The Nvidia Volta GPU tech should form the next generation of GeForce graphics cards. Well, if the stars align and the current technological drought shows some signs of abating. Well, and the strange Nvidia Turing and Ampere rumours prove false. Either way, we know we’re not going to see any new gaming cards launching at either GDC or Nvidia’s own GTC this month. Sad times.
Check out our pick of the best graphics cards you can buy today.
Nvidia Volta release date
Mid-2018 seems like the most likely release for Volta-based gaming cards, because they’re definitely not launching anything in March. But if you can’t wait Nvidia have launched the $3,000 Titan V…
Nvidia Volta specs
The professional-class VoltaAi??ai??i?? the Tesla V100Ai??ai??i?? uses TSMC’s 12nm FinFET design and the full GV100 GPU has 5,376 CUDA cores. Imagine a GTX 2080 Ti with that.
Nvidia Volta price
Don’t shoot the messenger, but it’s possible the consumer Volta cards will push
prices up again. $699 for the GTX 2080? Not beyond the realms of possibility…
Nvidia Volta performance
We’re too early in the release cycle for there to be any performance figures flying around, but we’re hoping for greater efficiency and better with DX12 and Vulkan.
The Nvidia Volta graphics architecture is the next big thing coming out of the green teamai??i??s GPU skunkworks and has had its first professional-level, ultra-expensive processor, the V100 officially unveiled. Volta is the silicon successor to the Pascal generation of graphics cards, the generation that brought us the mightyAi??GTX 1080 TiAi??andAi??Titan Xp, and is gearing up to deliver the GTX 1070 Ti soon. But can it deliver the same generational performance boost offered by Nvidia’s impressive last-gen graphics silicon?
If the gaming performance of the Volta GPUs we’ll end up seeing in our PCs next year provide anything like the performance delta the first Tesla V100 silicon is displaying against the last-gen professional Pascal cards then we’re in for a treat. Volta, in the pro space, is out-performing Pascal by 132%.
As is Nvidiaai??i??s wont the new GPU architecture is taking its name from a famous historical scientist. Alessandro Volta gave his name to the Volt having been a pioneer of electrical energy and its storage. He was also the discoverer of butt-gas – fun little science fact for you there.
But recently there have been the most paper-thin rumours about new Ampere and Turing Ai??codenames being released as the next graphics chips from the Nvidia skunkworks. Both GPU names have come from a single line in two disparate articles surrounding Nvidia financial results, and have no sources to make the names in any way believable. Hell, we can make up spurious codenames in our sleep…
But Volta is a GPU architecture that definitely exists and was first unveiled, at least in theoretical form, at Nvidiaai??i??s Graphics Technology Conference way back in 2013. It was originally meant to be the GPU silicon which followed directly on from the Maxwell architecture (which made up the GTX 900-series of graphics cards), but a year later up pops the Pascal design used in the most recent 10-series GeForce parts, pushing the prospective Volta chips further back.
But Volta now seems to be back on track, with devices appearing in drivers, and in vague rumours from Chinese websites, so letai??i??s delve into what it is, what itai??i??s going to mean for our gaming PCs and when you can get one jammed into your rig.
This is probably the most important question of all about the upcoming Nvidia Volta architecture: when is it going to be launched? Nvidia themselves have been relatively tight-lipped about the whole thing, apart from appearing on oft-changing GPU roadmaps, but weai??i??re starting to see more wider mentions of the technology, whether implicit or specific.
Mid-2018 is now probably the most realistic release date for the consumer version of the Volta GPU, despite the professional GPUs being delivered last year. We had thought that maybe Nvidia would launch something at GDC this year, but our sources within the green team tell us that they are categorically not going to be launching a new GPU in March. Nor will they be launching anything game-related at GTC a week later either – that’s all just deep-learning, AI, and wheely things.
Computex in June maybe? C’mon Nvidia, at least tell us about a new gaming chip at your Graphics Technology Conference. Jen-Hsun surely needs something to talk about at his keynote…
Nvidia did, however, launch the very first desktop spin of the Volta graphics architecture at the end of 2017. TheAi??Nvidia Titan VAi??is a only-very-slightly cut down version of the Tesla V100 card, using the same 5,120 CUDA cores, but with less memory and much lower memory bandwidth.
Considering the amount of computational power available in the new Titan V the $3,000 price tag actually looks pretty reasonable for a pro-level GPU. For a gaming card? Well, not so much.
The performance in-game of the Titan V is probably pretty impressive, given those 5,120 CUDA cores, but as they’re redesigned streaming multiprocessors housing those cores they might need some serious driver optimisation to git gud at gaming.
The extra CUDA core-count, HBM2 memory, and Tensor cores will make it a stunning AI processing GPU for the money, it might not deliver a gaming boost to match. But this isn’t a gamer’s card, so if you’re ready to drop $3,000 on a Titan V make sure you’ve got some proper workloads for it to attack on not just a little light 4K gaming.
But thatai??i??s a card still sporting the top-end, professional-level chip being built for supercomputers like the Summit machine likely to gain sentience at Oak Ridge in 2018 and doom us all. The GV104 silicon, however, is likely to be the consumer-facing version of Volta which makes its way into our graphics cards, and is also probably what SK Hynix were referring to when they announced the new GDDR6 memory standard.
When they announced the arrival of GDDR6 Hynix mentioned they were ai???planning to mass produce the product for a client to release high-end graphics card [sic] by early 2018 equipped with high performance GDDR6 DRAMs.ai??? Iai??i??m not the only one to think that is a reference to Nvidia and the release of GDDR6-based Volta GPUs.
Team red are now looking unlikely to be refreshing their AMD RX VegaAi??GPUs in the new year, so it doesn’t look like they were referring to the Radeon side of things.
Samsung have also announced their own GDDR6 memory is set to “play a critical role in early launches of next-generation graphics cards” which is playing into our suspicions of a relatively early 2018 launch. Though that could just because we’re desperate for a new GPU launch to potentially salve the cryptocurrency crisis that’s stolen all our graphics cards…
There have been many figures thrown around as to just what transistor lithography Nvidia will be jamming into their next-gen Volta graphics cards, with the latest being a return to the Taiwan Semiconductor Manufacturing Company (TSMC) for the manufacturing of 12nm Volta GPUs, having previously dallied with Samsung for their 14nm GP107 silicon.
The Tesla V100 Volta chip is sporting the 12nm FinFET design from TSMC so there’s no reason to think it’ll be any different for the consumer releases.
TSMC have also said their entering volume production of their 12nm FinFET chips in the fourth quarter of this year. That plays into the idea of a full launch of Nvidia’s consumer-facing Volta cards relatively early in 2018.
Originally Volta was supposed to be built using TSMC’s new 10nm process, but the pace of transistor shrinkage has become rather laggardly in recent years. Other rumours had Nvidia sticking with TSMCai??i??s existing 16nm tech in order to be able to stick to their roadmap and get actual Volta cards on the shelves in 2018. It’s now looking like somewhere between the two.
TSMCai??i??s 10nm process, unlike Intelai??i??s Cannonlake design, is expected to be more of a stop-gap measure between their current process and the almost mythical 7nm lithography. To put that into perspective 7nm is about the height of three Tom Cruises standing on each otherai??i??s shoulders. But getting down to 10nm seems to be more of a challenge than maybe even they expected. Iai??i??m guessing either the price is prohibitive or the yields too low to offset the performance benefits of a newer design. Or a combination of both.
TSMC have though announced this stop-gap measure, which theyai??i??re nominally calling 12nm. Itai??i??s apparently based on their existing 16nm design, but with density, performance and energy efficiency improvements. Whether this 12nm node will genuinely be packed with 12nm transistors, or whether itai??i??s just going to be clever marketing, is about as clear as thermal paste.
The current thinking then is Nvidia have now taken this updated 12nm design to use in their Volta GPUs to augment the new architectureai??i??s own performance and power improvements with the slight extra performance bump of a shrunken lithography. With it largely being based on the existing 16nm node there shouldnai??i??t be any concerns about launching a brand new GPU architecture with a brand new lithography. Nvidia have been burnt by those technological flames beforeai??i?? anyone remember the super-heated GTX 480?
We also know the top Volta GPUs will be running HBM2 memory, as detailed by Oak Ridgeai??i??s Summit specsAi??and subsequently in Jen-Hsun’s introduction of the architecture in May. The GV100 is the super-star GPU, sporting both the NVLink interface – instead of PCIe 3.0 – along with the second-gen stacked high-bandwidth memory design. The V100 chip comes with 16GB of HBM2 capable of an incredible 900GB/s of memory bandwidth.Ai??
The HBM2 memory features in the cut-down, PCIe-based GV100 version built for the Volta Titan card, and probably later the Quadro crowd. The strategy for the current Pascal generation, however, has been to run both HBM2 and the NVLink interconnect on the Tesla accelerators with a switch to GDDR5X and PCIe 3.0 for the less-expensive Quadro and consumer variants. I doubt that will change for subsequent Volta cards, with the expected GV102 parts probably sporting the newer GDDR6 technology.
The first consumer-focused Volta GPUs, however, will be the GV104 chips, potentially set to go into the GTX 2080 and GTX 2070 cards. These would be built using GDDR6, allowing Nvidia to pump large capacity, high-performance frame buffers into their cards without the extra expense of HBM2. GDDR6 isnai??i??t as pacey as HBM2 but is definitely speedier than GDDR5X, offering a data rate of 16Gbps as opposed to the formerai??i??s 14Gbps top speed. That also means it comes with a chunk of extra bandwidth too.
Hynix have stated that with a 384-bit memory bus, the sort of design usually favoured by Nvidiaai??i??s high-end graphics cards, they are able to offer memory bandwidth of up to 768GB/s, which isn’t far off the 900GB/s of the V100’s HBM2 design. The new Titan Xp is using GDDR5X and can only manage 548GB/s with its mighty 12GB setup.
With GDDR6 running so quickly it would be unlikely Nvidia would bother with HBM2 for the consumer cards.
In terms of the actual Volta GPU architecture itself there is not a lot to say until we get a little more detail from Nvidia. What their engineers do need to do, however, is to target improved performance from the newer low-level APIs. Compared with AMDai??i??s fourth-gen GCN architecture, used in the Polaris cards, Nvidiaai??i??s Pascal GPUs are generally off-the-pace when it comes to Vulkan and DirectX 12 gaming speed.
Oh hai, welcome to price speculation corner. Obviously we donai??i??t know how Nvidia are going to price their new cards, much of that will probably come down to their relative performance compared with AMDai??i??s high-end Vega parts, but we can still make some educated guesses based on Nvidia passim. And based on the ludicrous pricing of the Titan V.
There was, however, a worrisome section in the MyDrivers.com piece about a rumoured accelerated release for Volta. In it they suggest the GeForce 20-series of products will see re-planning of the prices and market positions which ai???will increase the price of single card to enhance profits.ai??? My understanding of Chinese is pretty woeful (read: nonexistent) so Iai??i??m relying on Mr. Google Robotai??i??s comprehension skills, so itai??i??s possible thereai??i??s a misunderstanding born of the translation into English.
But if true, and the prices of single cards are being adjusted upwards, that would be the second generation in a row where Nvidia have pushed prices skywards. With the 10-series Pascal-based cards the GTX 1080 was released at an unprecedented level, especially when taking the reference/Founders Edition shenanigans into account.
If $699 is going to become the de facto standard for Nvidiaai??i??s high-end graphics cards, as shown with the GTX 1080 Ti cards, then these are worrying times. Unless youai??i??re AMD and confident you can keep undercutting Nvidiaai??i??s GPUs with your own refreshed Vega graphics cards. Gotta get them built in volume first guys…
Although weai??i??ve seen reference to the GV100 Volta GPU in Nvidiaai??i??s graphics drivers, meaning there must be testing going on right now, there havenai??i??t been a lot of interesting and/or apocryphal benchmarks leaking out to whet our appetites. Thatai??i??s probably down to the fact the GV100 is going to be tested on applications which likely bare little relevance to any of the performance benchmarks we might recognise.Ai??
That said, in a pure generation-on-generation test between the pro-level Pascal P100 and Volta V100 GPUs the newer Nvidia architecture is posting performance that is 132% faster than the last-gen chips. If we get anywhere near that level of performance boost from gaming applications Volta will be stunning. Realistically that’s unlikely because these tests are based on Geekbench benchmarks, run in a Linux environment, using the specific CUDA API, and not Shadow of War at 4K.
The professional Volta cards have been tuned especially for AI workloads, with a new Tensor core design at its heart. These new silicon slices won’t have anything to do with gaming, at least for the moment, so if those are the only real improvements with Volta it’s possible there won’t be a huge performance uplift over Pascal at all.
Obviously weai??i??re still expecting some increased gaming performance from Volta, and Iai??i??ve already spoken about the necessity for it to better deal with the low-level APIs of Vulkan and DirectX 12, but we should also expect some improved efficiency born both of the slightly shrunken GPU production process as well as the redesigned architecture itself.Ai??
Given that Volta is the namesake of the Italian gentleman credited with the invention of the battery youai??i??d certainly hope for some improved efficiency. On the notebook side Pascal made great strides forward for the performance of mobile GPUs, and Volta ought to carry that progress even further.Ai??
When he announced the existence of the Volta code-name at GTC 2013 Jen-Hsun Huang explained that “I love that name Volta because it would suggest that it will be even more energy efficient.”
Fingers crossed itai??i??s not just a suggestionai??i??
This page is regularly updated with all the latest news, rumours and information Ai??about the Nvidia Volta GPU architecture. Bookmark away and let us know your thoughts about it in the comments below.
Click on the quick links to get back up to the section of your choice.