DEPLOYDatabase

Chip comparison

Google TPU v5p vs NVIDIA A100 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 v5pNVIDIA A100 Tensor Core GPU
Process nodeTSMC N7
Transistors (B)54
Die size826 mm²
TDP300 W400 W
Memory95 GB HBM2e80 GB HBM2e
Memory bandwidth2,765 GB/s2,039 GB/s
FP16 (dense)312 TFLOPS
BF16 (dense)459 TFLOPS312 TFLOPS
INT8 (dense)918 TOPS624 TOPS
Launch price (list)$10,000
Form factorOAM (per-chip in v5p pod)SXM4
Announced2023-12-062020-05-14
Released2024-01-012020-05-14

Robots running Google TPU v5p

No robots publicly running it yet.

Robots running NVIDIA A100 Tensor Core GPU

No robots publicly running it yet.

Data centers with Google TPU v5p

Data centers with NVIDIA A100 Tensor Core GPU

No data centers publicly running it yet.

Common questions

Google TPU v5p vs NVIDIA A100 Tensor Core GPU: which is faster for training?

Google TPU v5p has 1.47x the dense FP16/BF16 throughput of the other (Google TPU v5p: 459 TFLOPS; NVIDIA A100 Tensor Core GPU: 312 TFLOPS). Real-world training speed also depends on memory bandwidth, interconnect, and software stack: see the spec diff for those inputs.

Google TPU v5p vs NVIDIA A100 Tensor Core GPU: which has more memory?

Google TPU v5p carries more HBM (Google TPU v5p: 95 GB HBM2e; NVIDIA A100 Tensor Core GPU: 80 GB HBM2e). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.

Google TPU v5p vs NVIDIA A100 Tensor Core GPU: which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Google TPU v5p 1.53 (459 ÷ 300 W); NVIDIA A100 Tensor Core GPU 0.78 (312 ÷ 400 W). Google TPU v5p wins at the die level; system-level efficiency also depends on cooling and interconnect.

Google TPU v5p vs NVIDIA A100 Tensor Core GPU: which is more widely used?

Google TPU v5p: 0 robots and 1 data center publicly running it. NVIDIA A100 Tensor Core GPU: 0 robots and 0 data centers. Empty here means no operator has said publicly, not zero adoption in the market.

See also: every chip comparison · Google TPU v5p full page · NVIDIA A100 Tensor Core GPU full page.