Random low frame rate intervels no matter how much is running

I’ve created threads regarding multiple problems before, but have narrowed it down to just this now, so I thought I would create another. I can share the others and they can be deleted if needed.

I’m still getting nowhere regarding the main issue which are these spikes of low frame rates. Been through every driver version from 450 to 470 and none have made any difference to this problem. I’ve also got a new clean version of Ubuntu 20.04 and tested that and it has exactly the same problem.

At this point, some may suggest it is a hardware issue, but how can that be so intermittent? These periods of what feels like low frame rates occur so randomly, so is surely has to be driver related. Saying that, when I switch to the Nouveau drivers seems to not have the issue. While I can’t play games due to it not using the motherboard graphics as well as it could (still passing through the card), it no longer seems to have the random low FPS issue which can be noticeable when doing things as basic as dragging windows around the screen as they are choppy.
Just to be sure, I used Nouveau for over a week, and other than poor performing games, I never experienced the intermittent issue with any other program. I at one point wondered if it was related to lag from my input devices, as watching videos is perfectly smooth no matter what drivers I use, but watching others play 3D games while I stand still for example, the choppiness is still noticeable. It is just totally random. Sometimes totally smooth, sometimes a few frames per second. And it is exactly the same experience on my desktop with virtually no applications running.

If not using Nvidia drivers can help, surely there is a compatibility problem with my card, Ubuntu and the drivers all the way from 450 to 470. Before I had this issue, I never thought about graphics drivers as everything just worked, including the first month I had my new PC (all the way back in march). I believe it must have had 445 or below back then and it worked fine.

The other bizarre thing that is still the case is plugging another monitor into my graphics card seems to pretty much prevent these odd frame rate drops, but introduces other issues such as tearing. I’ve done a lot of research on this and I just can’t understand why I’m having such issues after having so many driver updates not fixing the problem. Also strange to not read up on anyone having the same experience. The only other possibility (that doesn’t seem realistic) is that this problem started to occur sometime after the first time I had used and them removed a 2nd monitor. I’m not sure if this or the driver update caused it. Surely the latter seems unrealistic though.

I am totally stuck with what to do as unless I find out it is a hardware fault, I can’t just send it back to the retailer that custom built it for me as there may not be any issues. I just don’t know how to find out exactly what they are. I still want to try the 445 drivers or earlier. I’ve tried “sudo apt-get install nvida-driver-445” but don’t seem to be able to fetch them.

If I can confirm that it is the card and software, then I think I will have to accept it and go for AMD instead. But I first need conformation of what the issue is, which I’ve been trying to work out for 5 months.

Have uploaded my latest bug report

nvidia-bug-report.log.gz (294.1 KB)

You do not want to be using motherboard graphics, just plug everything into the dedicated GPU.

If you get tearing issues, open nvidia-settings, then under display configuration click advanced and check force composition pipeline.

If you still get the “random” low FPS, what I would suggest is checking your PSU. If the card is not getting enough power, it will stutter.

Or something in your case is getting really hot.

Will try to reply to each of your lines by what i have tried.

1:
I am aware I won’t want to be using the motherboard graphics, but I am doing so for testing reasons. Is it correct that if I use the Nouveau drivers, and have my monitor connected though the graphics card, is it using the motherboard graphics rather than Nvidia? I did this as I want to confirm if the card can work with the outputs still being used, but not the Nvidia drivers. With Nouveau drivers, while i can’t play any demanding games at all, I never experience the intermittent choppiness of basic things like dragging windows about.

2:
I first wondered what was causing tearing many months ago, and I eventually figured out it was the force composition pipeline boxes in NVIDIA config not being ticked. When using 1 monitor, I never have this problem after I make sure the boxes are ticked. When I connect 2 monitors and have one as a 2nd screen, I don’t seem to be able to remove the tearing, no matter what I do. I don’t want to be using 2 monitors though as I don’t have the space. The most bizzare thing is other than the tearing glitches 2 monitors introduces, it seems to pretty much get rid of these bursts of low frame rate. I just can’t understand this. But this pattern has been the same every time I’ve tried using 2 monitors in the past 5 months I have had this problem.

3:
I would be surprised if it was this unless using the Nouveau drivers uses a lot less power. It can have this problem sometimes with in seconds of logging in and just dragging windows about, and I somehow think the card will be using next to no power at all at this stage. And I can for some reason watch videos without anything ever being choppy. My power supply is the Corsair RM650x and I think it will be capable of delivering more power than I actually need.

4:
I’ve recently been managing my case to check the temperatures and have been adding fans as well as overclocking the graphics card (just fan speed). The maximum temperature I allow it to get to is 75 degrees, which for my model is 13 degrees below critical, so it shouldn’t be an issue. Even when I haven’t been gaming at all all day, the GPU temperature remains below 35 degrees, and the CPU fan is automatically barely spinning and the power supplies metal body feels cool to the touch.

I am trying to confirm weather it is hardware related, or the Nvidia drivers. The most strange thing is that it worked fine for the first few weeks I had the PC.

If I work out it is a hardware fault, I will be able to get back to the retailer that built the PC for me.

If I work out that it is related to the drivers, then there could do with being an update that actually fixes this problem, or a method explained somewhere on how to get older drivers (prior to 450). My really old PC is still on Ubuntu 14.04 and is using 370 drivers or older, and that works fine for what it is capable of with an old GT630.

If I find out that it is certainly an Nvidia related problem that I just can’t find a solution to, I think I will have by an AMD card and to try and sell this card (if I know there is not a hardware issue with it).

Hopefully this explains what I’ve tried so far!

Thanks.

I created an account just to reply.

YES, you are absolutely correct.

N V I D I A F***D UP!

Their recent drivers screw over A TON of gamers.
This exact random FPS drop happens to a lot of us,
AND NVIDIA HAS FAILED TO ADDRESS IT.

I would appreciate if you could describe your issues and the amount of time you have been having them. I’m still having them and I’ve had the issue since April last year!

Since this post, I returned the graphics card to the retailer, and they told me it was faulty, however, when I got the new card back, everything was exactly the same. So whatever fault there was, that wasn’t related to this issue. I they sent the entire system back, and they again confirmed there were issues (though i don’t believe they were related to this). They generously replaced every internal part other than the graphics card which had already been replaced. They they confirmed it to be working, but when I got it back, the problem was still there. They clearly are finding different issues to me that I’m not noticing and don’t seem to be finding my big problem.

It still had the problem, so after doing yet more tests confirming this, I sent it back again and asked if they could also test it in Ubuntu, and they couldn’t find any problems.

I’ve had enough of returning this computer now!

Some strange things that I can confirm is that this problem does not occur on another desktop PC in my house that also has Ubuntu 20.04 and the same driver version. I’ve tried 3 nvidia cards in that computer, and all are fine. No matter what nvidia card is in my PC, this problem is always there when using the drivers.

What is now making me more certain that it is driver related is that I took home an old 256mb ram ATI graphics card from a computer from a skip at work for testing and that for what it is capable of, does not have these intermittent periods of lagging. I’ve been using a program called GpuTest_Linux_x64_0.7.0. Running the most basic test on that with any nvidia card on my machine makes all windows horrendously choppy to drag about. Probably 5 - 10 FPS. This is what intermittently happens in gaming. It can be perfect at times, but then randomly go choppy even when the scene is not demanding at all.

Given that ATI card worked and used open source drivers, I’ve just bought a Dell Radeon R5 430 2GB 2nd hand. It wasn’t much and I don’t expect amazing performance from it, but if this also works, then it 100% confirms that Nvidia has a major driver issue. However, it isn’t with every PC.

It seems to me the nvidia drivers for the past 9 months has a compatibility issue with Ubuntu and maybe certain motherboards? I have the asus b550m-a wifi. If the card I have purchased works, I will have to look into getting an AMD card that is of similar performance. I somehow expect it will be around have the cost of my PC which I’m not happy about, but I have no option and I’m sick of not being able to use this computer for much as I spend £1000 on it. Over the past month and each time it has been sent back to the retailer, I’ve been using my 10 year old PC, which graphics wise has far less issues as the old nvidia drivers were reliable, at least with older hardware too.

If this is down to nvidia and their drivers, and this issue has been around this long, i can’t trust them to fix it, so I think i will have to sell my card and hopefully someone with windows will buy it.

I’ve taken a look at your initial bug-report.log and it looks like the the gpu was unable to properly clock up. When hooking up a second monitor, a bug in the driver triggers the gpu to switch to the highest performance state P0, so that’s probably why your problems vanished in that case.
So it was your best choice to just return it to the vondor to get your money back.

As I said in my previous post, they did find a fault with my GPU that I sent back on it’s own first time. When they told me this, it got my hopes up, but the new one had the same issue, so I must have been noticing something different. I thin returned my whole custom built PC to the retailer that put it together and they found faults again and generously replaced every single component excusing the case and fans and the graphics card, as that seemed to be fine according to them, and it was replaced last time. However, when I got it back, it had exactly the same issues. I then did many tests and the issues were the same as I described in my original post. I returned my system yet again and they could not find any issue, even in Ubuntu, which puzzles me.

After finding out a bit more with the stress test, this is what I now know seems to cause the issue:

  • running the most basic test in GpuTest_Linux_x64_0.7.0 will make every window horrendously choppy to drag around the entire time it is in use, even though the end result of the test states it was running at 60FPS which is correct for my monitor.

  • If I select open source nouveau drivers, this test can go ahead and windows can be dragged around perfectly smoothly. However, as expected for these drivers, they do not work well with NVIDIA cards well at all. The in game performance is consistently very weak, and has issues with fullscreen and many 3D games in my case. But these drivers have been like this for years from what I remember.

  • If I use 2 monitors, it does seem to reduce the extent of the low FPS periods, but not fix it entirely. However, using this graphics card with the same drivers on another computer with the same version of Ubuntu, it has no issues at all.

It surely must be something to do with Linux NVIDIA drivers not getting on with my motherboard or something.

Please put some load on the gpu, e.g. by running an Unigine demo. After a while, please create a new nvidia-bug-report.log with the demo still running and attach that.

Would I be able to create this bug report while running the program i already have? GPUtest? I’m unfortunately poor at understanding even simple instructions at times as to how to install things. So I don’t know how to set up Unigine.

If I can just do this bug report while running GPUtest (which does create the issue i encounter), and that is enough, i would rather do that. Could you remind me how to create an nvidia-bug-report?

Many thanks.

Of course, this will work as well.

To create the report, run
sudo nvidia-bug-report.sh

@ [thegianthogweed97] I’ve had this issue starting around Sept 2021 with a driver update. Games play well, then frames per second drop to nearly 20% for about 30-45 seconds every 1-2 min. I restored my PC since my comment yesterday (I’m the one that said N V I D I A F***D UP). I couldn’t log in to the NVidia accound so I had to create another one. Anyway, yes… Clean windows install. Clean drivers install. STILL LAGGING and DROPPING FRAMES!

To confirm it is the same issue as mine, having Ubuntu is quite a big part of it. If you don’t use Ubuntu, we can’t exactly confirm the issues we are facing are caused by the same thing.

I seem to be having issues uploading any content at all. It says this every time:

“Sorry, there was an error uploading that file. Please try again.”

I have 2 logs, 1 created with my PC running the stress test and one created without.

I also want to upload 2 videos demonstrating this low FPS issue.

I’ve now also tested a Radeon R5 430 2GB and while the actual performance is lower than I expected, again, there is no issue at all when running the most basic GPU test. There is also, consistent, acceptable performance when at a low resolution in games.

This is something I am unable to replicate no matter what NVIDIA card I use with my new PC. Every game, or even when i just have many windows open, it can randomly feel slow and choppy when dragging stuff about when there is next to no load on the GPU at all. Since i can prevent this happening by switching to nouveau open source drivers, it simply has to be a glitch with the drivers? I think it has been an issue ever since 450 or maybe even 445. I remember the first month when i had this computer back in march and I did not have these issues.

If my computer company can’t find any issues, and this is not a widely reported issue, than just what is this problem? It seems ridiculous to think that using a 256mb ram ATI card is more tolerable, predictable and reliable with this PC than my 4gb GTX1650. I am pretty certain it is a driver compatibility problem with my motherboard. As my dad’s PC also uses Ubuntu 20.04 LTS like mine, has the same nvidia drivers but has no problems at all with this card!

I’m just going around in circles and have been since April last year. At this rate, I’m going to accept that it could be an expensive fix. I just want it to work.

One option I do have is having to buy windows. I have never had Windows as my operating system at home, and haven’t used it regularly in simply years. I personally don’t like it at all but can respect it is better for nvidia things. However, as it still is common for plenty of gamers to use Nvidia and Linux (I have for the past 8 years), a compatibility problem lasting this long is very poor in my opinion.

Please let me know how to upload things here, as this glitch I explained is still there.

Seems the forum’s virus scanner is broken, making it difficult to upload the .gz file. Please try unzipping it and uploading the plain.log file, some users had success with that.

I also tried uploading some simple PNG images showing the problem. It still won’t let me attach anything.

I think I’m now at the stage i will probably have to get used to windows, and use a virtual machine for Linux if I want it to feel familiar. Hopefully windows drivers will be much better.

Would still rather share what I wanted to, so hopefully it will be fixed soon.

If you have internet connection, you can use pastebinit to upload it from console.

  • install pastebinit (sudo apt install pastebinit)
  • unzip logfile (gunzip nvidia-bug-report.log.gz)
  • upload logfile (pastebinit -i nvidia-bug-report.log)
  • note down and post the url you’re given

Seems to have worked now, will see if it posts!

Number 2 is doing this bug report while not running GPUtest.

nvidia-bug-report.log (2.9 MB)
nvidia-bug-report2.log (2.9 MB)


The two videos show the test being done. The first one being on my PC built by SCAN (which has the choppy issue) and one on another PC, that is using the same drivers and version of Ubuntu. Only big difference really is the motherboard. This other PC visually is fine, and fine for gaming too.

There’s something really wrong with that system. Regarding the nvidia gpu, it seems while running gputest, the gpu has nothing to do because it doesn’t get enough data over the pcie bus, very low TX/RX loads as if something is strangling it.
The second oddity is the boot timing. The system is booting fast, finds your first samsung drive, then the second intenso, all fine. Until it hits your crucial ssd, the kernel/mobo having a hard time configuring it, setting it to dma takes 6s and afterwards the kernel pauses for 30s (possibly running into some timeout) until it finally mounts the root fs and continues boot.
Can you confirm your boot to login takes about 50s?
Please try disconnecting the crucial drive and also disable iommu, either in bios or by setting kernel parameter iommu=off. Then run gputest again and create a new nvidia-bug-report.log. Also, please post the output of

grep nvidia /etc/modprobe.d/* /lib/modprobe.d/*

Please forgive me for understanding very little. I really struggle to follow instructions.

I’m impressed with the level of information you know now. I’m assuming these things i sent give you a great deal of information.

You are correct that my current boot to log in does take around 50 seconds, then 30 or so to log in. This is on my 960gb drive however. This is odd as once it is logged in, the system is fast and responsive, but has these spikes of lag that I’ve described.

I am entirely aware that this drive is too clogged up with outdated stuff (for example, it hasn’t had a totally clean install of Ubuntu since around 14.04) but this isn’t related. Part of the reason I haven’t had a totally clean install for ages is that I’m not good with getting used to new interfaces or other changes.

However, as a test, I did try having a clean install of 20.04 on a fast kingston hyperx savage memory stick, and did the same GPU test. Also had the same inconsistent low fps issue in games. Did the same test by doing a clean install of Ubuntu on my crucial SSD. The experience is exactly the same.

one thing that my PC is doing since I put it back together is booting up by default from the crucial drive. I have to manually select my 960gb integral drive for my main OS drive that I’m wanting. This will probably be something related to boot priority. But I don’t think any of this should be related to the graphics issues should it? As I said, not using nvidia drivers removes this specific issue.

Your final points, i don’t fully understand where i need to go in the bios to disable “iommu”, or have any idea what it is. Also not sure about the the command at the bottom of your post.