GPT4All speeding up

[Peborgh]

I have played with GPT4All to be able to get information on what I have stored on my PCs.
It is as slow as a slow thing on a slow day that is professor of slowness at Oxford university (so to speak).
It does however seem to work in a quite useful way. If you can wait...
LLMs are meant to use beefy GPUs to operate efficiently. Which my PC(s) do not have.
So the question is this: Is there such a beefy(ish) GPU that I could add to my PC that GPT4All could detect and use in its functioning and thereby speed it up?
Thankyou,
peter

[Grogan]

Hello Peborgh. I'm not familiar with GPT4All, but what I do know is that you'd want an Nvidia graphics card, something like a 4090 card with 16 Gb of VRAM.

Your PC would have to be able to handle such a card, what type of PC do you have?

[Peborgh]

It's a Dell Optiplex, not really powerful or or recent!
Maybe I am over-optimistic...
Thanks,

peter

[Zema Bus]

I have GPT4All installed and have played with it some, I've been meaning to get into it more but just haven't gotten around to it. But it has been pretty responsive for me, the machine I'm using it on has an AMD Radeon RX 6750 XT. I'll see how it does on a machine I have with a lower end card.

[Peborgh]

Thank you,

peter

[Grogan]

The problem with that will be fitting a high end graphics card in the PC, and an inadequate power supply, likely lacking the connections for it. If you have a PCI-E x16 slot for a graphics card, even if it's wired for x8, you may be able to get a low powered card for it.

I said Nvidia simply because I thought it would be better recognized by the software but Zema has more knowledge of that. I see it should be able to use "Anything Direct3D 11/12 or OpenGL 2.1 capable" according to system requirements. So probably any graphics card you could install would help some.

[Zema Bus]

For something like ChatGPT that has held true, high end Nvidia cards have been optimal, though not essential. GPT4All is a little different in that it can run well on lower end hardware (at least to a degree), and it can run on just the CPU. Memory is a factor too, if there isn't very much available memory it probably won't run as well. Here's the FAQ, and here's a site that goes into more depth.

[Peborgh]

Thanks for all the helpful info, guys. I will close this conversation now. Turns out I was a bit too optimistic...
peter

[Grogan]

No problem, you're welcome, any time.