Graphics Cards

Nvidia tesla p100 with hbm2 memory arrives

Table of contents:

Anonim

It started talking about it months ago, but it was finally today when the new Nvidia Tesla P100 card with a PCI-Express / NVLink interface and advanced HBM2 memory was announced, thus Nvidia is ahead of AMD in being a pioneer in its use. of this memory stacked, even if not on a card intended for players.

Nvidia Tesla P100: features

The Nvidia Tesla P100 is a new graphics card oriented to the servers, Nvidia also has a version of this same card with the NVLink interface and 16 GB of HBM2 memory. The variant with PCI-Express interface comes in two versions with 12 GB and 16 GB of HBM2 memory.

The GeForce GTX Titan Xp is here, the most powerful graphics card.

Nvidia Tesla P100 is based on a powerful and highly efficient Pascal GP102 GPU fully unlocked with 3840 Cuda Cores and 240 TMUs, the same as the one used in the GeForce GTX Titan Xp announced a couple of days ago. As for the HBM2 memory, we find a 4096-bit interface that allows it to reach a maximum bandwidth of 720 GB / s, well above the 512 GB / s reached by AMD's Fiji silicon-mounted HBM memory. The 12GB memory variant reduces bandwidth to 540GB / s.

With these characteristics the Nvidia tesla P100 offers a maximum power of 9.3 TFLOPs in single precision and 4.7 TFLOPs in double precision in the PCI-Express model and a power of 10.6 TFLOPs and 5.3 TFLOPs in the NVLink variant, therefore the frequency of Core performance will be different in the two versions. Its TDP is kept at 250W and is passively cooled.

More information: Nvidia

Graphics Cards

Editor's choice

Back to top button