Nvidia Titan V Launched for AI and Deep Learning, First GPU Based on 'Volta' Architecture

While the Pascal architecture that powers Nvidia's GeForce 10-series gaming GPUs is far from showing any signs of age, the company has just released its first consumer product based on the next-gen Volta architecture. However, gamers still have to wait a while - the new Nvidia Titan V is a powerhouse designed primarily for machine learning, big data analysis and artificial intelligence development, and is priced at $2,999 (approximately Rs. 1,93,300 before taxes and duties). 

Nvidia released its Volta-based Tesla V100 data centre GPU in the middle of 2017, and has announced that the Volta architecture is being used for GPUs in several upcoming supercomputers for research facilities. The Titan V is a standard PCIe card, with a GPU codenamed GV100, manufactured on a new 12nm process. The GV100 has 21.1 billion transistors and can achieve throughput of 110 teraflops in optimised workloads, which Nvidia says is a massive 9X leap over its previous architecture.

Nvidia calls its new compute units Tensor Cores, and they are designed specifically for deep learning, with independent integer and floating-point data paths that can operate in parallel. There are 640 Tensor Cores and 5120 CUDA cores, compared to 3840 on the previous-gen Titan Xp. The GPU base clock speed is 1200MHz and it can boost up to 1455MHz. 

The Titan V has 12GB of stacked HBM2 memory which operates at 1.7Gbps on a 3072-bit bus, for a total bandwidth capacity of 652.8GBps. 3D performance should be a bit better than that of the GeForce GTX 1080 Ti's (Review) but the massive price premium will deter gamers as well as cryptocurrency miners. Hopefully though, this is still an indication of the performance we can expect when Nvidia decides to refresh its gaming GPU lineup.

Nvidia has given the Titan V a distinct black and gold paint job, but it's still a dual-slot PCIe card with the company's now-familiar angular stock cooler. TDP is rated at the same 250W as previous Titan models, and you need only one 8-pin and one 6-pin PCIe power connector. Given the drastic rise in processing power, it's clear that Nvidia has also managed to improve power efficiency thanks to the new manufacturing process as well as its architecture. There are three DisplayPort outputs and one HDMI port on the backplate. 

Data scientists can pick up a Nvidia Titan V from the company's website as of today, with a limit of two per customer. Like previous Titan models, it is unlikely that third-party manufacturers will offer their own graphics cards with this GPU. Pricing and availability in India are not yet known.