can get them both to run separately at 3600 or together at 2133 (haven't tested anything between.) But FPS drops 0 to occur in all scenarios now.

So I didn't have this issue prior to the RAM 4x16 install so is it possible that could have messed up memory controller?
You should include the versions of the rams.

Old: 3.44
New: 8.35

The fact that the platform works with either of the ram kits without hiccups

This isn't true.

4x16 boots at 2133. Still get FPS to 0 problem.
2x16 (new v8.35) boots at 3600. Still get FPS to 0 problem.
2x16 (old v3.44) boots at 3600. Still get FPS to 0 problem.

would indicate that the platform doesn't like you running mismatched ram kits. I say mismatched since the version numbers designate what IC's are used to make the sticks/kits of ram. Often times IC's are swapped in their manufacturing lines for favored prices or in other cases, sourcing materials faster/easier.

Further reading.

Using one kit of ram, manually input timings, frequency and voltage(maybe slightly higher) as what's shown on the stickered side of the ram kit in use.

Thanks that's useful but at this point I can't get my original configuration to work without the FPS to 0 drop. I can try manually setting the clocks.

 

[drivinfast247]

Go to userbenchmark.com and run the test. Link the results here.

 

[qsold]

Go to userbenchmark.com and run the test. Link the results here.

I haven't been able to reproduce the beahvior in any synthetic program. I'm not sure what this will tell you but I ran it 3 days ago:

https://www.userbenchmark.com/UserRun/57079446

This is Time Spy's Stress test which is also unable to reproduce the behavior.
https://www.3dmark.com/3dm/84428684

Even running at 2x16 with my old sticks @ 2133, I get the FPS to 0 drop.

Example of behavior (Note I've replicated in 4 games including one offline game):

The behavior only shows up every 3-5min give or take and I suspect it's CPU/RAM related at this point. So I would need a CPU dependent benchmark that runs for, say, 10min to test. I've also tried to replicate in Furmark but I can't seem to replicate there.

 

[drivinfast247]

I haven't been able to reproduce the beahvior in any synthetic program. I'm not sure what this will tell you but I ran it 3 days ago:

https://www.userbenchmark.com/UserRun/57079446

This is Time Spy's Stress test which is also unable to reproduce the behavior.
https://www.3dmark.com/3dm/84428684

Even running at 2x16 with my old sticks @ 2133, I get the FPS to 0 drop.

Example of behavior (Note I've replicated in 4 games including one offline game):

The behavior only shows up every 3-5min give or take and I suspect it's CPU/RAM related at this point. So I would need a CPU dependent benchmark that runs for, say, 10min to test. I've also tried to replicate in Furmark but I can't seem to replicate there.

What are your GPU temps. while under load?

 

[qsold]

What are your GPU temps. while under load?

They're normal; same with CPU (see pt 1 in OP).

Also same issue observed with 2 different GPUs so it' not GPU (see pt 15 in OP).