Has anyone tried it? Is it that big of a difference compared to the 5090? Curious to try it out. Vera pricing seems much better than Runpod’s
Honestly? After months of wrestling with my 5090’s 32GB limitations, I’m ready to move on.
The constant memory juggling is exhausting - you can’t just run a 70B model, you have to quantize it aggressively, pray your batch size fits, and forget about using the full context window. Every project becomes a memory optimization puzzle before you even start the actual work.