Nvidia Ampere graphics cards: Everything we know so far

Ampere is the successor to Nvidia’s Turing line of RTX 2000-series graphics cards and will power the next-generation 3000 series. These new cards will employ a new, smaller process node, bringing Nvidia’s cards in-line with AMD’s RX 5000 series and should improve performance and efficiency throughout the product stack by sizable margins.

Many details about the cards remain under wraps, but as we edge closer to their expected release date, more and more details have come to light, giving us a good idea of what we can expect from the next-gen line of Nvidia GPUs.

Pricing and availability

Ampere was originally expected to debut in the first half of 2020, but Nvidia later clarified that it would appear later in the year. As of June, we know that pre-production cards have gone out to researchers and cloud service providers, like Google Cloud. Gaming cards don’t have a firm release date, but with AMD’s next-gen RDNA2 cards slated to debut ahead of the new-gen console release (somewhere between October and December), it wouldn’t be a surprise to see Nvidia launch earlier, perhaps between August and September.

Pricing remains under wraps, too, and will likely do so until Nvidia makes an official announcement on models in an unveiling of the initial Ampere GPU lineup. The only Ampere GPU we know that’s in the wild at this time is the AI-focused DGX A100, which is priced at $200,000 — and doesn’t give us any clues about what to expect from mainstream cards.

That said, AMD’s next-gen RDNA2 GPUs are, for the first time in three years, expected to compete directly with some of Nvidia’s most powerful graphics cards, which could mean we see more competitive pricing from the green team than we saw with Turing. It famously increased pricing across the entire product stack, making for a much more costly upgrade path for gamers — especially amidst the cryptocurrency mining craze of 2018.

While it’s possible that the use of a new process node for Ampere card production could allow for some manufacturing cost reductions that will reduce costs for buyers, Ampere GPUs are reportedly very large. That reduces yield numbers for each batch produced, thereby potentially inflating costs in turn.


Nvidia’s Ampere will be based on a brand new process node, dropping from a 12nm TSMC design for the RTX series. What it’s dropping to, though, has yet to be finalized. Initial reports claimed it would be a 7nm EUV process from Samsung, with more recent rumors pointing to TMSC’s 7nm node. The latest rumors in June 2020, however, point instead to Samsung’s 8nm node. Whichever one is used, it should bring improvements in performance and efficiency, with Nvidia’s own design changes adding new features and bolstering performance further.

Ampere will support a third generation of Tensor cores, unlocking faster and more capable deep learning super sampling and potentially additional AI-driven functions. It will support the second generation of high bandwidth memory HBM2 (the expense of that could drive up prices on some cards), as well as GDDR6X, which should offer greater memory bandwidth at a reduced power requirement than GDDR6.

The new architecture will also support PCIExpress 4.0.

Despite its enhancements and changes, Ampere doesn’t appear to have reduced the size of the eventual GPUs. The Ampere-based A100 accelerator employs the largest die of any recent graphics card, even though it is denser, with far greater transistor counts than anything we’ve seen before. Its power draw is also greater, suggesting that Nvidia’s new cards could be more power-hungry than their predecessors, despite using a more efficient design.


Nvidia RTX 2060 Super and RTX 2070 Super review
Nvidia Super graphics cards upped performance in 2019, but Ampere should offer much more. Dan Baker/Digital Trends

With AMD’s rumored big Navi 20 graphics cards expected to offer serious top-tier gaming performance and introduce red team-powered ray tracing to the market, Nvidia is thought likely to drive performance even higher, with Ampere to retain its market lead and performance crown. Outside of the enhancements made with a process node reduction, Ampere has been shown to be capable of employing far greater numbers of streaming multiprocessors (and, therefore, CUDA cores) than its predecessors. The A100 Ampere AI GPU has almost 7,000 of them — a vast increase of almost 2,000 CUDA cores over anything we’ve seen before, even from enterprise-focused cards.

With a new generation of memory and reportedly as much as 24GB of it in the top-of-the-line models, this could see a sizable performance improvement over even the best Turing cards. In a HardwareLeaks report, an unannounced Nvidia Ampere GPU scored over 18,000 points in a 3DMark Time Spy run, making it more than 30% faster than a stock 2080 Ti and only slightly behind that of the heavily modded and overclocked Kingpin edition of the EVGA RTX 2080 Ti.

Further down the product line, WCCFTech has collected some rumored specifications that see the new generation of cards each taking one step up the performance ladder while likely retaining similar, if not improved, pricing. The new-gen RTX 3080 is said to have the same number of CUDA cores as the 2080 Ti and slightly less new-gen memory for comparable, if not slightly improved, performance.

The 3070 Ti is said to be slightly more capable than the stock 2080, perhaps more akin to the 2080 Super, with the standard 3070 taking the place of the RTX 2080.

All of this is highly speculative until we know more, but it suggests that for general gaming power, the Ampere-generation will be more akin to Turing in terms of how it performs compared to its predecessor, rather than the greater leap in capability we saw with 10th-gen Pascal cards.

Cooling overhaul

With the larger dies and potential added power draw of the new-gen cards, it’s not entirely unlikely that Nvidia will need to improve its Founders Edition cooling. While it is always testing out new designs internally, one has appeared in some leaked images that suggests Nvidia may radically overhaul its cooling solution for Ampere GPUs.

Nvidia RTX 3080 Ampere leak

Pictured in a production facility, these coolers are reportedly attached to the RTX 3080, suggesting that they would at least be employed for the high-end Ampere GPUs. They employ a unique cooling and PCB design that has one fan placed at the PCIe bracket end of the card, drawing air upward and driving it across the shrouded heatsink in the center of the card. The PCB ends in a concave “V” shape, allowing for a secondary fan to sit in-line with the PCB and exhaust the hot air above the card.

This is certainly unique, especially for an in-house Nvidia design, and raises a lot of questions about its potential and whether we’ll see anything similar from third parties in the future. Considering the usual layout of modern gaming machines, this would aid in exhausting the hot air rather than blowing it out of the (horizontal) “top” and “bottom” of the card, as many multi-fan designs do today.

Still, while this alleged Nvidia design isn’t confirmed, the only actual RTX 3000-series card that we have seen is a 3080 Ti model from Asus in a report from Videocardz, which employs a much more traditional triple-fan cooling solution.


It’s possible that the fans are higher performance or hide a denser array of heatsink fins underneath to cope with additional heat output, but it does appear that a unique cooling solution isn’t mandatory for Ampere. If a more robust solution is necessary, though, it could drive up material costs and the eventual price of higher-tier GPUs.

What about ray tracing?

As with performance, we don’t have hard details on what Ampere will do with ray tracing, but it seems almost certain that Nvidia will continue to push the new lighting technology’s support in its new-gen cards. It effectively made ray tracing part of the modern gaming conversation with Turing, even if few games support it almost two years after the release of those cards.

One rumor that appeared in May 2020 suggests that Ampere will improve ray tracing performance throughout the GPU lineup by as much as 4x, which would effectively eliminate any kind of bottlenecking when enabling the feature. That could allow top-tier cards to play games at 4K resolution with reasonable frame rates with ray tracing enabled and lower-end cards to take advantage of it without sacrificing higher frame rates at less intensive resolutions.

That would go a long way to making ray tracing more broadly available to gamers.

Editors' Recommendations