TPU v5e specs
Google · Shipping now · first available Aug 2023 (estimated). Dense compute, memory, bandwidth and power from the vendor's documents; each precision stays separate.
All accelerators · Compare with TPU v5p
Specifications
| Dense BF16 · FP32 accumulate | 197 TFLOPS |
|---|---|
| Dense FP8 · FP32 accumulate | Not published |
| Dense FP4 · FP32 accumulate | Not published |
| Memory | 16 GB · HBM |
| Memory bandwidth | 0.82 TB/s |
| Maximum chip power | Not published |
| Interconnect | ICI 2D torus |
| First availability | Aug 2023est — estimated / secondary-source figure |
| Availability | Shipping |
Efficiency-tier TPU; INT8 393 TOPS.
Dense tensor rates with FP32 accumulation, never sparse or FP16-accumulate headline figures. Memory is in decimal GB. The est marker flags a value its vendor-document review did not confirm.
Rent it
$1.20/chip-hrCheapest on-demand · Google Cloud · observed
1 provider: Google Cloud. Quotes checked through .
TPU v5e rental prices from every provider →Value ratios
| $ per dense BF16 PFLOP-hour | $6.09 |
|---|---|
| Dense BF16 TFLOPS per watt | No power figure |
| Bandwidth bytes per 1,000 BF16 FLOP | 4.16 |
| $ per dense FP8 PFLOP-hour | No dense FP8 rate |
| Dense FP8 TFLOPS per watt | No dense FP8 rate |
| Bandwidth bytes per 1,000 FP8 FLOP | No dense FP8 rate |
| $ per GB of memory per hour | $0.075 |
Prices divide the cheapest on-demand list price; theoretical ratios do not measure application speed.
Sources
Specifications reviewed . 197 TFLOPS BF16 and 393 INT8 TOPS per chip; no FP8, FP4 or power figure is published. 0.82 TB/s follows Google's Trillium comparison; the v5e table's 800 GiBps would read 0.86.
Rental price: Google Cloud published pricing ↗ (TPU v5e; per chip-hour).