P40 Enterprise GPU Pascal Architecture 3840 CUDA Cores Video Transcoding
Tesla P40: 24GB ECC GDDR5 memory, 3840 CUDA cores, 47 TOPS INT8 performance. Ideal for AI inference, video transcoding, and virtual workstations. Passive cooling with 250W TDP, supports Quadro vDWS and 24 NVENC sessions for multi-tenant virtualization.
Memory Interface
320-bit
Boost Clock
1710 MHz
Port
Tianjin Guangzhou
Gpu Memory Bandwidth
933 GB/S
Form Factor
4.4" (H) x 10.5" (L) Dual Slot
Type
Wireless
Product Description
P40 Enterprise GPU Pascal Architecture 3840 CUDA Cores Video Transcoding
Product Overview
The Tesla P40 is a dual-slot data center GPU based on the 16nm Pascal architecture, released in 2016. It features 24GB ECC GDDR5 memory with 346GB/s bandwidth, 3840 CUDA cores, and delivers 47 TOPS INT8 inference performance, ideal for AI inference, video transcoding, and virtual workstations.
With 250W TDP and passive cooling, it requires server airflow for heat dissipation. Supporting Quadro vDWS and 24 NVENC sessions, it efficiently handles multi-tenant virtualization and 3D rendering tasks. A reliable, cost-effective solution for legacy data center workloads and moderate AI inference needs.
Technical Specifications
| Architecture | 16nm Pascal |
| Memory | 24GB ECC GDDR5 |
| Memory Bandwidth | 346GB/s |
| CUDA Cores | 3840 |
| INT8 Performance | 47 TOPS |
| TDP | 250W |
| Cooling | Passive |
| Virtualization | Quadro vDWS, 24 NVENC sessions |
Recommended Products
Price: $5373-7000/unit
Price: $2955-4500/UNIT
Price: $3000-4500/UNIT
Interested in this product?
Contact us now for more details and a custom quote.