Nvidia have just now released their first HBM2 PCIe card, the Tesla P100, and in doing so have beaten AMD to the market with the new memory technology. While this card has been out for some time now with reports going back as early October 2016, it hasn’t been commercially available until now.
Nvidia have proclaimed this card to be “The most advanced data centre GPU ever built” for tasks such as “Predicting our climate’s future” and “Artificial intelligence for self-driving cars.”
The P100 has 2 different models a 16GB variant with 732GB/s memory bandwidth and a 12GB variant with 549GB/s memory bandwidth
Nvidia have also shown just how well these GPUs scale using NV-Link which will allow a system to scale up to 8 P100s in a single system at 5x the bandwidth of PCIe. Nvidia are saying that this has been designed to “Solve the world’s most important challenges that have infinite compute needs in HPC and deep learning.”
According to Nvidia this card will not only enhance performance greatly but it will save companies quite a bit of money too. A single GPU-accelerated node that’s powered by 4 Tesla P100s, interconnected with PCIe, can replace up to 32 commodity CPU nodes for a wide variety of applications. This in turn can bring potential savings of up to 70%, in saying that just one of these cards will set you back a whopping £6400.
What do you think of Nvidia’s new workstation card and are you disappointed that their first HBM2 card isn’t a gaming GPU? Tell us what you think in the comments section below.