DEPLOYDatabase

Chip comparison

Cambricon MLU370-X8 vs Tesla Dojo D1

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-X8Tesla Dojo D1
Process nodeTSMC N7TSMC N7
Transistors (B)50
Die size645 mm²
TDP250 W400 W
Memory48 GB LPDDR5
FP16 (dense)96 TFLOPS
BF16 (dense)362 TFLOPS
FP8 (dense)362 TFLOPS
INT8 (dense)256 TOPS
Form factorPCIeTraining tile (25 D1 chips per tile)
Announced2022-03-252021-08-19
Released2023-07-01

Robots running Cambricon MLU370-X8

No robots publicly running it yet.

Robots running Tesla Dojo D1

No robots publicly running it yet.

Data centers with Cambricon MLU370-X8

No data centers publicly running it yet.

Data centers with Tesla Dojo D1

No data centers publicly running it yet.

Common questions

Cambricon MLU370-X8 vs Tesla Dojo D1: which is faster for training?

Tesla Dojo D1 has 3.77x the dense FP16/BF16 throughput of the other (Cambricon MLU370-X8: 96 TFLOPS; Tesla Dojo D1: 362 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 Tesla Dojo D1: which is more power-efficient?

Dense FP16/BF16 TFLOPS per watt: Cambricon MLU370-X8 0.38 (96 ÷ 250 W); Tesla Dojo D1 0.91 (362 ÷ 400 W). Tesla Dojo D1 wins at the die level; system-level efficiency also depends on cooling and interconnect.

See also: every chip comparison · Cambricon MLU370-X8 full page · Tesla Dojo D1 full page.