NVIDIA Tesla K80 24GB GDDR5 Graphics GPU Server Accelerator 900-22080-0100-000

NVIDIA Tesla K80 24GB GDDR5 Graphics GPU Server Accelerator 900-22080-0100-000

Experience 10x Faster Application Performance: Accelerate your most demanding single- and double-precision workloads in scientific computing, seismic processing, and data analytics by upgrading to the NVIDIA Tesla K80 dual-GPU accelerator. It delivers up to 2.2x faster performance than the Tesla K20X, up to 2.5x faster than the Tesla K10, and up to 10x faster than CPUs on real-world applications.

Key Specifications:
Double-Precision Performance: Up to 2.91 Teraflops with NVIDIA GPU Boost™
Single-Precision Performance: Up to 8.74 Teraflops with NVIDIA GPU Boost
Memory: 24 GB GDDR5 (12 GB per GPU)
Memory Bandwidth: 480 GB/sec per board
Throughput: 2x application throughput with the dual onboard GPUs

As part of the Tesla Accelerated Computing Platform, the Tesla K80 utilizes a robust ecosystem of software, hardware, and support to meet the highest demands in datacenter applications.

Performance Advantages:
2x Speed-up: Compared to previous-generation accelerators
10x Faster: Compared to CPUs for selected applications

Detailed Product Specifications:
Brand: NVIDIA
Part Number: 900-22080-0100-000
MPN: 900-22080-0100-000, S26361-F2222-L81
Type: Graphics Card
Chipset Manufacturer: NVIDIA
GPU Model: NVIDIA Tesla K80
Memory Size: 24 GB GDDR5 (12 GB per GPU)
Memory Interface Width: 384-Bit
Memory Bandwidth: 480 GB/s
Core Speed:
Boost Clock Speed: 875 MHz
Base Clock Speed: 560 MHz
Memory Speed: 10 GHz Effective
APIs Supported: CUDA, DirectX 12, Direct3D, Vulkan, Mantle
Connectors: Component, DVI, Composite (RCA), DisplayPort, D-Sub
Maximum Power Consumption: 300 W
PCI Power Connectors: 1 x 8-Pin PCI-E
Compatible Slot: PCI Express 3.0 x16
CUDA Cores: 4992

Tesla K80 Exclusive Features:
GPU Boost: Dynamically scales clock speeds based on workload demands to optimize performance while adhering to power and thermal limits.
Double Shared Memory and Register File: Doubles shared memory and register file compared to Tesla K20X and K10, increasing effective bandwidth.
Zero-Power Idle: Powers down idle GPUs to enhance energy efficiency in data centers.
Multi-GPU Hyper-Q: Efficient scheduling of MPI ranks across GPUs to maximize GPU utilization.
System Monitoring: Integrated management of GPU processors in computing systems using common cluster and grid solutions.
Memory Protection: ECC memory protection for both internal memories and external GDDR5 DRAM, ensuring accuracy and reliability in supercomputing.