FP32 FFMA (CUDA cores)

Scalar FP32 fused multiply-add on the CUDA cores, which is the path the unlock never governed.

数据怎么来的

Register-resident fma.rn.f32 issue rate: eight independent accumulator chains per warp, best of three timed runs. Library SGEMM rows are grouped with the other cuBLAS GEMMs instead, because a fixed M.N.K call and an issue loop are not the same measurement.

是否受 SM speed-select 熔丝约束：否

实测行

共 8 行，覆盖 8 张卡。0 行提升超过 5%，8 行没有变化；提升最大的是 GeForce RTX 3050 6GB，为 1.01×。

模式	解锁前	解锁后	提升	显卡
`ffma` · FP32 was never rate-limited	8.40	8.50 TF/s	1.01×	GeForce RTX 3050 6GB · GA107 · 18 SM
`ffma` · FP32 was never rate-limited	49.4	49.7 TF/s	1.01×	GeForce RTX 4080 SUPER · AD103 · 80 SM
`ffma` · FP32 was never rate-limited	36.6	36.7 TF/s	1.00×	GeForce RTX 4070 SUPER · AD104 · 56 SM
`ffma` · FP32 was never rate-limited	40.1	40.2 TF/s	1.00×	GeForce RTX 4070 Ti · AD104 · 60 SM
`ffma` · FP32 was never rate-limited	79.1	79.1 TF/s	1.00×	GeForce RTX 4090 · AD102 · 128 SM
`ffma` · FP32 was never rate-limited	31.7	31.7 TF/s	1.00×	GeForce RTX 4070 (AD103) · AD103 · 46 SM
`ffma` · FP32 was never rate-limited	32.1	32.1 TF/s	1.00×	GeForce RTX 4070 (Gigabyte) · AD104 · 46 SM
`ffma` · FP32 was never rate-limited	50.8	50.7 TF/s	1.00×	GeForce RTX 4080 · AD103 · 76 SM

这组里最小的比值是 GeForce RTX 4080，为 1.00×。

1.00× 该怎么读

1.00× 不代表解锁失败。它说明那条路径本来就没有被限速，覆盖没有可恢复的东西。如果你在自己的卡上测出你关心的模式是 1.00×，解锁很可能确实生效了，只是对这个负载没有帮助。

FP32 FFMA (CUDA cores)

实测行

1.00× 该怎么读

其他测量路径