Question 17
UnclassifiedWhat is the theoretical peak FP16 Tensor Core performance of an H100 SXM GPU?
Correct answer: C
Explanation
The H100 SXM’s Tensor Core peak is listed as 1,979 TFLOPS for FP16 because NVIDIA specifies this as the GPU’s theoretical maximum half-precision Tensor Core throughput. That figure reflects the chip’s advertised peak compute rate, not sustained real-world performance.
Why each option is right or wrong
A. 312 TFLOPS
B. 989 TFLOPS
C. 1,979 TFLOPS
NVIDIA’s H100 SXM product specifications list the GPU’s theoretical peak Tensor Core throughput for FP16 at 1,979 TFLOPS. The question asks for the peak FP16 Tensor Core figure, so the correct value is the advertised maximum compute rate for that exact precision and form factor, not a measured or sustained benchmark.
D. 3,958 TFLOPS