NVIDIA® CUDA parallel computing architecture is enabled on GeForce, Quadro, and Tesla products. Whereas GeForce and Quadro are designed for consumer graphics and professional visualization respectively, the NVIDIA® TeslaTM product family is designed ground-up for parallel computing and offers exclusive computing features.
THE TESLA ADVANTAGE #1: PERFORMANCE
The NVIDIA® Tesla products are designed with exclusive features to maximize performance for computing professionals.
|Faster Double Precision Performance||515 GigaFlops||Much lower at 1/8th single precision|
|Faster PCI-E Communication||Two DMA engines for bi-directional communication||Only one DMA engine available|
|Larger Memory Size||3GB & 6GB available for HPC apps||Largest memory is 1.5GB|
|Optimized for InfiniBand||NVIDIA GPUDirect for accelerated communication||Extra memory copies and context switching|
|Performance Drivers||TCC driver on Windows reduces CUDA kernel overhead||Kernel launch overhead reduces GPU performance|
THE TESLA ADVANTAGE #2: RELIABILITY
Computational professionals depend on mission critical applications to run reliably without data error. All Tesla products undergo zero error tolerance testing in the most strenuous conditions for hundreds of times longer than any other NVIDIA products. Tesla exclusive features include:
|ECC Protection||All internal and external memories protected||No memories protected|
|Zero Error Stress Tested||Stress tested for several days with zero tolerance for error||Tested only for graphics apps|
|Manufactured by NVIDIA||Guaranteed by NVIDIA to be of highest quality||Manufactured by 3rd party board makers|
|Enterprise Level Support||Three year warranty and prioritized support for bugs and feature requests||Consumer level support|
|ISV Certification||HPC software vendors certify only on Tesla||Functionality/performance risk with software|
THE TESLA ADVANTAGE #3: HPC SOLUTIONS
HPC customers have unique requirements. Only Tesla business unit provides products and solutions to meet these special needs.
|Integrated Systems||Innovative Servers and workstations||Consumer oriented products|
|18-24 Month Availability||Longer life cycle for investment protection||Supply depends on consumer demand|
|Cluster management and GPU monitoring||Access GPUs in clusters for temperature, power, and usage||No support for datacenter features|
|HPC Expertise||Total solution expertise provided by CUDA/OpenCL engineers and technical staff||No HPC expertise or support|