Two players on one PC with a 2-GPU setup is doable.
Two players on one PC with a 2-GPU setup is doable.
the problem is straightforward: building software for this requires funding, the company needs fast CPU virtualization and is ready to invest heavily for it. no one would willingly pay such amounts for efficient GPU virtualization, since today's VM providers focus mainly on servers where GPU power isn't essential. if they require it, they simply use dedicated GPU clusters.
I use my main system for rendering and recording games, so buying two budget models isn't an option. Still, I just wanted to check if playing CS:GO like that works well. Can I split it up with my brother to play the cheapest way? That sounds straightforward, right? I'm hoping for a quick answer.
The issue lies in how GPUs are treated by the system—though they contain execution units grouped into cores, they're still perceived as a single unit. On the other hand, each processor core is seen independently. I'm uncertain whether drivers or hardware present the GPU as one cohesive part or as separate components. It seems possible to run multiple users on two GPUs, but this demands driver intervention to manage command submission and VRAM allocation.
I don't agree with the chosen response. Based on what I've observed, it's quite feasible to run several systems on one GPU today. This technique is known as GPU-partitioning in Windows. A specific driver configuration can help make it work, though you might not need a dedicated setup. I configured Aster and successfully ran multiple users on a single GPU, even without custom drivers or partitioning. I shared a guide on this process: while my configuration uses two GPUs, the steps remain identical for one GPU. It seems Windows can manage sharing GPU resources effectively. I was genuinely impressed by its performance.