I recently upgraded to a 7900 XTX GPU. Besides being great for gaming, I wanted to try it out for some machine learning.
It’s well known that NVIDIA is the clear leader in AI hardware currently. Most ML frameworks have NVIDIA support via CUDA as their primary (or only) option for acceleration. OpenCL has not been up to the same level in either support or performance.
That being said, the 7900 XTX is a very powerful card. It has 24GB of VRAM, a theoretical 60 TFLOPS of f32, and 120 TFLOPS of f16. The recent AI hype wave is also incentivizing AMD to beef ML support on their cards, and they seem to be making real investments in that space.

Casey Primozic

A hands-on benchmarking report of the AMD Radeon RX 7900 XTX GPU for machine learning workloads in mid-2023. The author tests TensorFlow with ROCm for neural network training and ResNet50, finding performance well below theoretical expectations — barely beating the older 6900 XT despite having 2.6x more FP32 TFLOPS. A raw RDNA3 assembly benchmark via tinygrad confirms the card can hit its advertised 61 TFLOPS peak, pointing to software/driver issues rather than hardware limitations. Porting the benchmark to tinygrad yielded mixed results — occasionally beating TensorFlow by ~10% but highly unstable across versions. The conclusion is that AMD GPU ML support is immature but improving, with official RDNA3 ROCm support planned for late 2023.

Machine Learning Benchmarks on the 7900 XTX

Simple TensorFlow Neural Network Training Benchmark ⌗

Porting the TensorFlow benchmark to TinyGrad ⌗