Strange no one has reported it earlier but I often see visual glitches and artefacts on GTX 1060 6GB while running NVIDIA drivers 381.09.
The bug manifests itself in a mouse cursor leaking/ghosting in various places when you Alt Tab between applications or e.g. if you’re in a web browser and the input field is scrolling.
I’m 100% sure previous stable drivers didn’t have this issue.
Here’s a video of this problem:
My GPU is running at stock speed/voltages.
I’m not running any compositing window managers - what you’re seeing is a usual plain X11 desktop session.
Thanks, birdie. Does this reproduce without the {ForceCompositionPipeline=On, ForceFullCompositionPipeline=On} MetaMode flags and TripleBuffer option? How about with a window manager other than kwin?
Finally, does it go away if you enable the composite extension and use kwin’s compositor?
What was the previous stable drivers didn’t have this issue? Is it regression? Can you share reproduction steps for this issue? Do we need to open any application to trigger this issue?
Hi gqman69, Can you share video showing this issue on your desktop?
Previous stable drivers (378.13) didn’t have this bug.
There are no steps to reproduce this issue - it’s very random and almost impossible to trigger at will.
You just move mouse and occasionally you notice this issue.
There’s one thing that makes the bug appear more often though - run any intensive CUDA application in background and you’ll get this mouse pointer ghosts quite often.
Running some application is mandatory - so far I’ve managed to notice the problem in Mozilla Firefox, Google Chrome and Akregator.
Hi gqman69,
Could you record and post a video showing this issue? And nvida bug report
Hi birdie,
[i]
if you’re in a web browser and the input field is scrolling.[/i]
Can you attach video of this?
>>There’s one thing that makes the bug appear more often though - run any intensive CUDA application in background and you’ll get this mouse pointer ghosts quite often.
What CUDA intensive application you are running?
Hi birdie, Our QA need some more info to repro this issue:
As the repro rate is very low and to reproduce, Can customer provide some consistent steps to repro? as this is regression and consistent repro could help us to find the regression point
From where did customer had downloaded the Zcash/Ethereum miner “As i am getting dependencies issue which required higher version of libraries”? Can we get the downloadable link, Steps to configure , installation and step to run the application? If dependencies are downloaded from outside what are they ?
Is there any other sample which will help to replicate this issue soon?
Do customer is using cuda sample?
What steps were performed to see this issue? do we need to keep the mouse moving continuously?
How often does this issue occur?
Plain X11 desktop - Is that mean you are able to repro this issue on bare X started with X or Xorg or xinit commands?
There are no consistent steps unfortunately however it’s worth noting that I’m running the X.org server from CentOS 6.7 which might also mean that newer X.org server versions are immune to this issue.
The reasons I’m running this X.org server release is because X.org from CentOS 6.8/6.9 is very buggy and starts consuming 100% of one CPU core shortly after logging into your X session.
This is not necessary at all, however once you’re running a Zcash miner (EWBF in my case) everything becomes so slow, you just cannot help but notice this bug better and sooner.
I would not recommend trying to reproduce this issue when running the miner since it requires too much patience.
Running EWBF is extremely easy. Generate a Bitcoin address or use an existing one and start mining this way:
miner --server equihash.usa.nicehash.com --port 3357 --user YOUR_BTC_ADDRESS --pass x
Nope, there isn’t one, sorry.
Nope, however EWBF surely runs using CUDA.
This is the trickiest part. You’re just moving your mouse and sometimes it stays where it shouldn’t be. There is no surefire way to reproduce this issue - it’s extremely random. Also I’ve just realized that it takes quite some (at least an hour) before you start seeing the problem. I’ve just started my X.org session and no matter how hard I try I cannot reproduce it.
If you’re switching from an application to an application, or scrolling or moving mouse it may happen as often as once every 5 minutes however if you don’t move your mouse or you’re working in a single application it may never occur.
I meant that running any compositing manager (based on OpenGL/EGL) is not required, but you’re right that it’s reproducible with the bare X.org server started as the only running application plus at least one other application is required, e.g. Mozilla Firefox.
If you don’t mind let’s make this way. I will try to find a way to reproduce it or certain steps that make it easier to reproduce and I’ll update this topic.
Meanwhile you could close this bug report as CANNOT REPRODUCE or anything like that because surely I don’t want to waste the precious time of your engineers.