Nvidia announces Tesla p40 and P4 accelerators on




Nvidia has announced two new Tesla cards with Pascal GPUs. The Tesla P40 is based on a GP102-GPU 3840 cudacores. That’s more than the Tesla P100, Nvidia’s most powerful accelerator, but the P40 does not hbm2 memory.

The Tesla P40 and P4 are the successors of the M40 and M4, which still contain Maxwell GPUs. Nvidia prices not yet announced, but the p40 and P4 cards will be more affordable than the high-end P100 , as they become available respectively in October and November.

Nvidia focuses cards on inferencing applications, with deep neural networking is used to recognize speech, image and text. Nvidia combines the introduction of the cards then two software tools. TensorRT is a library for the optimization of deep learning models, while the Deep Stream SDK decoding and analyzing video streams to accelerate by artificial intelligence.

It is striking that Nvidia has enabled the P40 3840 stream processors, this being at the flagship P100 3584. That card, however, has a considerably wider memory interface coupled with fast hbm2 memory. As a result, the memory bandwidth of the P100 relative to the P40 is considerably higher: 720GB / s in front of 346GB / s. With Nvidia Tesla P100 focuses on the training of deep-learning networks, thanks to the considerable fp16-processing power of 21.2 teraflops.


In: A Technology & Gadgets Asked By: [22628 Red Star Level]

Answer this Question

You must be Logged In to post an Answer.

Not a member yet? Sign Up Now »