Hope everyone is doing good in life, long story short i just setup a pc with ryzen 9 7950x3d and an AMD RX OC 9070xt but when i run cursor the pc seems to lag like its running pantium 1 or something, and ive heard compared to the cpu the gpu is really fast and good for LLMS. so i was wondering is there a way to run cursor on GPU as i am running Ubuntu as the OS. because the cursor running currently make the OS very slow somehow while not utilizing the full cores not even the RAM. so if anyone can provide a solution i shall be thankful.
1 Like
Are you running a local model, if not the GPUs are useless, but a 7950x3d should be a very good CPU. How much RAM do you have, they say you need only like 4GB, but if you want the app to run fast it’s best to have 32GB
i have 32GB ram. i have downloaded the app image for the cursor and running that as --no-sandbox. and the model i run only is sonnet 4 thinking.
I hope you didn’t buy a GPU specifically for cursor, because you have made a big mistake: sonnet doesn’t run on your computer, it runs in the cloud
I’m still not sure why cursor is lagging though
ofcourse i havent bought it for cursor i have bought the PC for normal productivity but because the LLMs etc are resource hungry so i wanted have good specs on hand but turns out the cursor is very laggy i dont know why although even my internet is good too.
Hey, it’s possible that some process is excessively consuming your CPU/RAM resources. Could you check which processes are being used while Cursor is running?
Start by checking the built-in Activity Monitor, then use Process Explorer in Cursor itself. There’s also a chance the issue might be related to one or more extensions, so try launching with the flag cursor --disable-extensions. The issue might also be related to the specifics of the project, have you tried checking this in other projects?
It would be great if you could share the info and screenshots of the processes used.
That’s a lot of RAM being used. Next router is known to be very resource intensive: https://stackoverflow.com/questions/77173892/nextjs-next-router-worker-consuming-more-memory-space-m1-macbook-gets-lock-auto. I don’t think it’s really the LLMs themselves that are requiring your GPU. All built-in LLMs in cursor run in the cloud and don’t use your GPU
you might be right but the issue is in cursor not in the Next Router and the way i confirmed it is running the same project using windsurf and it was working perfectly fine and also using VS-code with cline and its working fine but when i run the project with cursor it lags like hell.
any cursor member can shed some light on this issue because at this point the cursor IDE is just unable to compile the code for a proper 2,3 seconds it hangs even if i try to edit a css code what is happening.
In Cursor, does it start fast and then slow down or is it slow from the moment you open it. The former could suggest a memory leak.
A couple of things:
- If you’re on the new payment model, you sure you’re not hitting the throttle limits? Cursor slows to a crawl when that happens. It’s a massive pain
- I’m on the new payment model and running sonnet-4 in agent max mode to try and avoid the throttling (just testing this out) and it’s still painfully slow
- I have an old rig AMD 5600 and RTX 6700 BUT I have 64 GB RAM. I run 5 virtual desktops, 3 instances of Cursor each with a different project, about a million tabs open across multiple browser profiles and my rig doesn’t skip a beat. My RAM sits at about 50% capacity. Maybe throw some extra RAM in your computer.
throwing a ram doesnt have anything to do with this because as i said the same project runs buttery smooth on windsurf and vscode and my PC i have never seens utilizing my ram to its full capacity, i am very sure the issue is with cursor but the only issue i am not 100% sure is that where in cursor and why its happening, so cursor starts fine but as soon as i run the project its starts to behave crazy slow and at this point not even working for me i have to wait for even typing my code.
1 Like
For me, Cursor lags very badly when I have a long chat history and clearing it makes it stop lagging.
Indeed I also had missing hardware acceleration issues on the official AppImage. I will never understand why Cursor does it like that but here we are. Here’s a hack: use code-cursor package from Nix which repacks AppImage into sane distribution and hooks up all necessary stuff to get hardware acceleration to work. Thank me later.
never used or heard about the Nixshell just wondering what is it before i proceed because it says right there
that its just for trying something before permanently installing it so how is it like running cursor on your machine?
It says permanently because nix shell is often used to install software temporarily, for example you can do: nix-shell -p code-cursor --run cursor and when that commands exits it’s like nothing happened (it will run cursor without installing it). You can of course uninstall permanently installed packages later: nix-env --uninstall code-cursor 
(nix-env is for permanent installs, nix-shell for temporary).
If you ask AI there’s a high chance it will be able to help you with Nix stuff. It requires more work than AppImage but welp, that’s the best I know and it’s fairly universal.
In the perfect world, Cursor would simply fix their Linux distribution.