NVIDIA announces Pascal GPUs for AI and neural network

NVIDIA is pushing its Pascal architecture even further with the launch of Tesla P4 and P40 GPUs. The company’s latest GPUs are designed for artificial intelligence and neural network programming.


The powerful Tesla P40 GPU clocks in at 12 teraflops for single precision calculation and capable of 47 trillion 8-bit INT8 operations per second. The Tesla P40 GPU pairs a 24 GB GDDR5 GPU with 3,840 CUDA cores. While P4 is less powerful, it still offers 5.5 teraflops single precision calculation and 22 trillion INT8 operations per second. The GPU uses 8 GB DDR5 memory coupled with 2,560 CUDA cores. The Tesla P40 has a memory bandwidth of 346Gbps while the P4 has a memory bandwidth of 192Gbps.

“With the Tesla P100 and now Tesla P4 and P40, NVIDIA offers the only end-to-end deep learning platform for the data center, unlocking the enormous power of AI for a broad range of industries,” said Ian Buck, general manager of accelerated computing at NVIDIA. “They slash training time from days to hours. They enable insight to be extracted instantly. And they produce real-time responses for consumers from AI-powered services.”

NVIDIA’s Tesla P40 and P4 also support TensorRT, a library created for optimizing the deep learning models. In order to tap into the power of a Pascal server, the new GPUs add support for NVIDIA’s DeepStream SDK, which can simultaneously decode and analyse up to HD video streams in real-time.

NVIDIA is also collaborating with Chinese search engine Bride for delivering highest compute performance. NVIDIA Tesla P40 will be available next month while the P4 will only be available from November.

0 thoughts on “NVIDIA announces Pascal GPUs for AI and neural network