4 ms·

> and all you need is an RTX 30- or 40-series GPU with at least 8GB of VRAM Smells like artificial restriction to me. I have a 2080 Ti with 8GB of VRAM that is

> and all you need is an RTX 30- or 40-series GPU with at least 8GB of VRAM Smells like artificial restriction to me. I have a 2080 Ti with 8GB of VRAM that is still perfectly fine for gaming. I play in 3440x1440 and the modern games need DLSS/FSR on quality for nice 60++ - 90 FPS. That is perfectly enough for me and I have not had a game, even UE5 games where I really thought I really NEED a new one. I bet that card is totally capable of running that chatbot. They do the same with frame generation. There they even require you a 40 series card. That is ridiculous to me as these cards are so fast that you do not even need any frame generation. The slower cards are the ones that would benefit from it most so they just lock it down artificially to boost their sales.

magicalhippo 3y ago
> 2080 Ti with 8GB of VRAM Sure you don't mean 11GB[1]? Or did they make other variants? FWIW I have a 2080 Ti with 11GB, been considering upgrading but thinking I'll wait til 5xxx. [1]: https://www.techpowerup.com/gpu-specs/geforce-rtx-2080-ti.c3305 https://www.techpowerup.com/gpu-specs/geforce-rtx-2080-ti.c3...
- redder23 3y ago
  Yes of course, I have a 11GB of VRAM. My next card will be an AMD one. I like that they are open sourcing most of their stuff and I think they play better with Linux Wine/Proton. FSR 3 also not artificially restricts cards and runs even on the competition. I read today about at open source API that takes CUDA calls and runs them on AMD or everywhere. I am sure there will be some cool open source projects that do all kinds of things if I ever even need them.