Not long after Nvidia released the high-performance GTX 1080 Ti, the rumour mill is already picking up the pace for cards based on the next-generation GPU, codenamed Volta.
Those rumours started a couple of years ago, but at its GPU Tech Conference in May, Nvidia officially unveiled the first chip based on the new technology, called the Tesla V100.
Designed for data centres, the V100 is uninteresting to gamers, but it does point the way towards the consumer cards. No doubt these will be with us before too long.
If you’re looking to buy new card right now, check out our roundup of the best graphics cards.
What is Nvidia Volta?
Volta is the name for the microarchitecture – a design which will be used as the basis for the next generation of consumer graphics cards (as well as the workstation and enterprise stuff).
The current generation, including the Titan Xp and GTX 1080-1050, are based on Pascal architecture. Prior to that, it was Maxwell.
Originally, the plan was to go to Volta after Maxwell, but for whatever reason Pascal has arrived in the interim.
When will consumer Volta cards arrive?
That’s a question with no definitive answer right now. Some rumours point to a 2017 launch, which could theoretically happen, but the smart money is probably on a launch early next year because of comments by memory manufacturer SK Hynix (see below).
When Nvidia unveiled the Tesla V100, it showed only a prototype: this isn’t a product you can buy yet.
It uses the GV100 GPU, and it’s likely the consumer graphics cards will have a GV104 chip. Again, there are rumours that they’ll be called the GTX 2080 and 2070. This would be easier to understand than incrementing by 100 as ranges have done for the last few years.
Otherwise the successor to the GTX 1080 would be the GTX 1180.
What about performance and specs?
It’s way too early to say if Volta will deliver the kind of significant performance gains that Pascal did over Maxwell.
It’s not even clear what process the GPU will use. Intel is moving to 10nm for Cannon Lake later this year, and that was the plan for Volta.
However, shrinking the transistors is proving difficult (part of the reason why Intel is sticking with 14nm for four whole generations of Core processors) and there’s a chance that Volta will end up using a 12nm process from TSMC – the company that makes the chips for Nvidia.
Rumours have also surfaced about the kind of memory the cards will use. Memory manufacturer SK Hynix published a press release in April which “introduced the world’s fastest 2Znm 8Gb(Gigabit) GDDR6(Graphics DDR6) DRAM. The product operates with an I/O data rate of 16Gbps(Gigabits per second) per pin, which is the industry’s fastest. With a forthcoming high-end graphics card of 384-bit I/Os, this DRAM processes up to 768GB(Gigabytes) of graphics data per second.”
It didn’t specify which manufacturer that graphics card was coming from, but with AMD’s Vega expected to use HBM2 memory and the press release also stating that “SK Hynix has been planning to mass produce the product for a client to release high-end graphics card by early 2018”, you’d assume it was the next-gen GeForce cards.
Nvidia’s consumer cards have all used GDDR memory previously, and while the Tesla V100 will use HBM2 (and possibly Quadro and the next Titan X card) we expect the GeForce cards based on Volta will stick to this tradition.
The Titan Xp with 12GB of GDDR5X has a memory bandwidth of 547.7GB/s, so it’s interesting that SK Hynix’s GDDR6 will run at 16Gb/s and – with a 384-bit interface card – will offer up to 768GB/s. That’s quite a jump: the Tesla V100 – with HBM2 – has a bandwidth of 900GB/s, so a putative Volta-based Titan with GDDR6 wouldn’t be too far behind.
We’ll be following the development of Volta closely, and updating this article when any new rumours (and facts) about the consumer cards surface, so keep checking back.