Which GPUs does FlashAttention-2 support?
FlashAttention-2 2.8.3.post1: 22 of the 65 GPUs tracked here are supported, from the library's own requirements. Checked 2026-09-27.
The rule
- NVIDIA: builds compiled for compute capability 8.0, 9.0, 10.0, 12.0, which also cover newer minor versions of each (8.0 code runs on 8.6 and 8.9).
- AMD: gfx90a, gfx942, gfx950. Composable Kernel backend, ROCm 6.0+. RDNA 3/4 and Strix targets are allowed on the main branch but not in a release yet.
- Intel: not supported.
compiledFor is the default FLASH_ATTN_CUDA_ARCHS in setup.py at the v2.8.3.post1 tag (sm_100/sm_120 need a CUDA 12.8+ build). The README text names Ampere, Ada and Hopper; the build also covers Blackwell. Ada (8.9) runs the sm_80 code under CUDA's same-major compatibility. Turing (T4, RTX 20xx) is not supported here; the README points to the separate flash-attention-turing project. The AMD list is setup.py's allowed ROCm archs at the same tag. The main branch (2026-09-27) also allows gfx1100-1102, gfx1150/1151 and gfx1200/1201, which the README already advertises, but no release includes them yet.
Supported (22)
| GPU | Vendor | Target | Details |
|---|---|---|---|
| B300 (Blackwell Ultra) | NVIDIA | 10.3 | |
| DGX Spark | NVIDIA | 12.1 | |
| RTX 5060 Ti 16GB | NVIDIA | 12.0 | |
| RTX 5090 | NVIDIA | 12.0 | |
| RTX PRO 6000 Blackwell | NVIDIA | 12.0 | |
| B200 | NVIDIA | 10.0 | |
| H200 SXM 141GB | NVIDIA | 9.0 | |
| L40S | NVIDIA | 8.9 | |
| RTX 4060 Ti 16GB | NVIDIA | 8.9 | |
| H100 SXM5 80GB | NVIDIA | 9.0 | |
| RTX 4090 | NVIDIA | 8.9 | |
| RTX 6000 Ada Generation | NVIDIA | 8.9 | |
| RTX 3060 12GB | NVIDIA | 8.6 | |
| A100 SXM4 80GB | NVIDIA | 8.0 | |
| RTX 3090 | NVIDIA | 8.6 | |
| Instinct MI350X | AMD | gfx950 | |
| Instinct MI355X | AMD | gfx950 | |
| Instinct MI325X | AMD | gfx942 | |
| Instinct MI300A | AMD | gfx942 | |
| Instinct MI300X | AMD | gfx942 | |
| Instinct MI210 | AMD | gfx90a | |
| Instinct MI250X | AMD | gfx90a |
Not supported (42)
| GPU | Vendor | Target | Details |
|---|---|---|---|
| T4 | NVIDIA | 7.5 | no build for compute capability 7.5 |
| TITAN RTX | NVIDIA | 7.5 | no build for compute capability 7.5 |
| GTX 1080 Ti | NVIDIA | 6.1 | no build for compute capability 6.1 |
| TITAN V | NVIDIA | 7.0 | no build for compute capability 7.0 |
| V100 SXM2 32GB | NVIDIA | 7.0 | no build for compute capability 7.0 |
| P100 SXM2 16GB | NVIDIA | 6.0 | no build for compute capability 6.0 |
| Tesla P40 | NVIDIA | 6.1 | no build for compute capability 6.1 |
| TITAN X (Pascal) | NVIDIA | 6.1 | no build for compute capability 6.1 |
| M40 24GB | NVIDIA | 5.2 | no build for compute capability 5.2 |
| K80 | NVIDIA | 3.7 | no build for compute capability 3.7 |
| Radeon AI PRO R9700 | AMD | gfx1201 | |
| Radeon RX 9060 XT (16GB) | AMD | gfx1200 | |
| Radeon RX 9060 XT (8GB) | AMD | gfx1200 | |
| Radeon RX 9070 | AMD | gfx1201 | |
| Radeon RX 9070 XT | AMD | gfx1201 | |
| Ryzen AI Max+ 395 | AMD | gfx1151 | |
| Radeon RX 7600 XT | AMD | gfx1102 | |
| Radeon RX 7700 XT | AMD | gfx1101 | |
| Radeon RX 7800 XT | AMD | gfx1101 | |
| Radeon RX 7900 GRE | AMD | gfx1100 | |
| Radeon RX 6650 XT | AMD | gfx1032 | |
| Radeon RX 6750 XT | AMD | gfx1031 | |
| Radeon RX 6950 XT | AMD | gfx1030 | |
| Radeon RX 7900 XT | AMD | gfx1100 | |
| Radeon RX 7900 XTX | AMD | gfx1100 | |
| Radeon RX 6600 | AMD | gfx1032 | |
| Radeon RX 6700 XT | AMD | gfx1031 | |
| Instinct MI100 | AMD | gfx908 | |
| Radeon RX 6800 | AMD | gfx1030 | |
| Radeon RX 6800 XT | AMD | gfx1030 | |
| Radeon RX 6900 XT | AMD | gfx1030 | |
| Radeon VII | AMD | gfx906 | |
| Radeon Instinct MI50 | AMD | gfx906 | |
| Radeon Instinct MI60 | AMD | gfx906 | |
| Radeon Instinct MI25 | AMD | gfx900 | |
| Radeon Instinct MI6 | AMD | gfx803 | |
| Radeon Instinct MI8 | AMD | gfx803 | |
| Arc B580 | Intel | — | |
| Data Center GPU Max 1550 | Intel | — | |
| Arc A380 | Intel | — | |
| Arc A750 | Intel | — | |
| Arc A770 (16GB) | Intel | — |
Not enough data (1)
| GPU | Vendor | Target | Details |
|---|---|---|---|
| Instinct MI455X | AMD | — | gfx target not recorded |
Sources: github.com/Dao-AILab/flash-attention/blob/v2.8.3.post1/setup.py, github.com/Dao-AILab/flash-attention/blob/main/README.md