PNY Tesla K80M User Manual [ru]

NVIDIA® TESLA K80 by PNY

PART NUMBER:

TCSK80M-PB

NVIDIA TESLA K80 by PNY

THE WORLD’S FASTEST ACCELERATOR FOR DATA ANALYTICS AND SCIENTIFIC

Accelerate your most demanding single and double precision workloads in scientic computing, seismic processing, and data analytics applications 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 performance than the Tesla K10, and up to 10x faster performance than CPUs on real-world applications.

The Tesla K80 accelerator delivers more than 2x application speed-up compared to the previous generation of accelerators, and up to 10x faster performance compared to CPUs. With exclusive features like 24 GB of GDDR5 memory, 480 GB/s memory bandwidth, and improved GPU Boost technology, the Tesla K80 delivers the computational horsepower that allows you to crunch through petabytes of data and run simulations faster than ever before.

Tesla GPU Accelerators are built on the NVIDIA Kepler™ compute architecture and powered by CUDA,® the world’s most pervasive parallel-computing model. This makes them ideal for delivering record acceleration and compute performance efciency for applications in elds including: Machine Learning and Data Analytics, Seismic Processing, Computational Biology and Chemistry, Weather and Climate Modeling, Image, Video, and Signal Processing, Computational Finance/Physics, CAE and CFD.

The Kepler-based Tesla family of GPUs is part of the innovative Tesla Accelerated Computing Platform. As the leading platform for accelerating data analytics and scientic computing, it combines the world’s fastest GPU accelerators, the widely used CUDA parallel computing model, and a comprehensive ecosystem of software developers and software vendors.

The innovative design of the TESLA K80 compute architecture includes:

Zero-power Idle

Increases data center energy efciency by powering down idle GPUs when running legacy nonaccelerated workloads.

2x Shared Memory and 2x Register File

Increases effective throughput and bandwidth with 2x shared memory and 2x register le compared to the K40.

GPU Boost

Enables the end-user to convert power headroom to higher clocks and achieve even greater acceleration for various HPC workloads. Dynamically scales GPU clocks for maximum application performance and improved energy efciency

TESLA K801 - PRODUCT SPECIFICATIONS

MEMORY SIZE PER BOARD 24 GB GDDR5 (12 GB per GPU)

MEMORY INTERFACE 384-bit

MEMORY BANDWIDTH 480 Gb/s

CUDA CORES 4992

PEAK DOUBLE PRECISION FLOATING POINT PERFORMANCE

PEAK SINGLE PRECISION FLOATING POINT PERFORMANCE

2.91 Tops (GPU Boost Clocks)

1.87 Tops (Base Clocks)

8.74 Tops (GPU Boost Clocks)

5.6 Tops (Base Clocks)

SYSTEM INTERFACE PCI Express 3.0 x16

MAX POWER CONSUMPTION 300 W

THERMAL SOLUTION passive heat sink

FORM FACTOR

111.15 mm (H) x 267 mm (L) Dual Slot, Full Height

DISPLAY CONNECTORS None

POWER CONNECTORS 8-pin CPU power connector

PACKAGE CONTENT 1x Power Adapter (2 x PCIe 8-pit to single CPU 8-pin)

PART NUMBER TCSK80M-PB

More information: www.pny.eu/tesla

marks and/or registered trademarks of NVIDIA Corporation in the U.S. and other countries. The PNY logotype is a regis-

tered trademark of PNY Technologies. All other trademarks and copyrights are the property of their respective owners. Dez14

PNY Technologies Europe

sales@pny.eu | T: +33 (0)5 56 13 75 75

TELSA K80 - Features

NVIDIA® TESLA K80 by PNY

ECC MEMORY ERROR PROTECTION

datacenters and supercomputing centers. Both external and internal memories are ECC protected in Tesla K80.

Integrates the GPU subsystem with the host system’s monitoring and

Meets a critical requirement for computing accuracy and reliability in

SYSTEM MONITORING FEATURES

management capabilities such as IPMI or OEM-proprietary tools. IT staff can thus manage the GPU processors in the computing system using widely used cluster/grid management solutions.

Accelerates algorithms such as physics solvers, ray-tracing, and

L1 AND L2 CACHES

sparse matrix multiplication where data addresses are not known beforehand

ASYNCHRONOUS TRANSFER WITH DUAL DMA ENGINES

FLEXIBLE PROGRAMMING ENVIRONMENT WITH BROAD SUPPORT OF PROGRAMMING LANGUAGES AND APIS

Turbocharges system performance by transferring data over the PCIe bus while the computing cores are crunching other data.

Choose OpenACC, CUDA toolkits for C, C++, or Fortran to express application parallelism and take advantage of the innovative Kepler architecture.

End-user can convert power headroom to higher clocks and achieve

TESLA GPUBOOST

even greater acceleration for various HPC workloads on Tesla K80. Dynamically scales GPU clocks for maximum application performance and improved energy efciency

2X SHARED MEMORY AND 2X REGISTER FILE

ZERO-POWER IDLE

Increases effective throughput and bandwidth with 2x shared memory and 2x register le compared to the K40

Increases data center energy efciency by powering down idle GPUs when running legacy nonaccelerated workloads

Software and Drivers

SOFTWARE APPLICATION PAGE www.nvidia.com/teslaapps

TESLA GPU COMPUTING ACCELERATORS SUPPORTED OS Linux (64-bit) and Windows (64-bit)

DRIVERS www.pny.eu/drivers

LEARN MORE ABOUT TESLA DATA CENTER MANAGMENT TOOLS www.nvidia.com/softwarefortesla

TESLA in Comparison

TESLA K801TESLA K40 TESLA K20X TESLA K20 TESLA K10

Peak double-precision oating point performance (board, boost clocks)

Peak single-precision oating point performance (board, boost clocks)

Number of GPUs 2 x GK210 1 x GK110B 1 x GK110 2 x GK104

Number of CUDA cores 4992 2880 2688 2496 3072

Memory size per board (GDDR5) 24 GB 12 GB 6 GB 5 GB 8 GB

Memory bandwidth for board (ECC off)™ 480 Gb/s 288 Gb/s 250 Gb/s 208 Gb/s 320 Gb/s

Architecture features SMX, Dynamic Parallelism, Hyper-Q SMX

System Servers

2.91 Tops 1.66 Tops 1.31 Tops 1.17 Tops 0.19 Tops

8.74 Tops 5 Tops 3.95 Tops 3.52 Tops 4.58 Tops

Servers &

Workstations

Servers