L4 specs
NVIDIA · Shipping now · first available Mar 2023. Dense compute, memory, bandwidth and power from the vendor's documents; each precision stays separate.
All accelerators · Compare with L40S
Specifications
| Dense BF16 · FP32 accumulate | 121 TFLOPS |
|---|---|
| Dense FP8 · FP32 accumulate | 242 TFLOPS |
| Dense FP4 · FP32 accumulate | Not published |
| Memory | 24 GB · GDDR |
| Memory bandwidth | 0.3 TB/s |
| Maximum chip power | 72 W |
| Interconnect | PCIe 4.0 |
| Process node | TSMC 4N |
| First availability | Mar 2023 |
| Availability | Shipping |
Low-profile 72 W Ada inference card with 24 GB of GDDR6 and no NVLink.
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
No tracked rental quote. The site tracks providers' published list prices; this part may still be rentable under terms it does not record.
Value ratios
| $ per dense BF16 PFLOP-hour | No on-demand quote |
|---|---|
| Dense BF16 TFLOPS per watt | 1.68 |
| Bandwidth bytes per 1,000 BF16 FLOP | 2.48 |
| $ per dense FP8 PFLOP-hour | No on-demand quote |
| Dense FP8 TFLOPS per watt | 3.36 |
| Bandwidth bytes per 1,000 FP8 FLOP | 1.24 |
| $ per GB of memory per hour | No on-demand quote |
Prices divide the cheapest on-demand list price; theoretical ratios do not measure application speed.
Sources
Specifications reviewed . Dense BF16 121 and FP8 242 TFLOPS, the unstarred halves of the sparse figures; 24 GB GDDR6 with ECC at 300 GB/s, 72 W, PCIe Gen4 x16; TSMC 4N. No FP4 format is listed (INT4 only).
First availability: NVIDIA GTC 2023 inference platforms launch ↗ (2023-03-21): “The NVIDIA L4 GPU is available in private preview on Google Cloud Platform and also available from a global network of more than 30 computer makers”