Google Cloud Will Add GPU Services in Early 2017

Google Cloud will add GPUs as a service early next year, the company has said. Amazon Web Services, Microsoft Azure and IBM's Bluemix all already offer GPU as a service. From a report on GeekWire: Google may be seeking to distinguish itself, however, with the variety of GPUs it's offering. They include the AMD FirePro S9300 x2 and two offerings from NVIDIA Tesla: the P100 and the K80. And Google will charge by the minute, not by the hour, making GPU usage more affordable for customers needing it only for short periods. CPU-based machines in the cloud are good for general-purpose computing, but certain tasks such as rendering or large-scale simulations are much faster on specialized processors, Google explained. GPUs contain hundreds of times as many computational cores as CPUs and excel at performing risk analysis, studying molecular binding or optimizing the shape of a turbine blade. Google's GPU services will be available in early 2017 through Google Compute Engine and Google Cloud Machine Learning.

Excellent

[110010001000]

This can be used for bitcoin mining and "deep learning" or possibly "deep learning about bitcoin mining". All I need is some VC backing.

Re:

[lgw]

Naturally, the slashvertizement for Google cloud doesn't mention that AWS has had GPU instances for years. Azure seems to have had it for a year now. Google is really playing catch-up in cloud services (though they aren't a joke like Oracle).

I hope Whipslash got a check for this slashvertizement - be good to see some advertizement flowing out of Goggle for once (and anything that keeps /. afloat).

Re:

[BronsCon]

Amazon Web Services, Microsoft Azure and IBM's Bluemix all already offer GPU as a service.

Re:

[haruchai]

Try reading the 2nd sentence of the 1st line of the story after the headline

Re:Excellent

[ShanghaiBill]

This can be used for bitcoin mining

Nope. You cannot mine bitcoins profitably with a GPU. Not even FPGAs are sufficient. You need ASICs.

A GPU can give you a few mega-hashes per joule. An FPGA can do about 20 MH/j. The best ASICs can do about 10,000. It is not even close.

Mining hardware comparison [bitcoin.it].

Re:

[110010001000]

That is why I combine "Deep Learning" with it. Deep learning makes everything possible.

Re:

[K. S. Kyosuke]

Note that "deep learning" doesn't refer to your rectal extractions.

Re:

[ShanghaiBill]

For instance, you could mine Ethereum on a GPU.

Not anymore. Ethereum ASICs have been available since August. GPUs can no longer compete. Ethereum mining takes much more memory than bitcoin, so it doesn't scale as well when going from GPU to ASIC, but the cost saving is still substantial.

Re:

[ShanghaiBill]

AWS already offers it, why should I care?

1. Because competition is good.
2. Because renting by the minute, rather than the hour, is useful. When developing OpenCL [wikipedia.org] applications, I spend more time debugging and tuning in short bursts, than I do running in production.

Re:

[vision33r]

Mining bitcoins using GPU is the most power consuming operation, mining BTC in general is a waste of power what little BTC you can get that requires a ton of electrical power to do so.

1000 GPU "Hello World" program

[Ukab the Great]

The ultimate symbol of wealth, status and power in the geek community.

Cost?

[Visarga]

What counts is price and performance. Is there a comparison to other offers?

GaaS

[unixisc]

Should solve our energy issues w/o leaving a big carbon footprint, right?

Re:

[subk]

Why is this not +5 Funny?