DEPLOYDatabase

Chip comparison

Google TPU v4 vs NVIDIA H200 Tensor Core GPU

Side-by-side specs

Straight from each vendor's datasheet. Winner in bold: higher throughput, memory, or transistor count; lower power draw or price. Sparse (2:4) throughput roughly doubles the FP8/INT8 numbers shown.

FieldGoogle TPU v4NVIDIA H200 Tensor Core GPU
Process nodeTSMC N7TSMC 4N
Transistors (B)80
Die size814 mm²
TDP192 W700 W
Memory32 GB HBM2141 GB HBM3e
Memory bandwidth1,200 GB/s4,800 GB/s
FP16 (dense)989 TFLOPS
BF16 (dense)275 TFLOPS989 TFLOPS
FP8 (dense)1,979 TFLOPS
INT8 (dense)275 TOPS1,979 TOPS
Form factorOAM (per-chip in v4 pod)SXM5
Announced2021-05-182023-11-13
Released2022-05-012024-03-01

Robots running Google TPU v4

No robots publicly running it yet.

Robots running NVIDIA H200 Tensor Core GPU

No robots publicly running it yet.

Common questions

Google TPU v4 vs NVIDIA H200 Tensor Core GPU: which is faster for training?

NVIDIA H200 Tensor Core GPU has 7.20x the dense FP16/BF16 throughput of the other (Google TPU v4: 275 TFLOPS; NVIDIA H200 Tensor Core GPU: 1,979 TFLOPS). Real-world training speed also depends on memory bandwidth, interconnect, and software stack: see the spec diff for those inputs.

Google TPU v4 vs NVIDIA H200 Tensor Core GPU: which has more memory?

NVIDIA H200 Tensor Core GPU carries more HBM (Google TPU v4: 32 GB HBM2; NVIDIA H200 Tensor Core GPU: 141 GB HBM3e). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.

Google TPU v4 vs NVIDIA H200 Tensor Core GPU: which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Google TPU v4 1.43 (275 ÷ 192 W); NVIDIA H200 Tensor Core GPU 2.83 (1,979 ÷ 700 W). NVIDIA H200 Tensor Core GPU wins at the die level; system-level efficiency also depends on cooling and interconnect.

Google TPU v4 vs NVIDIA H200 Tensor Core GPU: which is more widely used?

Google TPU v4: 0 robots and 2 data centers publicly running it. NVIDIA H200 Tensor Core GPU: 0 robots and 4 data centers. Empty here means no operator has said publicly, not zero adoption in the market.

See also: every chip comparison · Google TPU v4 full page · NVIDIA H200 Tensor Core GPU full page.