nVidia L4 (7680CudaCores/240TensorCores,24GB GDDR6 ECC, PCI-E16g4LP, 1slot, 72W), pasivní chlazení

FP32                        30.3 TFLOPS
TF32 Tensor Core    120 TFLOPS | Sparsity
FP16 Tensor Core    242 TFLOPS | Sparsity
BFLOAT16 Tensor Core       242 TFLOPS | Sparsity
FP8Tensor Core       485 TFLOPS | Sparsity
INT8 Tensor Core    485 TOPS | Sparsity
GPU Memory           24GB
GPU Memory Bandwidth   300 GB/s
Display Connectors   None | vGPU Only
NVENC | NVDEC     2 | 4 | 4| JPEG Decoders | AV1 Encode and Decode
Form Factor             Single Slot | Low-Profile
System Interconnect    PCIe Gen4 x16 | 64GB/s
Thermal Solution       Passive
Maximum Power Consumption    72 W
Server Options      Partner and NVIDIA-Certified Systems with 1-8 GPUs

Render Config
Shading Units   7680
TMUs    240
ROPs   80
SM Count   60
Tensor Cores   240
RT Cores    60
L1 Cache    128 KB (per SM)
L2 Cache    48 MB

Product website:

https://www.nvidia.com/en-us/data-center/l4/