Nvidia's hugely powerful $3,000 Titan V PC GPU is fastest ever

[steve2470]

https://www.cnet.com/news/nvidias-3000-titan-v-gpu-offers-massive-power-for-machine-learning/

(a few weeks old but still interesting)

Looking for a way to turn your home computer into a deep-learning AI super-monster? Nvidia has an expensive answer.

The new Titan V GPU promises a crazy amount of processing for deep learning and AI applications. It's nine times more powerful -- at 110 teraflops -- than last year's Titan X, Nvidia's last massive desktop graphics processor aimed at machine learning applications.

The Titan V is based on Nvidia's newer Volta chip architecture, which is also being used in Nvidia's Xavier self-driving car system and for data centers. This is the first desktop-based use of the new chip, and it costs $2,999 or £2,700 (about AU$4,800). Hardly a consumer purchase, but it's meant for people who want a massive dose of machine learning processing. Developers, labs, AI researchers... or anyone who wants an insane supercomputer.

It's available now on Nvidia's store.

[htuttle]

1. They need to rename 'GPU's

So, humans figured out a new use for graphics cards, and now they are used for all sorts of other things.

If you go to Nvidia's homepage for the Titan X, there's not a single mention of 'display monitor' or 'graphics' or anything related to that.

For nostalgia's sake, I'd suggest bringing back the term, "Math Co-Processor".

[Voltaire2]

2. vector matrix multiplication unit just rolls off the tongue. nt.

[lapfog_1]

3. DOe and IBM are installing these new GPUs in the next three

fastest Supercomputers in the world... the so-called CORAL project (Collaboration Oak Ridge Argonne Livermore).

https://www.hpcwire.com/2017/12/06/ibm-begins-power9-rollout-backing-doe-google/

Up to 6 of the Nvidia plus 2 PowerPC nextgen CPUs, lots of memory, and NVLink (Nvidia's proprietary replacement for PCIe) AND PCIe Gen4 (next gen PCIe) all in a 2U water chilled (external chilled water required) server.

I think Livermore is installing like 230+ racks of these servers plus next gen Mellanox Infiniband communications between servers. 125+ Pflops at 64 bit floating point. At 16 bit "AI" or "deep learning" should be close to .5 exaflops.

And they are still known as GPUs and not math co-processors because they only do certain floating point math operations at speed...

[htuttle]

4. True, but Graphics Processing Unit doesn't really capture it either, nowadays

It's like referring to a rocket as an 'Up Goer'. I'm liking Vector Matrix Multiplication Unit, more and more.

"VMMU" rolls off the tongue just as easily as 'GPU'.

Then we can get back to rating 'video cards' by their RAM, as the Gods intended.