DEPLOYDatabase

Chip comparison

Cambricon MLU370-X8 vs Google TPU v6 (Trillium)

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.

FieldCambricon MLU370-X8Google TPU v6 (Trillium)
Process nodeTSMC N7
TDP250 W200 W
Memory48 GB LPDDR532 GB HBM3
Memory bandwidth1,600 GB/s
FP16 (dense)96 TFLOPS
BF16 (dense)918 TFLOPS
FP8 (dense)1,836 TFLOPS
INT8 (dense)256 TOPS1,836 TOPS
Form factorPCIeOAM (Trillium pod, up to 256 chips)
Announced2022-03-252024-05-14
Released2024-12-01

Robots running Cambricon MLU370-X8

No robots publicly running it yet.

Robots running Google TPU v6 (Trillium)

No robots publicly running it yet.

Data centers with Cambricon MLU370-X8

No data centers publicly running it yet.

Data centers with Google TPU v6 (Trillium)

Common questions

Cambricon MLU370-X8 vs Google TPU v6 (Trillium): which is faster for training?

Google TPU v6 (Trillium) has 19.13x the dense FP16/BF16 throughput of the other (Cambricon MLU370-X8: 96 TFLOPS; Google TPU v6 (Trillium): 1,836 TFLOPS). Real-world training speed also depends on memory bandwidth, interconnect, and software stack: see the spec diff for those inputs.

Cambricon MLU370-X8 vs Google TPU v6 (Trillium): which has more memory?

Cambricon MLU370-X8 carries more HBM (Cambricon MLU370-X8: 48 GB LPDDR5; Google TPU v6 (Trillium): 32 GB HBM3). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.

Cambricon MLU370-X8 vs Google TPU v6 (Trillium): which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Cambricon MLU370-X8 0.38 (96 ÷ 250 W); Google TPU v6 (Trillium) 9.18 (1,836 ÷ 200 W). Google TPU v6 (Trillium) wins at the die level; system-level efficiency also depends on cooling and interconnect.

Cambricon MLU370-X8 vs Google TPU v6 (Trillium): which is more widely used?

Cambricon MLU370-X8: 0 robots and 0 data centers publicly running it. Google TPU v6 (Trillium): 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 · Cambricon MLU370-X8 full page · Google TPU v6 (Trillium) full page.