GPU and CPU Usage Drops - Bug Bounty!

[NickNaylor]

Hi everyone,

 

I’ve had an issue with my relatively new (~3 months old) PC that is driving me absolutely insane and that I can’t seem to troubleshoot. I’m hoping that someone might be able to help me out. I’m so frustrated with this and have spent so much time trying to solve this issue that if someone provides a solution that solves the problem, I will send you a $50 Steam gift card as a gigantic thank you. All you need to do is send me your Steam name once I’ve identified a successful solution. I’m also happy to send you a $50 Amazon gift card if you’d prefer, though you would need to send me your e-mail address. To prevent any controversy, in the event that multiple people suggest a solution that solves the problem, the first person to make the suggestion will get the “bounty.”

 

THE PROBLEM:

When playing most games, I get intermittent, abrupt drops in CPU usage and GPU usage that result in substantial FPS drops. Usage on both the CPU and GPU will drop to 0% at the same time before returning back to their previous levels. There is no detectable slowing that I can see before these drops - they really do occur abruptly. Sometimes, the drops are associated with an abrupt drop in GPU core voltage, but not always. Games I have noticed this on include Rust, Halo, and Parkitect. It seems to happen somewhat randomly - for example, it doesn’t occur just when loading into new areas or in graphically “intense” scenes - and with some degree of regularity (maybe once every 1-2 minutes or so, sometimes more frequently, sometimes less). On Rust, I go from 100-110 FPS to 0 FPS for 0.5-2 seconds, then I jump immediately back to 100-110 FPS. On Halo, I can get 300+ FPS, drop to 0 FPS, then return immediately back to 300 FPS. The same thing occurs in Parkitect. I haven’t noticed this issue in Fallout 4, for whatever that’s worth. My system otherwise seems stable - I’ve never had a BSOD, abrupt shut down, etc. while gaming or otherwise

 

Here’s a short video showing what I’m talking about in Rust (running at max settings, 2560x1440). The drops occur at 1:07, 3:06, 3:21, 5:22, 7:22, 7:37, and 9:38 in this video. https://www.youtube.com/watch?v=9aYpAuR_n-4

 

And in Halo (running at max settings, 2560x1440). Sorry about the choppy video, that’s an artifact of OBS recording. The drops occur at 0:13, 2:15, 4:14, 4:30, 6:30, 8:30, 8:45, and 10:45: https://www.youtube.com/watch?v=NwRCJ7zu_is

 

Whether or not RTSS or HWinfo is running doesn’t make a difference and aren’t contributing to the issue as far as I can tell. The issue didn’t occur more or less frequently with OBS recording, so what you see in these videos is typical of how I experience the issue.

Interestingly, this problem does NOT occur when running benchmarks. I have had no problem running 3DMark Time Spy and Unigine Heaven (in the latter case, for long periods of time), and I was able to run Prime95 overnight with no errors. I also don’t notice it when using my CPU for non-gaming purposes, nor when I’m playing VR (I have a Valve Index that I run at max resolution and 144Hz). Half Life: Alyx, Beat Saber, and Superhot VR all run well without this behavior occurring.  It seems to be limited strictly to gaming on my desktop monitors, and unfortunately it completely ruins games that I like to play (FPS). This would be a tolerable problem if it were fixable by simply dropping settings, but even at low settings and low resolutions in these games, the issue still occurs.

 

BUILD (all of my components were bought new):

Motherboard: ASUS ROG Strix X570-E (though I’ve also tried putting everything on an ASUS X570 PRIME-PRO board)

 

CPU: Ryzen 9 3900X

 

GPU: PNY RTX 2080 Ti (inb4 “y no 30XX”)

 

RAM: Corsair Vengeance 16 GB DDR4 x 4, running in DOCP at 3200 MHz

 

Drives: Samsung 970 EVO 1 TB SSD (OS and games are installed on this drive), WD Black 4 TB 7200 RBP HDD (data only)

 

PSU: SuperFlower Leadex Gold 1300W

 

Monitors: Samsung 32” 1440p 144Hz x 2, LG 32” 1080p ultrawide

 

OS: Windows 10 with all updates installed

 

3DMark Time Spy Score: https://www.3dmark.com/3dm/51076894

 

During gaming, my CPU temps typically run around 65-70C and my GPU runs around 70-80C. I know my temps are high and I’m planning on doing a custom loop in the near future to try and cool things down, but I don’t think I’m getting any thermal throttling. My CPU never goes above 80C and my GPU never goes above 85C. My CPU and GPU are running stock with no overclocking applied on my part.

 

THINGS I’VE TRIED, NONE OF WHICH HAVE SOLVED THE PROBLEM:

Hardware:

Updating chipset drivers

Changed PCI power saving to off in Windows power settings

Changed wireless adapter power saving to low power saving in Windows power settings

Turning on/off HPET

Running RAM at stock

Running Memcheck and Windows Memory Diagnostic - no issues identified in RAM

Re-seating the GPU and installing the GPU in an alternate PCIe slot

Changing the motherboard (from an X570 PRIME-PRO to a ROG Strix X570-E)

Unplugging the front USB panel panel on the case

Running only a single monitor using different ports on the GPU

Removing my CPU cooler and reapplying thermal paste (resulted in a slight drop in temps, otherwise no change in the behavior)

Unplugging my HDD from the motherboard and only running the SSD

Disabling FreeSync on my 144Hz monitors

Yes, my monitor cables are plugged into the GPU and not the motherboard

 

Software:

Re-installing Windows - I get the same issue when the only things installed are stock Windows, Nvidia drivers (without GeForce Experience), and Steam - the drive was formatted before installation

Updating GPU drivers (currently running the most recent Game Ready drivers) and BIOS

Changing Windows power plan settings (currently running Ryzen High Performance, no change when running with Ryzen Balanced or the Windows High Performance setting)

Tweaking a variety of Nvidia driver settings (namely, changing the power management mode to “prefer maximum performance” though I’ve also tried settings suggested in a variety of guides on YouTube)

Using DDU to uninstall and reinstall Nvidia drivers

Uninstalling GeForce Experience and only running the drivers

Running games in full screen/borderless/windowed modes

Disabling the Steam overlay

Disabling the Windows Game Bar

Lowering settings in-game (including running at 1080p max settings and 1080p low settings) - my average FPS increase and GPU usage drops, but I still get the behavior

LatencyMon shows no “red” readings when gaming

As far as I can tell, there are no “rogue processes” that are acting up to cause the slow down that I can notice. I’ve watched processes in Task Manager and Process Manager and haven’t seen anything indicating weird disk usage or network usage or processes that abruptly utilize significant amounts of CPU/GPU.

 

In short, essentially anything mentioned in a thread on LTT forums or Tom’s Hardware forums by searching “CPU and GPU usage drops while gaming” I’ve tried apart from replacing the CPU, GPU, and PSU (mostly because this is my first PC and I don’t have extra components hanging around). Most of the threads I’ve read include the above recommendations.

 

I’m more than happy to post anything else that might be helpful and answer any questions. I really appreciate any help anyone can provide. If this was something that only occurred at max settings I wouldn’t mind because I could just lower quality settings or the resolution (even though this hardware should be able to run these games with no problem) - I’m not too bothered by not being able to run things at the best quality possible. The fact that this seems to occur irrespective of the settings, though, makes this an extremely frustrating issue.

[Guest]

Only 2 things I can think of off my head.

Try running an older GPU driver.

Update the motherboard bios.

 

[NickNaylor]

3 minutes ago, ShrimpBrime said:

Only 2 things I can think of off my head.

Try running an older GPU driver.

Update the motherboard bios.

 

Do you have a particular driver version you would recommend? I've tried a couple of older versions to no avail, but they were chosen somewhat at random.

 

As far as updating the BIOS, I'm running the most recent BIOS for my motherboard. I had the same issue with the most recent BIOS on the original motherboard I had as well.

 

[Guest]

22 minutes ago, NickNaylor said:

Do you have a particular driver version you would recommend? I've tried a couple of older versions to no avail, but they were chosen somewhat at random.

 

As far as updating the BIOS, I'm running the most recent BIOS for my motherboard. I had the same issue with the most recent BIOS on the original motherboard I had as well.

 

OK, you hadn't mentioned these, so thought to start there. 

 

"Corsair Vengeance 16 GB DDR4 x 4, running in DOCP at 3200 MHz"

edit.

Is this always enabled through all the testing?   I see you tried all 4 at stock. sry.

Have you tried just two sticks at Jedec frequency?

Edited October 1, 2020 by ShrimpBrime
corrections

[NickNaylor]

9 hours ago, ShrimpBrime said:

OK, you hadn't mentioned these, so thought to start there. 

 

"Corsair Vengeance 16 GB DDR4 x 4, running in DOCP at 3200 MHz"

edit.

Is this always enabled through all the testing?   I see you tried all 4 at stock. sry.

Have you tried just two sticks at Jedec frequency?

I've thought about using only 2 DIMMs but haven't actually tried it yet. I thought maybe the 4 DIMMs was causing some sort of issue, although that would be irritating to not be able to use all of my RAM slots. I can try running just 2 DIMMs this afternoon, though, and see if that results in an improvement.

 

Edit: And just to make sure I understand, by "JEDEC frequency" I assume you mean stock DDR4 frequencies, right? I can certainly try that. Running at stock vs. OC hasn't made a difference with 4 DIMMs, but I will try both settings with 2 DIMMs.

[Guest]

3 hours ago, NickNaylor said:

I've thought about using only 2 DIMMs but haven't actually tried it yet. I thought maybe the 4 DIMMs was causing some sort of issue, although that would be irritating to not be able to use all of my RAM slots. I can try running just 2 DIMMs this afternoon, though, and see if that results in an improvement.

 

Edit: And just to make sure I understand, by "JEDEC frequency" I assume you mean stock DDR4 frequencies, right? I can certainly try that. Running at stock vs. OC hasn't made a difference with 4 DIMMs, but I will try both settings with 2 DIMMs.

Right stock.

Pay attention to which Jedec profile the board chooses. The memory will come with more than one profile. 

 

But yeah, try just 2 dimms stock and overclocked.

32GB of system memory should be plenty for you though.

I understand filling four slots looks good, but not a requirement to run games smoothly.

[NickNaylor]

4 hours ago, ShrimpBrime said:

Right stock.

Pay attention to which Jedec profile the board chooses. The memory will come with more than one profile. 

 

But yeah, try just 2 dimms stock and overclocked.

32GB of system memory should be plenty for you though.

I understand filling four slots looks good, but not a requirement to run games smoothly.

Sir, you are a damn American hero.

 

After taking out 2 of the DIMMs, the system wouldn't post for some reason (got a Q-code error that wasn't listed in the motherboard manual, so I'm not sure what the issue was) and kept restarting itself but, after getting it to post, I was able to game smoothly. Definitely took a FPS hit without DOCP on, but who cares - at least it was stable. I was curious to see if this was a temperature issue since even with just 2 DIMMs my temps were at ~50C. I plugged in all 4 running at stock frequencies and temps jumped to ~55-58C, then applied the overclock and they jumped further to ~65C. Interestingly, though, the issue seems to have resolved (???). Even with 4 DIMMs and DOCP on, I'm no longer getting the drops. And my baseline FPS has increased - I'm now getting a steady 130-140 FPS in Rust compared to 100-110 before (???).

 

I wonder if I maybe something about how the DIMMs were seated was screwed up? Dust in the slot or on the contacts? I also changed the order of the DIMMs in the slots... I'm not sure why that would matter but maybe it had some effect? Maybe booting in the system into a safe mode (since it wouldn't post) did something... somehow? Before this all of the DIMMs were identified and my system was listing 64 GB as available, but maybe something was just off enough to cause instability. Also weird since if it were just a seating issue I would've expected it to be resolved after reinstalling everything on the new motherboard unless I managed to screw it up twice.

 

Regardless, this seems to have solved the issue. Not sure how, not sure why, but I'll take it. PM me your Steam username and I'll send you the gift card or your e-mail address if you'd prefer an Amazon gift card. THANK YOU.

[Guest]

Ah good information.

Yes, you should put a fan on the memory if it's running 65c.

Definitely a ram reseat could have done the trick.

Maybe after the memory was pulled and then reinstalled, the system may have trained the memory with different sub timings.

Or even all of the above?

You're welcome but I dont need a payment, but if you insist, my son (8) uses my account so I can moderate, and I'd just let him have at it.

Steam account name is right under my avatar. And thanks.

[NickNaylor]

On 10/1/2020 at 4:48 PM, ShrimpBrime said:

Ah good information.

Yes, you should put a fan on the memory if it's running 65c.

Definitely a ram reseat could have done the trick.

Maybe after the memory was pulled and then reinstalled, the system may have trained the memory with different sub timings.

Or even all of the above?

You're welcome but I dont need a payment, but if you insist, my son (8) uses my account so I can moderate, and I'd just let him have at it.

Steam account name is right under my avatar. And thanks.

Well, sadly it looks like I declared victory too soon - the issue has returned. No real big changes on the system other than installing Windows updates, but I'm not sure that that has anything to do with it. On the plus side, looking at my RAM temps caused me to install another case fan blowing right on the DIMMs, which dropped my temps ~20C, lol.

 

The fact that messing around with the RAM (temporarily) resolved the issue makes me think this is fundamentally a RAM problem. I dug through my motherboard QVL and my RAM actually isn't listed there. I'm unfamiliar with how important it is to use QVL-listed parts... could this be the issue? Thinking of running to Microcenter today to get some different RAM that is actually listed on the QVL.

[Guest]

51 minutes ago, NickNaylor said:

Well, sadly it looks like I declared victory too soon - the issue has returned. No real big changes on the system other than installing Windows updates, but I'm not sure that that has anything to do with it. On the plus side, looking at my RAM temps caused me to install another case fan blowing right on the DIMMs, which dropped my temps ~20C, lol.

 

The fact that messing around with the RAM (temporarily) resolved the issue makes me think this is fundamentally a RAM problem. I dug through my motherboard QVL and my RAM actually isn't listed there. I'm unfamiliar with how important it is to use QVL-listed parts... could this be the issue? Thinking of running to Microcenter today to get some different RAM that is actually listed on the QVL.

QVL is and isn't important. Mostly a list of memory that was tested before release time. That memory was available then, while there are new modules released after the fact that wheren't tested. We'll call it a grey area. This information only states compatibility with the board, doesn't speak for the memory controller.

 

I can tell you first hand my Vengence LED memory kit (3000mhz Hynix) is Very difficult to play with. It does not like going past it's rated speed at all, and on Gen 1 chips, it struggles to get past 2667mhz, the memory controllers top rated frequency.

 

Unfortunately, with an issue like this one, it's a 3 part story. Cpu, board and memory modules. 

[Guest]

Additional Comment.

To further back our support with memory kits, this thread below, has similar issues to yours. The memory kit looks similar.

This one is also a kit of Corsair Vengence memory modules. Looks like common frequency issues.

 

[NickNaylor]

4 hours ago, ShrimpBrime said:

QVL is and isn't important. Mostly a list of memory that was tested before release time. That memory was available then, while there are new modules released after the fact that wheren't tested. We'll call it a grey area. This information only states compatibility with the board, doesn't speak for the memory controller.

 

I can tell you first hand my Vengence LED memory kit (3000mhz Hynix) is Very difficult to play with. It does not like going past it's rated speed at all, and on Gen 1 chips, it struggles to get past 2667mhz, the memory controllers top rated frequency.

 

Unfortunately, with an issue like this one, it's a 3 part story. Cpu, board and memory modules. 

Well, unfortunately a RAM replacement (3600 CL16 on the QVL list) didn't seem to make a difference, either with DOCP on or off. I've been tweaking voltages in BIOS to see if that makes a difference... bumping the DRAM voltage to 1.4V and gradually bumping the SoC voltage to 1.15 also didn't make much of a difference.

 

At this point, I think I've eliminated the motherboard, RAM, and software-related issues as causes. Maybe I just need to do some very meticulous tweaking the BIOS to sort it out, I'm not sure.

[AbdulBais7]

12 hours ago, NickNaylor said:

Well, unfortunately a RAM replacement (3600 CL16 on the QVL list) didn't seem to make a difference, either with DOCP on or off. I've been tweaking voltages in BIOS to see if that makes a difference... bumping the DRAM voltage to 1.4V and gradually bumping the SoC voltage to 1.15 also didn't make much of a difference.

 

At this point, I think I've eliminated the motherboard, RAM, and software-related issues as causes. Maybe I just need to do some very meticulous tweaking the BIOS to sort it out, I'm not sure.

I solved my Issue, are you running your Windows on UEFI or Legacy Mode? because for my issue I just reinstalled my windows on UEFI Mode and Everything is fine now, so do check that out

[Mark Kaine]

3 hours ago, AbdulBais7 said:

I solved my Issue, are you running your Windows on UEFI or Legacy Mode? because for my issue I just reinstalled my windows on UEFI Mode and Everything is fine now, so do check that out

that's something I've been wondering reading your thread, why did you even run in legacy mode, any particular reason to do that?  seems counter productive using newer hardware.

[AbdulBais7]

3 hours ago, Mark Kaine said:

that's something I've been wondering reading your thread, why did you even run in legacy mode, any particular reason to do that?  seems counter productive using newer hardware.

well I shifted from i5 3570K which was running fine in Legacy and I just installed my windows with the same Legacy USB Drive so I didn't knew that It would cause the issue, tho I am still facing freezes but not like before 

 

[NickNaylor]

12 hours ago, AbdulBais7 said:

I solved my Issue, are you running your Windows on UEFI or Legacy Mode? because for my issue I just reinstalled my windows on UEFI Mode and Everything is fine now, so do check that out

I'm running in legacy BIOS mode. I'm going to try a different GPU to see if this helps the issue - if it does, then I think that would pinpoint the issue as the GPU. If that doesn't work, then I'll do yet another Windows install and be sure to do a UEFI install.

 

Edit: the fact that you're still having the issue would suggest to me that this isn't the actual root of the problem. But if it's improved, that's at least something.

[NickNaylor]

Well, I picked up an RX 5700 XT (only GPU I could get today) and get the exact same behavior, so I think I have ruled out a GPU issue.

 

Guess I'll try a Windows UEFI install and see where that gets me...

[AbdulBais7]

13 hours ago, NickNaylor said:

I'm running in legacy BIOS mode. I'm going to try a different GPU to see if this helps the issue - if it does, then I think that would pinpoint the issue as the GPU. If that doesn't work, then I'll do yet another Windows install and be sure to do a UEFI install.

 

Edit: the fact that you're still having the issue would suggest to me that this isn't the actual root of the problem. But if it's improved, that's at least something.

Can you exactly tell me your PSU model, etc? I have completely fixed my problem

 

[AbdulBais7]

7 hours ago, NickNaylor said:

Well, I picked up an RX 5700 XT (only GPU I could get today) and get the exact same behavior, so I think I have ruled out a GPU issue.

 

Guess I'll try a Windows UEFI install and see where that gets me...

UEFI Windows do works faster then Legacy, while a quick fix which I found myself is Turn off the PBO (Precision Boost Overdrive), down clock Memory a bit like from 3200MHz to 3000MHz and Install RYZEN MASTER and Auto OC your CPU, after these 3 things the issue seems to be Fixed completely 

[NickNaylor]

59 minutes ago, AbdulBais7 said:

UEFI Windows do works faster then Legacy, while a quick fix which I found myself is Turn off the PBO (Precision Boost Overdrive), down clock Memory a bit like from 3200MHz to 3000MHz and Install RYZEN MASTER and Auto OC your CPU, after these 3 things the issue seems to be Fixed completely 

Great, thanks for the suggestions. I'll give this a go this morning.

[NickNaylor]

1 hour ago, AbdulBais7 said:

Can you exactly tell me your PSU model, etc? I have completely fixed my problem

 

My PSU is a 1300W Super Flower Leadex Gold.

[AbdulBais7]

18 minutes ago, NickNaylor said:

My PSU is a 1300W Super Flower Leadex Gold.

thats more then plenty for your PC, yeah do check the Tweaks I suggested 

[NickNaylor]

44 minutes ago, AbdulBais7 said:

thats more then plenty for your PC, yeah do check the Tweaks I suggested 

Well, those tweaks didn't seem to resolve the issue - installed Windows on a UEFI drive, disabled PBO in BIOS, and am using the "auto OC" setting in Ryzen Master. Thanks for the suggestions, though - worth a shot.

[NickNaylor]

I thought I'd also add that I ran Memtest86 overnight on my current RAM setup (the 64 GB Corsair Vengeance Pro 3200 kit) the other day and had no issues reported.

[AbdulBais7]

3 minutes ago, NickNaylor said:

I thought I'd also add that I ran Memtest86 overnight on my current RAM setup (the 64 GB Corsair Vengeance Pro 3200 kit) the other day and had no issues reported.

your pc freezes while playing games right ?