# Does 5060 Ti 16GB work well with AMD GPUs like 7900XTX for LLM? — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/95681732c39665e7

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-13 09:59 ingested 2026-10-14 02:43

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1x4sk1y/does_5060_ti_16gb_work_well_with_amd_gpus_like/)

## retained text



```
Hi everyone,

Recently I got myself a RTX 5060 Ti 16GB and am building a PC for it. Since it only has 16GB VRAM, so I am considering getting a second GPU in the near future, maybe in 2027.

Since 5090 is getting really, really expensive and it's currently uncertain whether 5080 will deliver a 24GB version, despite recenet rumors. So I have been considering options such as AMD GPUs (I'm not saying Intel ones because I assume their software/hardware stack may be even worse than ROCm?), especially 7900XTX which is quite popular on this sub.

So I guess my question for people who have both Nvidia GPUs and AMD GPUs, **from your experience, do you think they work together well, for instance, can I use CUDA with the former one and ROCm with the latter one "at the same time"?** Currently, I think my usage is mostly fine-tuning, inference, and maybe learning kernel work?

Thank you a lot for your time!
```



document fetched 2026-10-14 00:42 · canonical https://reddit.com/r/LocalLLaMA/comments/1x4sk1y/does_5060_ti_16gb_work_well_with_amd_gpus_like
