NVIDIA Tesla K40 GPU Accelerator (Passive Cooling)

NVIDIA Tesla K40 GPU Accelerator (Passive Cooling)

NVIDIA Tesla K40 GPU Accelerator (Passive Cooling)

B&H # NVTK40 MFR # 900-22081-0040-000
Special Order
Expected availability: 3-7 business days

Product Highlights

  • GPU-Accelerated Parallel Processing
  • 745 MHz GPU - Up to 875 GHz with Boost
  • 2880 CUDA Cores
  • 12GB GDDR5 RAM
  • 3 GHz Effective Memory Clock Speed
  • 288 GB/s Memory Bandwidth
  • 4.29 TFLOPs Singe Precision Performance
  • 1.43 TFLOPs Double Precision Performance
  • Supports CUDA and Other APIs
  • PCI Express 3.0 x16 Interface
Show moreShow less

This item is noncancelable and nonreturnable

$0.00 Tax Collected Outside NY and NJ

You Pay: $3,359.95

Add to Wish List Item in Wish List

NVIDIA K40 overview

  • 1Description

The Tesla K40 GPU Accelerator with passive cooling from Nvidia is a GPU with no video outputs designed exclusively for providing acceleration to assist computational intensive tasks such as transcoding video, rendering 3D models, cryptography, and analysis of complex data sets. It features a 745 MHz core which can be boosted (overclocked) by the end user up to 875 MHz, and has 2880 Kepler generation CUDA cores (stream processors) which allow it to perform hundreds of computations simultaneously. To ensure data moves to and from the processors as efficiently as possible, the card features 12 GB of GDDR5 RAM over a 384-bit memory interface. A 750 MHz native memory clock speed gets effectively boosted to 3 GHz thanks to the fact GDDR5 type memory moves data at four words per cycle. Altogether this yields a memory bandwidth of 288 GB/s.

4.3 TFLOPs Single Precision Performance
The card boasts a double precision (64-bit length) benchmark of 1.43 TFLOPs (trillion floating point operations per second), and a single precision (32-bit length) benchmark of 4.3 TFLOPs. Floating point math is commonly found in applications such as 3D modeling as well as certain forms of data analytics
Supports CUDA and Other APIs
APIs provide the software interface that exposes applications to the compute power of the GPU for processing-intensive tasks, including transcoding video, cryptography, analysis of complex data sets, rendering 3D environments, CAD modeling, and more
SMX (Streaming Multiprocessor)
Redesigned architecture with more cores, delivers up to 3x more performance per watt than the SM technology in previous-generation Nvidia Fermi GPUs
Dynamic Parallelism
Enables GPU threads to automatically spawn new threads. By adapting to the data without going back to the CPU, this greatly simplifies parallel programming
Hyper-Q
Allows multiple CPU cores to simultaneously use the CUDA cores on a single Kepler GPU. This dramatically increases GPU utilization while reducing CPU idle times
Multi-Language SDKs
For software developers, SDKs are available for OpenACC, CUDA, C, C++ and Fortran
ECC Error Correction
Protects the data integrity from interference caused by cosmic background radiation
System Monitory Features
Integrates the GPU subsystem with the host system's monitoring and management capabilities, such as IPMI or OEM-proprietary tools. IT staff can now manage the GPU processors in the computing system using widely used cluster/grid management solutions
L1 and L2 Caches
Accelerates algorithms such as physics solvers, ray tracing, and sparse matrix multiplication where data addresses are not known beforehand
Tesla GPUBoost
End-user can convert power headroom to higher clocks and achieve even greater acceleration for various computationally-intensive workloads on Tesla K40
Passive Cooling for High MTBFs
A heatsink cooling system means no moving parts to wear out over time. And since there is no fan, the card operates silently for use in edit suits and other noise-sensitive environments
Quad Data Rate (QDR) GDDR5 RAM
Because GDDR5 RAM can transfer up to four words per cycle, as opposed to one word per cycle with traditional RAM, the effective data rate is quadrupled, allowing this card to process data transactions as quickly as if it had a 3.0 GHz memory clock speed
PCI Express 3.0 x16 Support
Provides twice the bandwidth of PCI Express 2.0; for example, an x8 PCI Express 3.0 slot has the same bandwidth as a x16 PCI 2.0 slot. In addition, the card remains backward-compatible with motherboards using earlier generations of PCI Express
In the Box
NVIDIA Tesla K40 GPU Accelerator (Passive Cooling)
  • Limited 3-Year Warranty
  • Table of Contents
    • 1Description

    NVIDIA K40 specs

    Performance
    GPU 1 x Kepler GK110B
    GPU Clock Base: 745 MHz
    With Boost: 810 / 875 MHz
    Stream Processors (CUDA Cores) 2880
    Memory 12 GB
    Memory Type GDDR5
    Memory Clock Actual: 750 MHz
    Effective (QDR): 3 GHz
    Memory Interface 384-bit
    Memory Bandwidth 288 GB/s
    BIOS 2 Mbit serial ROM
    BAR1 Size: 16 GB
    Floating Point Performance Double Precision: 1.43 TFLOPs
    Single Precision: 4.29 TFLOPs
    System Requirementsdyywxtwtcazy
    Software Windows (32- and 64-bit): 8.1, 7, Vista, XP, Server 2008 R2
    Windows (64-bit only): Server 2012 R2
    Linux (x86 64-bit and ARM): Support varies depending on distribution
    Hardware 1 x Free PCI-E x16 slot for electrical interface (PCI-E 3.0 x16 recommended for best performance)
    2 x Free PCI-E slots physical
    1 x 6-pin PCI-E power connector
    1 x 8-pin PCI-E power connector
    General
    Thermal Solution Passive heatsink
    Power Consumption Maximum: 235 W
    Idle: 16 W
    MTBF GB@ 35C: 282,847 hours
    GF@ 35C: 252,222 hours
    Form Factor Dual Slot: Full-height, full-length
    Dimensions 4.4 x 10.5" x dual slot / 111.2 x 267 mm x dual slot
    Weight Not specified by manufacturer
    Packaging Infodyywxtwtcazy
    Package Weight 2.4 lb
    Box Dimensions (LxWxH) 11.9 x 9.1 x 2.4"

    NVIDIA K40 reviews

    Be the first to review this item

    NVIDIA K40 Q&A

    See any errors on this page? Let us know.