AMD FP8 blockwise GEMM
FP8-blockwise matrix multiply optimized for MI300: A (m×k) and B (n×k) in e4m3fnuz column-major with fp32 blockwise scales (128-wide k blocks), output C (m×n) in bf16 row-major. Shapes are drawn from DeepSeek-R1.
Reported evidence · last observed 2025-05-25. Reported by source; not independently reproduced. License unknown.
Current records
Workloadk = 512 · m = 1024 · n = 4096 · seed = 17512 · fp32/fp8_e4m3fnuz18 cases
Not measured on H100, B200 for this workload. Challenges →
Estimated floor 1.64 µs · record 20.15× above itestimate, not evidence ›
1submission 27020KernelBot user 333.1µs1.00×Reported · license unknown · source2025-05-25stale
Measured exactly what you asked. Reported by source; not independently reproduced.
2submission 27258KernelBot user 333.5µs1.01×Reported · license unknown · source2025-05-25stale
Measured exactly what you asked. Reported by source; not independently reproduced.
3submission 28436KernelBot user 333.8µs1.02×Reported · license unknown · source2025-05-27stale
Measured exactly what you asked. Reported by source; not independently reproduced.
4submission 29878KernelBot user 37634.9µs1.05×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
5submission 25056KernelBot user 334.9µs1.06×Reported · license unknown · source2025-05-19stale
Measured exactly what you asked. Reported by source; not independently reproduced.
6submission 29876KernelBot user 37635.0µs1.06×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
7submission 25269KernelBot user 37635.2µs1.07×Reported · license unknown · source2025-05-20stale
Measured exactly what you asked. Reported by source; not independently reproduced.
8submission 18889KernelBot user 37635.6µs1.07×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
9submission 30673KernelBot user 4435.7µs1.08×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
10submission 27935KernelBot user 335.8µs1.08×Reported · license unknown · source2025-05-26stale
Measured exactly what you asked. Reported by source; not independently reproduced.
11submission 30689KernelBot user 4435.8µs1.08×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
12submission 30771KernelBot user 4436.0µs1.09×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
13submission 9828KernelBot user 336.2µs1.10×Reported · license unknown · source2025-04-27stale
Measured exactly what you asked. Reported by source; not independently reproduced.
14submission 29949KernelBot user 4436.5µs1.10×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
15submission 25255KernelBot user 37636.6µs1.10×Reported · license unknown · source2025-05-20stale
Measured exactly what you asked. Reported by source; not independently reproduced.
16submission 18667KernelBot user 37636.6µs1.11×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
17submission 27162KernelBot user 16036.8µs1.11×Reported · license unknown · source2025-05-25stale
Measured exactly what you asked. Reported by source; not independently reproduced.
18submission 30688KernelBot user 4436.9µs1.12×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
19submission 34192KernelBot user 1637.4µs1.13×Reported · license unknown · source2025-08-22stale
Measured exactly what you asked. Reported by source; not independently reproduced.
20submission 19014KernelBot user 37637.4µs1.13×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
21submission 30110KernelBot user 4437.5µs1.13×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
22submission 31540KernelBot user 29238.4µs1.16×Reported · license unknown · source2025-06-06stale
Measured exactly what you asked. Reported by source; not independently reproduced.
23submission 27143KernelBot user 29239.2µs1.18×Reported · license unknown · source2025-05-25stale
Measured exactly what you asked. Reported by source; not independently reproduced.
24submission 29932KernelBot user 4439.4µs1.19×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
25submission 24658KernelBot user 29240.1µs1.21×Reported · license unknown · source2025-05-19stale
Measured exactly what you asked. Reported by source; not independently reproduced.
26submission 18833KernelBot user 16040.3µs1.22×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
27submission 9306KernelBot user 340.4µs1.22×Reported · license unknown · source2025-04-26stale
Measured exactly what you asked. Reported by source; not independently reproduced.
28submission 18884KernelBot user 16040.7µs1.23×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
29submission 21570KernelBot user 21240.9µs1.24×Reported · license unknown · source2025-05-13stale
Measured exactly what you asked. Reported by source; not independently reproduced.
30submission 30355KernelBot user 41541.1µs1.24×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
31submission 24311KernelBot user 29241.4µs1.25×Reported · license unknown · source2025-05-18stale
Measured exactly what you asked. Reported by source; not independently reproduced.
32submission 7548KernelBot user 37641.8µs1.26×Reported · license unknown · source2025-04-23stale
Measured exactly what you asked. Reported by source; not independently reproduced.
33submission 33220KernelBot user 34642.0µs1.27×Reported · license unknown · source2025-07-11stale
Measured exactly what you asked. Reported by source; not independently reproduced.
34submission 23187KernelBot user 742.9µs1.30×Reported · license unknown · source2025-05-16stale
Measured exactly what you asked. Reported by source; not independently reproduced.
35submission 6008KernelBot user 343.1µs1.30×Reported · license unknown · source2025-04-20stale
Measured exactly what you asked. Reported by source; not independently reproduced.
36submission 19003KernelBot user 37643.1µs1.30×Reported · license unknown · source2025-05-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
37submission 33210KernelBot user 34643.4µs1.31×Reported · license unknown · source2025-07-10stale
Measured exactly what you asked. Reported by source; not independently reproduced.
38submission 23139KernelBot user 42343.9µs1.33×Reported · license unknown · source2025-05-16stale
Measured exactly what you asked. Reported by source; not independently reproduced.
39submission 29705KernelBot user 8744.0µs1.33×Reported · license unknown · source2025-06-01stale
Measured exactly what you asked. Reported by source; not independently reproduced.
40submission 18428KernelBot user 29244.2µs1.33×Reported · license unknown · source2025-05-08stale
Measured exactly what you asked. Reported by source; not independently reproduced.
41submission 30576KernelBot user 41544.3µs1.34×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
42submission 30578KernelBot user 41544.5µs1.34×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
43submission 21778KernelBot user 33445.0µs1.36×Reported · license unknown · source2025-05-13stale
Measured exactly what you asked. Reported by source; not independently reproduced.
44submission 17835KernelBot user 29245.0µs1.36×Reported · license unknown · source2025-05-08stale
Measured exactly what you asked. Reported by source; not independently reproduced.
45submission 17326KernelBot user 29245.0µs1.36×Reported · license unknown · source2025-05-07stale
Measured exactly what you asked. Reported by source; not independently reproduced.
46submission 30582KernelBot user 745.0µs1.36×Reported · license unknown · source2025-06-02stale
Measured exactly what you asked. Reported by source; not independently reproduced.
47submission 32127KernelBot user 21345.2µs1.37×Reported · license unknown · source2025-06-15stale
Measured exactly what you asked. Reported by source; not independently reproduced.
48submission 17827KernelBot user 29246.0µs1.39×Reported · license unknown · source2025-05-08stale
Measured exactly what you asked. Reported by source; not independently reproduced.
49submission 18398KernelBot user 16046.3µs1.40×Reported · license unknown · source2025-05-08stale
Measured exactly what you asked. Reported by source; not independently reproduced.
50submission 22782KernelBot user 33446.9µs1.42×Reported · license unknown · source2025-05-15stale
Measured exactly what you asked. Reported by source; not independently reproduced.
51submission 13557KernelBot user 16047.7µs1.44×Reported · license unknown · source2025-04-30stale
Measured exactly what you asked. Reported by source; not independently reproduced.
52submission 10058KernelBot user 16048.1µs1.45×Reported · license unknown · source2025-04-27stale
Measured exactly what you asked. Reported by source; not independently reproduced.
53submission 33182KernelBot user 34648.6µs1.47×Reported · license unknown · source2025-07-09stale
Measured exactly what you asked. Reported by source; not independently reproduced.
54submission 5810KernelBot user 348.6µs1.47×Reported · license unknown · source2025-04-20stale
Measured exactly what you asked. Reported by source; not independently reproduced.
55submission 9617KernelBot user 16048.9µs1.48×Reported · license unknown · source2025-04-26stale
Measured exactly what you asked. Reported by source; not independently reproduced.
56submission 8734KernelBot user 16048.9µs1.48×Reported · license unknown · source2025-04-25stale
Measured exactly what you asked. Reported by source; not independently reproduced.
57submission 21793KernelBot user 33449.0µs1.48×Reported · license unknown · source2025-05-13stale
Measured exactly what you asked. Reported by source; not independently reproduced.
58submission 21355KernelBot user 33449.7µs1.50×Reported · license unknown · source2025-05-13stale
Measured exactly what you asked. Reported by source; not independently reproduced.
59submission 21230KernelBot user 749.9µs1.51×Reported · license unknown · source2025-05-12stale
Measured exactly what you asked. Reported by source; not independently reproduced.
60submission 5798KernelBot user 351.0µs1.54×Reported · license unknown · source2025-04-20stale
Measured exactly what you asked. Reported by source; not independently reproduced.
31 more rows in this comparison group. Open all in search →