Nvidia unveils 350lb 'DGX-2' supercomputer with a $399,000 price tag

Polycount

Posts: 3,017   +590
Staff

While Nvidia may be best-known for their popular GeForce line of gaming GPUs, those are far from the only products the company sells. Nvidia's Quadro GPUs, for example, are geared more towards industrial design and advanced special effects rendering than gaming.

Nvidia has also dabbled in the realm of AI processing with their DGX-1 supercomputer. However, it seems one supercomputer offering isn't enough for Nvidia.

At the GPU Technology Conference on Tuesday, company CEO Jensen Huang unveiled the next evolution of their commercial supercomputer, simply dubbed the DGX-2.

Much like the DGX-1, the DGX-2 is geared towards machine learning and AI research in general. That said, despite being announced a mere 6 months after the DGX-1 officially began shipping out, the DGX-2 is roughly 10 times as fast as its predecessor.

Weighing in at a whopping 350lb, it isn't hard to see how Nvidia pulled off that speed increase; the DGX-2 is filled to the brim with top-of-the-line hardware. To begin with, the system houses an impressive array of 16 Tesla V100 GPUs spread across two separate GPU boards.

Each V100 contains 32GB of HBM2 memory, adding up to 512GB in total. The system also contains 1.5TB of standard RAM and two Intel Xeon Platinum CPUs. If storage space is your primary concern, the DGX-2 will launch with 30TB of NVME SSD storage by default.

"The extraordinary advances of deep learning only hint at what is still to come," said Huang. "Many of these advances stand on NVIDIA’s deep learning platform, which has quickly become the world’s standard. We are dramatically enhancing our platform’s performance at a pace far exceeding Moore’s law, enabling breakthroughs that will help revolutionize healthcare, transportation, science exploration and countless other areas."

Whether or not Nvidia's lofty ambitions will come to pass remains to be seen but machine learning companies will be able to see for themselves soon enough. If you handle equipment acquisition for such a company and you can stomach its $399,000 price tag, the DGX-2 is available to order right now. The supercomputer is expected to ship out to customers sometime in Q3 2018.

Permalink to story.

 
Hmmmmmm ..... I'll stick with my Cray ..... nicer shape and it makes a great footstool .....
 
I saw this kind of machine being used by a power company in my city. They wereteaching the machine how to discover electricity theft. A common pratice where I'm from (Brazil). Last year they discovered an ice cream factory. Their power consumption alone was enough for 80 houses.
So yes... Company who buy this are not doing for bragging rights, like a gaming GPU, but they see it as an investment.
 
I wonder how many miners are going to buy these

Begin paste>

I'm guessing ~100Mh/s from each GPU, x16 = 1.6Gh/s

Power will be ~220W each (maybe less for these newer babies) x16 = 3.52kW

Add in a couple of meaty Platinum 8180's, that can draw 205W each, but will likely draw 40 something watts (off the cuff guess), whilst idling away. Couple that will some mammoth m/b that want a couple hundred watts, all that RAM & unused NVME SSDs we'll round that up to 500W.

So my guestimation is ~4kW, lets throw in 5% PSU conversion losses (I suppose they are gonna be good), so we are looking at 4.2kW continuous power draw, for Ethereum mining. Less than 5kW total for sure though.

So $?

Annually:
46.46973093 coins mined.
Power Cost [10c/kWh] (in USD) $3,679.20
Profit (in USD) $17,398.55 per year.
Days to break even: 8391.50 Day(s).

If we re-run this with my UK energy costs:

Days to break even: 9610.94 Day(s).

Clearly I am on holiday too - to even bother with this response.

In addition, someone ran the V100's on an Amazon AWS cluster for nearly an hour, and with all costs considered, came out at *negative* $25k USD/year. Interesting though, and well written up.

End paste>
 
There's a lot of stuff to care about in regards to nvidia right now: new GPUs, supply and pricing concerns related to cryptocurrency mining, and antitrust law violations by way of GPP.
 
Back