NVIDIA L40: an Ada Lovelace data center accelerator for AI and visual computing
The L40 covers several workloads at once in a server rack - neural network training and inference, production rendering, NVIDIA Omniverse Enterprise and RTX vWS virtual workstations. The Ada Lovelace architecture provides 18,176 CUDA cores, 568 4th-generation Tensor cores with support for FP8, BF16, TF32 and structural sparsity, as well as 142 3rd-generation RT cores for hardware ray tracing with accelerated motion blur.
48 GB of GDDR6 memory with ECC and 864 GB/s of bandwidth (384-bit bus) are designed for memory-intensive scenes, large datasets and simulations where the capacity of a regular graphics card is not enough. Three NVENC engines and three NVDEC engines with hardware AV1, H.265 and H.264 encoding offload the CPU during streaming and video post-production.
- GPU: NVIDIA Ada Lovelace, 18,176 CUDA cores
- Memory: 48 GB GDDR6 ECC, 384-bit bus, 864 GB/s
- Tensor: 568 4th-generation cores; RT: 142 3rd-generation cores
- Interface: PCIe Gen4 x16, 16-pin power (PCIe CEM5), TGP 300 W
- Form factor: dual-slot, passive cooling, 4 x DisplayPort 1.4a
- vGPU (vApps, vPC, RTX vWS), Secure Boot, NEBS Level 3
The passive dual-slot design with a TGP of 300 W is intended for continuous 24x7 operation in servers and NVIDIA-Certified Systems. This is the retail version in PNY branded packaging with a full manufacturer warranty. We will select a platform to match the number of slots and the power - call +380 93,594 00 77.
Not sure this card fits your system?
We will check compatibility with your platform, calculate the power draw for your configuration and tell you whether the memory is enough for your models.
The model you need is out of stock? ETE is listed among official PNY partners, so we bring professional NVIDIA cards to order with the manufacturer warranty.
Message us on Telegram or leave a request.
