AMD ROCm GPU Runs FLUX.2-dev | Generated by AI

Home 2026.07

Here’s the full picture:

✅ GPU Confirmed — AMD ROCm

Component Details
GPU gfx1100 (RDNA 3 arch) — likely AMD Radeon PRO W7900 or cloud-partitioned Instinct
VRAM 48 GB (Samsung)
ROCm v7.2.1 (HIP 7.2.53211)
Driver v6.16.13
PCIe Gen4 x16 @ 16GT/s
Power 241W cap, currently 17W
OS Ubuntu 24.04, Python 3.12.3
CPU AMD EPYC 9334 — 128 threads
RAM 503 GB

📦 PyTorch (ROCm 7.2) Available

Official PyTorch wheels with ROCm 7.2 support exist on PyTorch’s index:

pip install torch --index-url https://download.pytorch.org/whl/rocm7.2

Available versions: 2.13.0+rocm7.2 (latest), 2.12.1, 2.12.0, 2.11.0

🎨 Can FLUX.2-dev Run on This?

Yes, absolutely. Here’s the breakdown:

FLUX.2-dev VRAM Requirements

Component Est. Size Notes
Text Encoder (T5-XXL) ~10 GB Loaded once, usually fp16
Transformer (flux2-dev) ~12-16 GB bf16 weights
VAE ~1 GB Tiny
Total to run ~25-30 GB At default 1024×1024
Inference overhead ~2-4 GB Intermediate activations

You Have 48 GB → ✅ More than enough

Headroom for higher resolutions (e.g. 1536×1536) or larger batch sizes.

Expected Performance

Hardware Time per image at 1024×1024 (4-step schnell / 28-step dev)
Your GPU (RDNA 3, 48 GB) ~3-10 sec/image (estimate, similar to RX 7900 XTX)
NVIDIA A100 (80 GB) ~2-5 sec
CPU-only (128 cores) Several minutes

⚠️ One Catch

FLUX.2-dev is a gated model — you need a HuggingFace token (must accept the license on huggingface.co first) to download it.

Want me to install PyTorch + diffusers on the server and test FLUX inference? I’d need to:

  1. pip install torch --index-url https://download.pytorch.org/whl/rocm7.2
  2. pip install diffusers transformers accelerate sentencepiece
  3. Log in with your HF token and download FLUX.2-dev

Generated by AI. Curating and sharing still takes effort. If you find it useful, feel free to donate. WeChat: @lzwjavaWeChat QR · X: @lzwjava · Say hi 👋

Back Donate