Nvidia has announced the Titan V, the "world's most powerful PC GPU".
It's based on Nvidia's Volta, the same architecture as the Nvidia Tesla V100 GPUs behind Amazon Web Service's recently launched top-end P3 instances, which are dedicated to artificial-intelligence applications.
The Titan V for PCs is also aimed at researchers and scientists and was unveiled at the big Neural Information Processing System (NIPS) AI conference on this week in Long Beach.
The graphics card consists of six graphic processing clusters, featuring 21.1 billion transistors, 12GB of memory, and 640 Tensor Cores, which are capable of performing 110 teraflops, or a 110 trillion floating-point operations per second.
That performance trounces one of Nvidia's best gaming GPUs, the Maxwell-based GTX 980, which is capable of 4.6 teraflops, as well as the 11 teraflops that the newer Pascal-based GTX 1080 Ti is capable of. At $2,999, it also costs about six times as much as those top-of-the-line gaming GPUs.
Nvidia says the Titan V has nine times the power of its predecessor, the $1,200 Titan Xp, and double the energy efficiency of Pascal GPUs. The gear is fabricated on TMCS's 12-nanometer FFN manufacturing process.
"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 Nvidia founder and CEO Jensen Huang.
"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."
The Titan V has a similar form factor to its gaming GPUs but comes in black and gold. As with the Titan Xp, each customer can only purchase two units.
Nvidia is also making its recently released AI software packages available via its GPU cloud.
The growth of AI and adoption of GPUs in the cloud to support AI applications have been a major win for Nvidia. Gaming is still Nvidia's biggest market by revenue but data center income from the likes of Microsoft, Baidu and AWS have seen the sector's revenue more than double over the past year to $500m.
Nvidia's video details the Titan V specification.
Previous and related coverage
AWS has released its most powerful and expensive instance aimed at AI developers.
With more than 500 high-performance computing applications that incorporate GPU acceleration, Nvidia is aiming to make them easier to access.
It's been decades since AMD and Intel last collaborated, but pressure from Nvidia has the two chip giants teaming up once again.
Read more on Nvidia
- Nvidia looks to reduce AI training material through 'imagination'
- Nvidia, Nuance team up to bring AI, machine learning to radiologists
- GE Healthcare turns to Nvidia for AI boost in medical imaging
- Nvidia's data center business hits $2bn annual revenue run rate
- Google Assistant adds TV voice control to Nvidia Shield (CNET)
- NVIDIA GPU Cloud 'democratizes AI' for developers on AWS (TechRepublic)