Skip to main content
Loading...

Advantech SKY-TESL-L40S-48P NVIDIA L40S 48GB GDDR6 Passive Data Centre GPU for AI and Graphics with 4x DisplayPort

The SKY-TESL-L40S-48P is an NVIDIA L40S data centre GPU on the Ada Lovelace architecture, with 48 GB of GDDR6 ECC memory, 91.6 TFLOPS FP32, RT and Tensor Cores and 4x DisplayPort 1.4, in a passive 350 W dual-slot card. It suits servers that mix AI, video, graphics and virtual desktops on one GPU.

  • Universal Compute and Graphics GPU: 18,176 CUDA cores, 568 fourth-generation Tensor Cores and 142 third-generation RT Cores, rated at 91.6 TFLOPS FP32.
  • 48 GB GDDR6 with ECC: 864 GB/s of memory bandwidth with error correction, sized for AI models, large scenes and several virtual workstation users.
  • AV1 Encode and Decode: Three NVENC and three NVDEC engines with AV1 support handle video transcoding and analytics streams on the same card.
  • NVIDIA vGPU Support: Supports NVIDIA virtual GPU software, including RTX Virtual Workstation and Virtual PC, to share one card across remote users.
  • Passive Dual-Slot, 350 W: A full-height PCIe 4.0 x16 card drawing up to 350 W through a 16-pin PCIe connector, cooled by server airflow.
Form Factor: PCI ExpressPCI Express
Multi Display Support: Quad Display SupportQuad Display Support
UK Support Team: +44 (0)1785 558 300 Call
Select section
  • Overview
  • Specifications
  • Downloads

SKY-TESL-L40S-48P Overview

Many server estates run AI inference, video and remote graphics side by side, and a GPU that handles all three can simplify the build. The SKY-TESL-L40S-48P is the NVIDIA L40S from Advantech, a universal compute and graphics data centre GPU on the Ada Lovelace architecture. It is a passive, dual-slot PCIe card that suits video, AI, virtualised desktop and graphics applications in the enterprise, in the cloud and at the edge.

The GPU has 18,176 CUDA cores, 568 fourth-generation Tensor Cores and 142 third-generation RT Cores, rated at 91.6 TFLOPS FP32, and it supports DLSS 3. Memory is 48 GB of GDDR6 with ECC at 864 GB/s. With NVIDIA vGPU software, the SKY-TESL-L40S-48P can serve mid-to-high-end design workflows on NVIDIA RTX Virtual Workstation and productivity applications on NVIDIA Virtual PC. Multi-Instance GPU and NVLink are not supported.

The card connects over PCI Express 4.0 x16 and draws up to 350 W through a 16-pin PCIe connector. Three NVENC and three NVDEC engines include AV1 encode and decode for video streams, and four DisplayPort 1.4 outputs are there when a local display is needed. It is a dual-slot, full-height card with a passive heatsink, so it relies on the server’s own airflow.

We supply the SKY-TESL-L40S-48P tested, and fit, configure and burn-in test it in the GPU servers we build, with power and airflow checked for a 350 W passive card. We can also advise on how many cards a chassis can feed and cool. The SKY-TESL-L40-48P uses the same 48 GB of GDDR6 at 300 W, the low-profile SKY-TESL-L4-24P draws 72 W for inference-led builds, and the SKY-TESL-6000BS-96 is the Blackwell Server Edition with 96 GB of GDDR7.

SKY-TESL-L40S-48P Specifications

Overview
Brand & ModelAdvantech SKY-TESL-L40S-48P
Performance
GPU / AI AcceleratorDiscrete GPU (PCIe)
Supported GPUsNVIDIA L40S: 18176 CUDA Cores, 568 Tensor Cores, 48GB GDDR6 ECC, 350W
AI Performance1,466 TOPS
GPU FamilyNVIDIA Data Center
GPU ArchitectureNVIDIA Ada Lovelace
PCIe Gen SupportPCIe 4.0
Graphics
Display StandardsDisplay Port 1.4a
Multi Display SupportQuad Display Support
Display Outputs4x Display Ports
Mechanical Properties
Dimensions267mm x 112mm (10.50" x 4.40")

Official Advantech UK Distributor

Things Embedded is an official Advantech UK distributor, and we supply the SKY-TESL-L40S-48P tested, and fit, configure and burn-in test it in the GPU computers and servers we build. Within the Advantech range, the SKY-TESL-L40-48P has the same 48 GB of GDDR6 and 4x DisplayPort at a lower 300 W maximum draw.

Across our range, the SKY-TESL-L40S-48P sits with our server GPUs, edge AI servers and computing components.

Consult. Design. Integrate.

Things Embedded is a leading UK supplier of industrial computing and edge AI systems. We combine technical expertise with a collaborative approach to deliver optimal solutions for your application. Whether it’s an off-the-shelf platform or a custom design, we stay at the forefront of the latest technologies to ensure your success at the edge. Our team works closely with you from concept through deployment, providing guidance, engineering support, and integration expertise every step of the way.

Connect with our solutions experts today to review your requirements and find the ideal solution fit for your application.