Nvidia Tesla M2090 6GB TCSM2090-PB データシート

Features

Benefits

HUNDrEDS oF cUDA corES

Delivers up to 665 Gigaflops of double-precision peak 

performance in each GPU, enabling servers from leading 

oEMs to deliver more than a teraflop of double-precision 

performance per 1 rU of space.  Single precision peak 

performance is over one Teraflop per GPU.

Ecc MEMory Error ProTEcTIoN

Meets a critical requirement for computing accuracy and 

reliability in datacenters and supercomputing centers.  

Internal register files, L1/L2 caches, shared memory, and 

external DrAM all are Ecc protected.

UP To 6GB oF GDDr5 MEMory  
PEr GPU

Maximizes performance and reduces data transfers by 

keeping larger data sets in local memory that is attached 

directly to the GPU.

SySTEM MoNITorING FEATUrES 

Integrates the GPU subsystem with the host system’s 

monitoring and 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.

L1 AND L2 cAcHES AS PArT oF THE 
NVIDIA PArALLEL DATAcAcHE

™

Accelerates algorithms such as physics solvers, ray-tracing, 

and sparse matrix multiplication where data addresses are 

not known beforehand.

NVIDIA GIGATHrEAD

™

 ENGINE

Maximizes the throughput by faster context switching that  

is 10X faster than previous architecture, concurrent kernel 

execution, and improved thread block scheduling.

ASyNcHroNoUS TrANSFEr WITH 
DUAL DMA ENGINES

Turbocharges system performance by transferring data  

over the PcIe bus while the computing cores are crunching 

other data.  

FLEXIBLE ProGrAMMING 
ENVIroNMENT WITH BroAD SUPPorT 
oF ProGrAMMING LANGUAGES  
AND APIS

choose c, c++, opencL, Directcompute, or Fortran to express 

application parallelism and take advantage of the innovative 

“Fermi” architecture.  

 
To learn more about NVIDIA Tesla, go to www.nvidia.com/tesla

© 2011 NVIDIA corporation. All rights reserved. NVIDIA, the NVIDIA logo, NVIDIA Tesla, cUDA, GigaThread, Parallel Datacache, and 

Parallel NSight are trademarks and/or registered trademarks of NVIDIA corporation. All company and product names are trademarks or 

registered trademarks of the respective owners with which they are associated. Features, pricing, availability, and specifications are all 

subject to change without notice.

TESLA M-cLASS GPU coMPUTING MoDULE

SoFTWArE AND DrIVErS

 

> Software applications page:  

 

> Tesla M-class GPU computing modules are supported for both Linux and 

Windows 64-bit

 

– Drivers – NVIDIA recommends that users get drivers for M-series 

products from their System oEM to ensure that driver is qualified by the 

oEM on their system. Latest drivers can be downloaded from  

 

– Learn more about Tesla data center management tools at  

 

> Software development tools are available at