PNY NVIDIA A800 40GB Active OEM - GA100 accelerator in tray supply for integrators
VCNA800-SB is the OEM version (tray, factory packaging) of the same A800 compute card built on the NVIDIA Ampere GA100 architecture. Technically it is identical to the retail one: 6912 CUDA cores, 432 third-generation Tensor cores, 40 GB of HBM2 memory on a 5120-bit bus with 1.5 TB/s bandwidth. The OEM format is convenient when assembling servers and workstation batches, where the retail box is not needed and what matters is the card itself and the supply terms.
Compute performance corresponds to the A100 class level: FP64 - 9.7 TFLOPS, FP32 - 19.5 TFLOPS, peak Tensor - up to 623.8 TFLOPS. The A800 covers AI model training and inference, full-precision scientific simulations and analytical data science workloads.
- Ampere GA100 architecture, 6912 CUDA + 432 third-generation Tensor cores
- 40 GB HBM2, 5120-bit bus, 1.5 TB/s
- Third-generation NVLink 400 GB/s - scaling two cards up to 80 GB
- MIG - up to 7 isolated GPU instances for multi-user environments
- PCIe 4.0 x16, TGP 240 W, dual-slot, active cooling
- 3-year NVIDIA AI Enterprise subscription included
The card has no display outputs - in a workstation it is complemented by a companion card (RTX 4000 Ada, RTX A4000 or T1000). Despite the OEM format, the official distributor warranty is retained. If you are planning a cluster or a multi-GPU node on the A800, an ETE.UA engineer will calculate the configuration and compatibility with your platform. Phone +380 93,594 00 77.
