麻烦自查一下

随便举个例子，懒得验证更多了
同一代的A6000 48G和3090 24G，价格前面的一直2w5以上，后者大概9000上下。

The NVIDIA RTX A6000 has a higher theoretical performance than the NVIDIA RTX 3090 in FP16 and FP32,
Theoretical performance
FP16 (half): The A6000 has 38.71 TFLOPS, while the 3090 has 35.58 TFLOPS
FP32 (float): The A6000 has 38.71 TFLOPS, while the 3090 has 35.58 TFLOPS
FP6.4 (double): The A6000 has 1210 GFLOPS, while the 3090 has 556 GFLOPS

【 在 huy2 的大作中提到: 】
: 但消费级的浮点运算不行吧，整数运算可以
:
--
FROM 119.57.72.*