The future of AI is on your desktop - the Nvidia Titan V and Volta GPU architecture are here

[ack]

AI is coming to our desktops - AI, deep learning and super-computing power will soon be in our homes, starting now with the new Nvidia Volta GPU architecture and the Titan V graphics card, aimed at the AI world. This is exciting and terrifying news, especially given that the cost is only $3000

https://www.fool.com/investing/2017/12/11/nvidias-new-titan-v-gpu-turns-pc-into-ai-supercomp.aspx

NVIDIA (NASDAQ:NVDA) introduced TITAN V, what it calls "the world's most powerful graphics processing unit (GPU) for the PC," on Thursday evening at the annual NIPS (Neural Information Processing Systems) conference. This monster graphics card isn't targeted at gamers, however, but at developers working in the artificial intelligence and high-performance computing fields.

TITAN V's 21.1 billion transistors deliver 110 teraflops of raw horsepower, or more than 9 times that of its predecessor, Pascal-based TITAN Xp, and "extreme energy efficiency," according to the company's press release. TITAN Xp was also made as a compute card, though it's been adopted by some gamers as a high-end gaming card. It launched just eight months ago and the company's Pascal GPU architecture is only about one-and-a-half years old, making TITAN V's nine-fold jump in performance and doubling of energy efficiency particularly impressive.

The TITAN V graphics card is the second product that NVIDIA has launched based on its Volta GPU architecture, which it unveiled last May and calls "the world's most advanced GPU architecture." Late in the second quarter, NVIDIA began shipping its Tesla V100 platform for data centers, and fully ramped production of this product in the third quarter.

NVIDIA's Tesla V100 has met with great success. All of the world's major internet companies and cloud-service providers are upgrading their data centers from NVIDIA Pascal-based systems to Volta-based ones. These companies include Amazon, Google, Facebook, IBM, and Microsoft, and the big Chinese players, Alibaba, Baidu, and Tencent. Major server makers are also incorporating Tesla V100 into their products.

"Our vision for Volta was to push the outer limits of high performance computing and AI. We broke new ground with its new processor architecture, instructions, numerical formats, memory architecture and processor links," said CEO Jensen Huang in the press release. "With TITAN V, we are putting Volta into the hands of researchers and scientists all over the world. I can't wait to see their breakthrough discoveries.

https://techcrunch.com/2017/12/08/n...eraflop-gpu-for-ai-processing-bitcoin-mining/

The Titan V features 5120 Cuda cores and additional 640 tensor cores that have been optimized to speed up machine learning workloads.

The card, which features 12GB of HBM2 memory, will retail for $2,999. For that, though, users will see a 9x increase in raw power compared to the Titan Xp, the card’s predecessor, which retailed for “only” $1,299.

It’s worth noting that this isn’t really Nvidia’s most powerful card yet — that honor still belongs to the 112 teraflop Tesla V100, which tends to retail for around $10,000. It is Nvidia’s most powerful GPU for PCs, though. Both the Titan V and V100 cards seem to have quite a bit in common, including the V100 GPU at their core, though the Titan V features both less memory and a reduced memory bus width.

https://www.nvidia.com/en-us/data-center/volta-gpu-architecture/

https://www.nvidia.com/en-us/titan/titan-v/?nvid=nv-int-tnvptlh-29190

Early benchmarks: http://www.pcgamer.com/early-titan-v-benchmarks/

 

[Ron Resnick]

What does it mean?

In your opinion what is the significance of this development?

 

[ack]

Super computing power at your fingertips - if that's what you care for. 25+ years ago, I was developing software on a Cray supercomputer, and long before that, on half a Thinking Machine (32K CPUs - https://en.wikipedia.org/wiki/Thinking_Machines_Corporation though long out of business) during my thesis; both of them were in the millions of dollars and mega structures with elaborate cooling systems, consuming incredible amounts of power. Now, we have so much computing power for a mere couple of grand, consuming about 400W, and still air-cooled.

I see this development enabling AI on our desktops, and with high-speed networking already in place, help us make intelligent decisions - like how to better invest, by deep learning of the numerous factors affecting financial markets, for example. Brilliant minds at Nvidia, no wonder their stock has been doing exceptionally well for years.

 

[audioguy]

That brings back old memories. I was with Control Data in the late 60's/early 70's where Seymore Cray came from. My Apple watch probably has more power in it than the early CDC computers. At the time the Control Data 6600 was announced, it was determined that FIVE of them would do all of the computing we would ever need. We have come a very long way. Very long.

 

[ack]

Control Data - wow, now there's a relic LOL

 

[BMCG]

wow...had not thought of Thinking Machines for quite some time...was a refuge for Course 6 (MIT) types in the mid to late eighties .

 

[ack]

Huh, I found a partner in crime - so you know about the C* language; I worked with one of its creators for a couple of years! Fun stuff!

 

[BMCG]

nah i was Course 10....but had friends/classmates/frat buddies worked there....

although that said did have the privilege of discussing the topic of AI with Marvin Minsky...

 

[GMKF]

It would be interesting to see what AI would do to amplifier circuits

AI is a very powerfull thing and should to be regulated somehow

PS: Nice to see computer history here, I have an intrest in "tube" computers

 

[Bso]

What does it mean?

In your opinion what is the significance of this development?

Positive for audiophiles:
Prices will go down on all CUDA units so we can use HQPlayer and its ilk more effectively.
Streaming, downloading and processing in hardware real-time of true, native high resolution (>96kHZ/24 bit + and DSD files) will become a reality since Interweb bandwidth keeps increasing in lock step and soon will not be a bottleneck.

Negative:
Privacy and security will become increasingly more difficult for the individual.
Governments, NGOs and (social media) companies will collect more and more data on individuals for control purposes.

Humourous:
Repetition: More companies will make claims about the power of AI and machine learning alone that mimic the claims of Thinking Machines and the MIT projects of yore.
More "undesirable" people and the ACLU will clamour for the "right to be left alone."
Individuals will have to develop a public persona and a private identity. This technology should help. Remember, if your SSN is stolen, you can't change it!Use of VPNs, Tor browsers, and so on will be become more and more necessary to protect your private identity.
Like-minded Individuals will create alternate schemes of credit ratings to displace those based on the "big 3", etc.

Just some thoughts....

 

[Ron Resnick]

Which publicly-traded companies are in the lead to achieve quantum supremacy through the development and deployment of quantum computers?