Tesla Advantage(Tesla與GeForce比較)

張貼者:2010年6月29日 上午12:55蔡長明   [ eddie liu 已於 2013年1月10日 下午11:06 更新 ]
Tesla系列擁有更快速的雙精準的浮點運算速度,可以達到將近515 GigaFlops,
Tesla系列擁有更快速的PCI-E通訊,擁有兩個DMA引擎,針對bi-directional 通訊。而GeForce系列僅有一個DMA引擎。
Tesla系列擁有更大的記憶體,高達3GB&6GB的選擇,針對HPC的應用程式可以發揮更大的功效。GeForce系列最大的記憶體為1.5GB
相關優勢如下表比較
   




Advantages

Tesla

GeForce

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




Advantages

Tesla

GeForce

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



Advantages

Tesla

GeForce

Integrated Systems

Innovative servers and workstations

System design and support not on consumer 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









Comments