NVIDIA · 2016
NVIDIA P4 GPU accelerator
The NVIDIA P4 GPU accelerator is a server-oriented graphics card based on the Pascal architecture, designed for inference workloads.
The NVIDIA P4 GPU accelerator, released in 2016, is a server-focused graphics card built on the Pascal architecture. It features the GP104 chip on a 16 nm process, with 2560 shading units. The card operates at a base clock of 1138 MHz and a boost clock of 1531 MHz. It includes 8 GB of GDDR5 memory on a 256-bit bus, providing a memory bandwidth of 192 GB/s. With a TDP of only 75 W, it requires no external power connectors and uses a PCIe 3.0 x16 interface. Notably, it has no display outputs, as it is intended for server environments, particularly for inference tasks. It supports DirectX 12.0, Vulkan 1.2, and OpenGL 4.6.
Specifications
| Architecture unverified | Pascal |
|---|---|
| GPU chip unverified | GP104 |
| Process node unverified | 16 nm |
| Shading units unverified | 2560 |
| Base clock unverified | 1138 MHz |
| Boost clock unverified | 1531 MHz |
| Memory size unverified | 8 GB |
| Memory type unverified | GDDR5 |
| Memory bus unverified | 256 bit |
| Memory bandwidth unverified | 192 GB/s |
| Board power (TDP) unverified | 75 W |
| Bus interface unverified | PCIe 3.0 x16 |
| Display outputs unverified | No display outputs |
| DirectX support unverified | 12.0 |
| Vulkan support unverified | 1.2 |
| OpenGL support unverified | 4.6 |
| Release date unverified | 2016-09-13 |
| Release year unverified | 2016 |
| Notes unverified | No display outputs; designed for server/inference. |