DEPLOYDatabase

Chip comparison

Google TPU v5e vs NVIDIA T4 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 v5eNVIDIA T4 Tensor Core GPU
Process nodeTSMC 12FFN
Transistors (B)13.6
Die size545 mm²
TDP170 W70 W
Memory16 GB HBM216 GB GDDR6
Memory bandwidth819 GB/s320 GB/s
FP16 (dense)65 TFLOPS
BF16 (dense)197 TFLOPS
INT8 (dense)393 TOPS130 TOPS
Form factorOAM (v5e pod, up to 256 chips)PCIe (single-slot, low-profile)
Announced2023-08-292018-09-13
Released2023-11-012019-01-01

Robots running Google TPU v5e

No robots publicly running it yet.

Robots running NVIDIA T4 Tensor Core GPU

No robots publicly running it yet.

Data centers with NVIDIA T4 Tensor Core GPU

No data centers publicly running it yet.

Common questions

Google TPU v5e vs NVIDIA T4 Tensor Core GPU: which is faster for training?

Google TPU v5e has 3.03x the dense FP16/BF16 throughput of the other (Google TPU v5e: 197 TFLOPS; NVIDIA T4 Tensor Core GPU: 65 TFLOPS). Real-world training speed also depends on memory bandwidth, interconnect, and software stack: see the spec diff for those inputs.

Google TPU v5e vs NVIDIA T4 Tensor Core GPU: which has more memory?

NVIDIA T4 Tensor Core GPU carries more HBM (Google TPU v5e: 16 GB HBM2; NVIDIA T4 Tensor Core GPU: 16 GB GDDR6). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.

Google TPU v5e vs NVIDIA T4 Tensor Core GPU: which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Google TPU v5e 1.16 (197 ÷ 170 W); NVIDIA T4 Tensor Core GPU 0.93 (65 ÷ 70 W). Google TPU v5e wins at the die level; system-level efficiency also depends on cooling and interconnect.

Google TPU v5e vs NVIDIA T4 Tensor Core GPU: which is more widely used?

Google TPU v5e: 0 robots and 1 data center publicly running it. NVIDIA T4 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 v5e full page · NVIDIA T4 Tensor Core GPU full page.