Tesla Data Center Solutions

Tesla GPU Computing Solutions for Data Centers
Accelerate your science with NVIDIA® Tesla™ 20-series GPUs. A companion processor to the CPU in a server, Tesla GPUs speed up HPC applications by 10x. Based on the NVIDIA CUDA™ GPU architecture codenamed "Fermi", Tesla 20-series GPUs feature up to 665 gigaflops of double precision performance, 1 teraflop of single precision performance, ECC memory error protection, and L1 and L2 caches.

Deploy Tesla GPGPUs in the data center:

Integrated GPU-CPU servers with embedded Tesla M-class GPU modules (M2050 / M2070Q / M2070 / M2075 / M2090)

SELECTING THE RIGHT TESLA GPU

	Key Features	Applications	Where to Buy
TESLA M2090
Product brief Download (412 KB PDF)	Highest performance Fermi-based GPGPU 665 GFlops Peak DP 6 GB memory size (ECC off) 177 GB/sec memory bandwidth (ECC off) 512 CUDA cores	Seismic processing CFD, CAE Supercomputing
TESLA M2070/M2075
Product brief Download (412 KB PDF)	High performance, large memory Fermi-based GPGPU 515 GFlops Peak DP 6 GB memory size (ECC off) 150 GB memory bandwidth (ECC off) 448 CUDA cores	Seismic processing CFD, CAE Supercomputing Satellite imaging, GIS Weather modeling
TESLA M2070Q
Product brief Download (412 KB PDF)	Designed for remote graphics visualization and high performance computing in the data center 515 GFlops Peak DP 6 GB memory size (ECC off) 150 GB/sec memory bandwidth (ECC off) 448 CUDA cores	Visualization Applications CAD, CAM, CAE pre/post processing Medical Imaging 3D Animation Seismic interpretation GPU Computing applications Seismic processing CFD, CAE Supercomputing Satellite imaging, GIS Weather modeling
TESLA M2050
Product brief Download (412 KB PDF)	High performance Fermi-based GPGPU 515 GFlops Peak DP 3 GB memory size (ECC off) 148 GB/sec memory bandwidth (ECC off) 448 CUDA cores	Computational Finance Bio-informatics Molecular dynamics Data analysis

SOFTWARE AND DRIVERS

NVIDIA recommends getting drivers for M-class products from system OEMs. Please visit the NVIDIA Driver Downloads page for M-class drivers.

Tesla M-class products are supported under

Windows Server 2008 and 2008 R2 (all editions), 64-bit
Windows 7 Support (Tesla M2070Q Only)
Linux 32-bit and 64-bit

RHEL 5.4 Server
Ubuntu 9.10 Server
RHEL 4.8 Server
SLES 11

Click here to learn more about drivers and cluster management software tools for Tesla data center products.
Click here to learn more about C, C++, Fortran, and OpenCL software development tools for Tesla GPUs.

PRODUCT SPECIFICATIONS

	Tesla M2090	Tesla M2070/M2075	Tesla M2050
Peak double precision floating point performance	665 Gigaflops	515 Gigaflops	515 Gigaflops
Peak single precision floating point performance	1331 Gigaflops	1030 Gigaflops	1030 Gigaflops
CUDA cores	512	448	448
Memory size (GDDR5)	6 GigaBytes	6 GigaBytes	3 GigaBytes
Memory bandwidth (ECC off)	177 GBytes/sec	150 GBytes/sec	148 GBytes/sec

* Note: With ECC on, 12.5% of the GPU memory is used for ECC bits. For example, 3 GB total memory yields 2.625 GB of user available memory with ECC on.

Hardware Support

Knowledgebase
Our knowledgebase is available online 24x7x365 and contains answers to the most common questions and issues.

User Forums
Discuss Tesla products, talk about CUDA development, and share interesting issues, tips and solutions with your fellow NVIDIA Tesla users on the CUDA discussion forums.

For RMA requests, replacements and warranty issues regarding your NVIDIA based product, please contact the OEM or reseller that you purchased it from.

LEGACY PRODUCTS

Tesla 10-series products:

Tesla S1070-400 and Tesla S1070-500 1U systems (Product Brief)
Tesla M1060 GPU module

Tesla 8-series products:

Tesla S870 1U system (System Specification Document)