NVIDIA Tesla GPU cards
 |
Buy NVIDIA Tesla in Singapore from official NVIDIA Tesla Preferred Partner |
for Workstations:
These Tesla cards are the cards with an active cooling heat sink to ensure proper card cooling and functioning in the workstation chassis. Workstation cases of all vendors do not have a GPU card focused airflow within a case, so GPUs with own cooling fan is a requirement.
| NVIDIA® Tesla™ Kepler K40 |
|
 |
|
| Brochure | NVIDIA Tesla K40 Fact Sheet |
| Peak double precision floating point performance |
1.43 Tflops |
| Peak single precision floating point performance |
4.29 Tflops |
|
Memory bandwidth |
388 GB/sec |
| Memory size (GDDR5) | 12 GB |
| CUDA cores | 2,880 |
| Core clock speed | 745MHz |
| Boost clock speed NEW Feature! Read more... |
Could be set to 810MHz or 875MHz |
| Architecture | Kepler |
| Power Consumption | 235W TDP |
for Servers:
Tesla cards for Data Centers have passive heat sinks and are designed for dedicated airflow tunnel cooling where airflow is directed straight to the heat sink of a Tesla card to maximize cooling efficiency. These Tesla cards are recommended strictly for the server use.
|
NVIDIA®Tesla™ |
NVIDIA®Tesla™ |
|
 |
 |
|
| Peak double precision floating point performance | 2.91 Tflops | 1.43 Tflops |
| Peak single precision floating point performance | 8.74 Tflops | 4.29 Tflops |
| Memory bandwidth | 480 GB/sec (2x 240 GB/sec) |
388 GB/sec |
| Memory size (GDDR5) | 24 GB (2x 12GB) |
12 GB |
| CUDA cores | 4,992 (2 x 2,496) |
2,880 |
| Core clock speed | 560MHz | 745MHz |
| Boost clock speed NEW Feature! Read more... |
Have 10+ levels: From 560MHz to 875MHz finer granularity between levels |
Have 3 levels: 745MHz, 810MHZ and 875MHz |
| Architecture | Kepler | Kepler |
| Power Consumption | 300W TDP | 235W TDP |
