NVIDIA GPU Products
NVIDIA® Tesla™ GPU Technology
Tesla™ computing solutions enable users to process large datasets with a massively multi-threaded computing architecture. By developing a parallel architecture from the ground up, NVIDIA® has designed its Tesla computing products to meet the requirements of HPC software. With the introduction of the Tesla 20-Series, the next-generation CUDA architecture (codenamed "Fermi"), you will enjoy a powerful new array of features:
| Tesla 10-Series | Tesla 20-Series | |
|---|---|---|
| Processing Cores | 240 | 448 |
| Double Precision Floating Point Capability | 78 Gflops | 515 Gflops |
| Single Precision Floating Point Capability | 933 Gflops | 1003 Gflops |
| Memory | 4GB DDR3 | 3GB GDDR5 (or 2.625GB GDDR5 w/ ECC) |
| L1 Cache (per streaming multiprocessor) | - | Configurable 48KB or 16KB |
| L2 Cache | - | 768KB |
| ECC Memory Support | No | Yes |
| Concurrent Kernels | No | Up to 16 |
| With ECC on, a portion of the dedicated memory is used for ECC bits, so the available user memory is reduced by 12.5%. For example, 3GB total memory yields 2.625GB of user-available memory. | ||
| Features | Tesla K20X | Tesla K20 | Tesla K10 |
| Number and Type of GPU | 1 Kepler GK110 | 1 Kepler GK110 | 2 Kepler GK104s |
| GPU Computing Applications | Seismic processing, CFD, CAE, Financial computing, Computational chemistry and Physics, Data analytics, Satellite imaging, Weather modeling | Seismic processing, CFD, CAE, Financial computing, Computational chemistry and Physics, Data analytics, Satellite imaging, Weather modeling | Seismic processing, signal and image processing, video analytics |
| Peak double precision floating point performance | 1.31 Tflops | 1.17 Tflops | 190 Gigaflops (95 Gflops per GPU) |
| Peak single precision floating point performance | 3.95 Tflops | 3.52 Tflops | 4577 Gigaflops (2288 Gflops per GPU) |
| Memory bandwidth (ECC off)1 | 250 GB/sec | 208 GB/sec | 320 GB/sec (160 GB/sec per GPU) |
| Memory size (GDDR5) | 6 GB | 5 GB | 8 GB (4 GB per GPU) |
| CUDA Cores | 2688 | 2496 | 3072 (1536 per GPU) |
In addition to the power of GPU parallel processing, you can benefit from the CUDA software development environment for parallel programming (including support for C, C++, Fortran, Open CL and Direct Compute) and a steadily expanding spectrum of high-performance computing applications.
