6700 XT has 4GB more VRAM that it may be very important in the long run. There are already games like The Last of Us that required a lot of VRAM even at 1080p. Also programming is a bit general. Are you a University Student right now in Computer Science? Do you use tools or will use tools like Tensorflow, Pytorch for AI (maybe in University Courses)? All these require CUDA and NVIDIA GPUs. Personally i never owned an AMD GPU and once that i tried to test an Unreal Engine Project to a computer with an AMD RX 580 GPU it was full of problems (maybe because of drivers issues). In my opinion for development i would go with NVIDIA. For just programming with tools like (Visual Studio , Android Studio etc) then the GPU does not matter but on the other hand CPU/RAM/SSD does.
I'm a university student, that's why I want to get 32gb RAM and a fast SSD. I problably won't need good gpu for programming.
My concern is not if you need a "good gpu for programming" but a compatible GPU for the tools you will need during your studies as someone who already finished BSc and MSc in Computer Science. Your CPU/SSD/RAM in the list should be enough for just programming. You said you want to play AAA titles at MAX settings and i guess you want to keep the GPU for a while so you have to go with the GPU with more VRAM. 100$ more may seems as a significant difference now but its better than to have to change the GPU in a year because the VRAM wont be enough to play the new AAA titles as you want at MAX settings.
I didn't say I want max settings (I said high or ultra). I can lower them a little bit if the 8GB VRAM wouldn't be enough to play on ultra, but I'm still not sure about the 7600. Maybe I should get the 6700XT.
I would still suggest the higher VRAM GPU. Its just a safer buy. There were games like TLOU and Jedi Survivor that required more than 8GB VRAM even at 1080p. Now there are many big releases in the near future like for example Starfield and AC Mirage that we will see how these new games will perform. For Deep Learning the tools are designed with CUDA so its recommended to get an NVIDIA GPU as these GPUs are also are performing better for these kind of tasks not to mention the extra features that these libraries support for NVIDIA GPUs like mixed precision (Tensor Cores). You may be able to run these tools also with AMD GPU with ROCm but just in Linux-like platforms. You can also train just on CPU but its really slow because GPUs have thousand cores for parallel computations. There are also other cloud solutions from Google that you can use for Deep Learning without using your own computer but its not free.
I have rx6600, 16gb ram and I must say that everything works perfectly in 1080, until I am satisfied with this achievement.
Facebook
Twitter
Instagram