Top GPUs & CPUs & AI accelerators Ranked by FP8 FLOPS of Matrix(Tensor) cores

This is NOT a ranking of gaming performance, it's a ranking for AI workloads.
This is NOT a ranking of processors—many of them may be missing since their specifications aren’t available. If you know them, feel free to comment.
This is NOT a ranking of actual task performances, as they may vary due to many factors like thermal throttling, driver support, and framework compatibility...

#1

GPU
192 GB
NVIDIA B100

3500 T

8 TB/s

2614.90 T

6 TB/s

2614.90 T

5.30 TB/s

#4

GPU
80 GB
NVIDIA H100 SXM

1979 T

3.36 TB/s

#5

GPU
141 GB
NVIDIA H200 SXM

1979 T

4.80 TB/s

1961.20 T

5.30 TB/s

#7

GPU
141 GB
NVIDIA H200 NVL

1930.20 T

4.80 TB/s

1678 T

3.70 TB/s

865 T

2.46 TB/s

419.01 T

1.79 TB/s

389.28 T

644 GB/s

#13

GPU
48 GB
NVIDIA L40

362.07 T

864 GB/s

330.30 T

1.01 TB/s

289.01 T

644 GB/s

225.10 T

960 GB/s

208.90 T

736 GB/s

194.95 T

716 GB/s

#19

GPU
24 GB
NVIDIA L2 PCIe

193 T

300 GB/s

176.39 T

672 GB/s

175.76 T

896 GB/s

160.36 T

504 GB/s

141.93 T

504 GB/s

131.91 T

576 GB/s

127.21 T

896 GB/s

123.47 T

672 GB/s

118.70 T

768 GB/s

116.58 T

504 GB/s

98.89 T

432 GB/s

98.30 T

896 GB/s

92.90 T

512 GB/s

88.26 T

288 GB/s

72.25 T

512 GB/s

62.48 T

256 GB/s

60.46 T

272 GB/s

46.45 T

256 GB/s

35.94 T

192 GB/s