VPX3-4937
3U VPX GPU Card with 9728 CUDA Cores, up to Four Video Outputs
The rugged VPX3-4937 high-performance card combines an NVIDIA® RTX™ 5000 Ada embedded GPU and a PCIe Gen4 switch in a 3U VPX module.
Designed and manufactured by WOLF Advanced Technology, this chip down design works with NVIDIA TensorRT™, CUDA, and CUDA Deep Neural Network (cuDNN). There is a significant performance increase per watt compared to the past GPUs due to more CUDA and Tensor cores operating at higher clock frequencies at the same power.
The VPX3-4937 supports SOSA aligned profiles, which makes it a plug-in upgrade for the VPX3-4935.
- Key Features
- Applications
- Specifications
- Related Products
- Developed in alignment with the SOSA® Technical Standard for plug-in upgrades
- NVIDIA RTX™ 5000 (AD103) GPU with 9728 CUDA cores, 304 Tensor cores, and 76 RT cores
- 16 GB GDDR6 256-bit memory with up to 576 GB/s
- Up to 3 DisplayPort outputs and 1 DVI output
- Module power: 90W to 130W, configurable
- ISR and EW applications that require the highest-performing GPU processing
- SWaP-constrained deep learning inference that can benefit from the largest number of Tensor Cores in 3U
- High-performance radar, SIGINT, EO/IR, sensor fusion, processing and display, and autonomous vehicles
Description
3U VPX NVIDIA Ada GPU Card with 9728 CUDA, 304 Tensor, & 76 RT cores
Generation
NVIDIA RTX™ 5000 (AD103) GPU with 9728 CUDA cores, 304 Tensor cores, and 76 RT cores
Memory Bandwidth
576 GB/s
PCIe Configuration
x16 Gen 4
Processor / FPGA
NVIDIA
Profile
3U SOSA Payload 14.6.11-0 or 14.2.3, 3U OpenVPX
Performance
27.7 TFLOPS
Backplane Fabric
16-lane PCIe4
Type
Up to 3 DisplayPort outputs and 1 DVI output, AdvMC-DSP-GPU
Card Type
GPU
Form Factor
3U VPX
Fabric Speeds
< 10 GbE
Video Interfaces
DisplayPort, DVI
Resources
Data Sheet
White Paper
Brochure
VPX3-4937
3U VPX GPU Card with 9728 CUDA Cores, up to Four Video Outputs
The rugged VPX3-4937 3U VPX GPU processor card combines an NVIDIA® RTX™ 5000 Ada embedded GPU and a PCIe Gen4 switch in a 3U VPX module.