DEPLOYDatabase

Chip comparison

Google TPU v4 vs Google TPU v5e

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 v4Google TPU v5e
Process nodeTSMC N7
TDP192 W170 W
Memory32 GB HBM216 GB HBM2
Memory bandwidth1,200 GB/s819 GB/s
BF16 (dense)275 TFLOPS197 TFLOPS
INT8 (dense)275 TOPS393 TOPS
Form factorOAM (per-chip in v4 pod)OAM (v5e pod, up to 256 chips)
Announced2021-05-182023-08-29
Released2022-05-012023-11-01

Robots running Google TPU v4

No robots publicly running it yet.

Robots running Google TPU v5e

No robots publicly running it yet.

Common questions

Google TPU v4 vs Google TPU v5e: which is faster for training?

Google TPU v4 has 1.40x the dense FP16/BF16 throughput of the other (Google TPU v4: 275 TFLOPS; Google TPU v5e: 197 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 Google TPU v5e: which has more memory?

Google TPU v4 carries more HBM (Google TPU v4: 32 GB HBM2; Google TPU v5e: 16 GB HBM2). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.

Google TPU v4 vs Google TPU v5e: which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Google TPU v4 1.43 (275 ÷ 192 W); Google TPU v5e 1.16 (197 ÷ 170 W). Google TPU v4 wins at the die level; system-level efficiency also depends on cooling and interconnect.

Google TPU v4 vs Google TPU v5e: which is more widely used?

Google TPU v4: 0 robots and 2 data centers publicly running it. Google TPU v5e: 0 robots and 1 data center. Empty here means no operator has said publicly, not zero adoption in the market.

See also: every chip comparison · Google TPU v4 full page · Google TPU v5e full page.