Graphics Cards

Nvidia tesla v100 from nvidia humiliates the tesla p100 gpu

Table of contents:

Anonim

Nvidia's new Tesla V100 GPU is already starting to ship to customers around the world, which is today the world's most powerful graphics chip. In the last few hours we have been able to see the performance improvements that Tesla V100 offers compared to its predecessor, the Tesla P100 that was launched in 2016.

Nvidia begins shipping Tesla V100 to its customers

The new Nvidia Tesla V100 GPU was created for heavy computing and artificial intelligence, but not for video games, for that market they already have the Pascal chip. The most knowledgeable are amazed at what Tesla V100 can do and the technology that has been implemented in it, 12nm manufacturing process , NVLINK 2.0, HBM 2.0, Tensor Cores and others.

The previous tesla P100 was based on the Pascal architecture, but Tesla V100 does it based on Volta, which is a completely new architecture and optimized for tasks that require a lot of computational processing. In the following table we can see the performance improvements that the V100 offers compared to the P100.

Comparison of performance against P100

From the start, we are already seeing 50% more teraflops performance in FP32 / FP64. 20% more bandwidth of HBM2 memory, 50% more L2 cache memory and up to 770% more level 1 cache memory, which goes from 1.3MB to 10MB. You can also see the breakthrough in DL Training and DL Inferencing, which is used for 'deep learning' tasks on computers.

In tests we can see the DGX-1 computer that comes with Tesla V100 and two 20-core Intel Xeon E5-2698 V4 processors. The score he gets is 743537 CUDA Score points, impressive.

Source: wccftech

Graphics Cards

Editor's choice

Back to top button