Posted by alphaatlas 10:11 AM (CDT)
Thursday September 13, 2018
Nvidia revealed the Turing powered Tesla T4 in an announcement today. The Tesla T4 features 2560 CUDA cores and 16GB of memory with an effective bandwidth of 320 GB/s. With no display outputs and passive cooling, Nvidia is marketing the T4 as an AI inferencing card for data centers, and say it offers a peak formance of 8.1 teraflops for FP32, 65 teraflops for FP16,130 teraflops for INT8, and 260 teraflops for INT4. Nvidia squeezes all that into a 75W TDP, which means the card can be powered by a PCIe slot alone.
NVIDIA estimates that the AI inference industry is poised to grow in the next five years into a $20 billion market.