The new 1060 6GB of this generation just arrived!

The new 1060 6GB of this generation just arrived!

I now understand why amd being open source is a big deal. It's so nice just to plug your card in and have it fully functional.

Unigine benchmarks pictures coming next.

Attached: dual.jpg (1500x972, 1.15M)

Other urls found in this thread:

github.com/ROCm-Developer-Tools/HIP/blob/develop/docs/markdown/hip_faq.md
twitter.com/SFWRedditVideos

1060

Attached: gtx1060.png (2045x1277, 2.85M)

rx6600 DOUBLE the performance. This is the amdgpu/mesa stack included in oracle linux, not the amd provided one. That actually scored a little lower 14077.

Attached: RX6600.png (2048x1281, 2.86M)

This is what lmsensors picks up.

Attached: sensors.png (605x228, 18.93K)

>(CUSTOM)
Use standard settings to get meaningful numbers that can be compared to benchmarks posted online.

I was just using it to compare my two cards, but that's a good idea. Both of the cards would likely have given nearly stock values since they're both standard models.

also no problems with gpu passthrough

Why would you run an amd graphics card on linux? It just doesn't make sense at all. You don't play games on linux, so at best you only need an igpu. The only case you need a gpu card on linux is you run deep learning which requires nvidia cards because of cuda. If you're thinking about multi-gpu passthrough for virtualization then you shouldn't. Just dualboot or run 2 PCs.

Here's an official one, but I ran the others with nothing else open so a little lower is expected.
Plenty of games on steam.
AMD has HIP:Heterogenous-computing Interface for Portability which can convert cuda code, but I haven't looked into much.
github.com/ROCm-Developer-Tools/HIP/blob/develop/docs/markdown/hip_faq.md

Attached: Superposition_Benchmark_v1.1_14164_1652929925.png (2048x1280, 1.57M)

>I haven't looked into much
Because it's non-functional. Everyone who works in ML/DL fields knows amd ecosystem is a joke