SambaNova SN40L vs SambaNova SN50
The SN50 delivers 2.5x the BF16 throughput of the SN40L (1,600 vs 638 TFLOPS dense).
SN50: RDU, 2026
BF16 dense lead
2.5x
SN50: 1,600 vs 638 TFLOPS
Memory
64 vs 64 GB
Same capacity
SambaNova SN40L
Full specs →Architecture
RDU
Memory
64 GB
SambaNova SN50
Full specs →Architecture
RDU
Year
2026
Memory
64 GB
Specifications
Architecture
RDURDU
Launch Year
—2026
Form Factor
—Custom
Memory
64 GB64 GB
Memory Bandwidth
1.8 TB/s—
TDP
——
Process Node
5nm5nm
Spec SN40L SN50
Architecture RDURDU
Launch Year —2026
Form Factor —Custom
Memory 64 GB64 GB
Memory Bandwidth 1.8 TB/s—
TDP ——
Process Node 5nm5nm
Performance (TFLOPS)
FP64
No verified data available No verified data available
FP32
No verified data available No verified data available
TF32
No verified data available No verified data available
BF16
638 TFLOPS 1,600 TFLOPS
FP16
No verified data available No verified data available
FP8
No verified data available3,200 TFLOPS
FP6
No verified data available No verified data available
FP4
No verified data available No verified data available
INT8
No verified data available No verified data available
Precision SN40L SN50
FP64 No verified data available No verified data available
FP32 No verified data available No verified data available
TF32 No verified data available No verified data available
BF16 638 TFLOPS 1,600 TFLOPS
FP16 No verified data available No verified data available
FP8 No verified data available3,200 TFLOPS
FP6 No verified data available No verified data available
FP4 No verified data available No verified data available
INT8 No verified data available No verified data available
FLOPS by Precision
What actually differs
Memory is 64 GB against 64 GB. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.
For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.
Frequently asked questions
Is the SN40L faster than the SN50?
At BF16 precision the SN50 reaches 1,600 TFLOPS dense against 638 TFLOPS for the SN40L. The performance table on this page lists every published precision for both GPUs.
Which has more memory, the SN40L or the SN50?
Both GPUs carry 64 GB of memory.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.