NVIDIA Tesla P40 24GB PCIe GPU Accelerator
Passive-cooled Pascal data-center GPU with 24 GB GDDR5 memory, PCIe 3.0 x16 host interface and 250 W board power for virtual workstation, rendering and inference workloads.
- NVIDIA Pascal GPU with 3,840 CUDA cores.
- 24 GB GDDR5 framebuffer with up to 347 GB/s memory bandwidth.
- PCI Express 3.0 x16, dual-slot rack-server form factor.
- 250 W total board power with passive cooling; system airflow is required.
- NVIDIA product brief identifies INT8 support for deep-learning inference.
No promotion photo is attached to this SKU yet. Ask sales for current unit and packaging photos before ordering.
Product Overview
NVIDIA positions the Tesla P40 as a Pascal-generation GPU accelerator for virtualized visual computing, simulation, rendering and compute workloads. The official datasheet highlights support for Quadro vDWS software and CUDA/OpenCL workloads in virtualized environments.
The Tesla P40 product brief identifies the board as PG610 SKU 200 / NVPN 699-2G610-0200-100, with 24 GB GDDR5 memory, PCIe 3.0 x16 connectivity, passive thermal design and a 250 W maximum power limit.
For AI inference deployments, NVIDIA notes INT8 instruction support in the product brief, enabling high-throughput inference use cases when paired with a compatible system and software stack.
Key Highlights
Architecture
PascalOne NVIDIA Pascal GPU according to the official Tesla P40 datasheet.
CUDA Cores
3,840CUDA core count listed in NVIDIA's Tesla P40 product specifications.
Memory
24 GB GDDR5Large framebuffer for graphics, visualization and compute workloads.
Board Power
250 WPassive data-center card requiring host-system airflow.
Verified Capabilities
Compute and virtualization
- vGPU profiles: Official datasheet lists 1 GB, 2 GB, 3 GB, 4 GB, 6 GB, 8 GB, 12 GB and 24 GB profiles.
- NVENC density: NVIDIA lists support for 24 H.264 1080p30 streams.
- Workload support: Datasheet references CUDA and OpenCL workloads with NVIDIA virtual GPU software.
Board integration
- Host interface: PCI Express 3.0 x16 interface.
- Thermal solution: Passive cooling; product brief states system airflow is required.
- Auxiliary power: One CPU 8-pin auxiliary power connector listed in NVIDIA's product brief.
Technical Specifications
| Product SKU | PG610 SKU 200 |
|---|---|
| NVIDIA NVPN | 699-2G610-0200-100 |
| GPU architecture | NVIDIA Pascal |
| CUDA cores | 3,840 |
| Memory | 24 GB GDDR5 |
| Memory interface | 384-bit |
| Memory bandwidth | Up to 347 GB/s |
| Host interface | PCI Express 3.0 x16 |
| Form factor | 4.376 in x 10.5 in, dual-slot |
| Total board power | 250 W |
| Thermal solution | Passive |
| Power connector | One CPU 8-pin auxiliary power connector |
| Operating temperature | 0 C to 45 C |
| Storage temperature | -40 C to 75 C |
| Condition | New |
| Manufacturer P/N | 699-2G610-0200-100 |
| Inventory listed date | 2026-04-29 |
Specifications are summarized from NVIDIA Tesla P40 GPU Accelerator Product Brief: https://images.nvidia.com/content/pdf/tesla/Tesla-P40-Product-Brief.pdf
Use Cases
Virtual Workstations
For environments using NVIDIA virtual GPU software to deliver remote visualization and design workloads.
Inference Servers
Pascal GPU with INT8 support can be used in compatible deep-learning inference deployments.
Rendering and Simulation
The official datasheet references simulation, HPC rendering, CUDA and OpenCL workloads.
Frequently Asked Questions
Need verified availability or a formal quote?
Send us the part number, target quantity and destination. We will confirm stock, packaging condition, lead time and current promotional pricing.