Chip comparison
Google TPU v4 vs SambaNova SN40L
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.
| Field | Google TPU v4 | SambaNova SN40L |
|---|---|---|
| Process node | TSMC N7 | TSMC N5 |
| TDP | 192 W | 700 W |
| Memory | 32 GB HBM2 | 1500 GB DDR5 (three-tier: 64 MB SRAM + 64 GB HBM3 + 1.5 TB DDR5) |
| Memory bandwidth | 1,200 GB/s | — |
| FP16 (dense) | — | 638 TFLOPS |
| BF16 (dense) | 275 TFLOPS | — |
| FP8 (dense) | — | 1,250 TFLOPS |
| INT8 (dense) | 275 TOPS | — |
| Form factor | OAM (per-chip in v4 pod) | PCIe (SN40L DataScale system) |
| Announced | 2021-05-18 | 2023-09-19 |
| Released | 2022-05-01 | 2024-01-01 |
Robots running Google TPU v4
No robots publicly running it yet.
Robots running SambaNova SN40L
No robots publicly running it yet.
Data centers with Google TPU v4
Data centers with SambaNova SN40L
No data centers publicly running it yet.
Common questions
Google TPU v4 vs SambaNova SN40L: which is faster for training?
SambaNova SN40L has 4.55x the dense FP16/BF16 throughput of the other (Google TPU v4: 275 TFLOPS; SambaNova SN40L: 1,250 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 SambaNova SN40L: which has more memory?
SambaNova SN40L carries more HBM (Google TPU v4: 32 GB HBM2; SambaNova SN40L: 1500 GB DDR5 (three-tier: 64 MB SRAM + 64 GB HBM3 + 1.5 TB DDR5)). Memory capacity sets the largest model that fits on one chip without partitioning; memory bandwidth sets inference throughput.
Google TPU v4 vs SambaNova SN40L: which is more power-efficient?
Dense FP16/BF16 TFLOPS per watt: Google TPU v4 1.43 (275 ÷ 192 W); SambaNova SN40L 1.79 (1,250 ÷ 700 W). SambaNova SN40L wins at the die level; system-level efficiency also depends on cooling and interconnect.
Google TPU v4 vs SambaNova SN40L: which is more widely used?
Google TPU v4: 0 robots and 2 data centers publicly running it. SambaNova SN40L: 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 v4 full page · SambaNova SN40L full page.