What to Expect From Nvidia's Next-Gen RTX 3080 GPUs

It’s been two years since Nvidia brought real-time ray tracing to the mainstream with its GeForce RTX 20-series of graphics cards, and now rumors are swirling about the next generation of consumer-focused GPUs, which could be coming as soon as next week. Here’s everything we (think we) know about (what we assume will be called) the GeForce RTX 3080 and its brethren.

Release Date

Nvidia often launches new GPUs a couple years after the previous generation, so we’re definitely in a reasonable ballpark for an upcoming release. TweakTown predicted an August 2020 reveal of Nvidia’s new Ampere architecture, with a September 2020 launch, a date range that other rumors have backed up. Nvidia’s GeForce Twitter account has also teased an #UltimateCountdown to a big GeForce announcement on September 1 – it doesn’t get much more telling than that. [widget path="global/article/imagegallery" parameters="albumSlug=cyberpunk-2077-pc-ray-tracing-screenshots&captions=true"] Note that based on its previous launches, Nvidia is likely to release its high-end cards, like the rumored RTX 3080, RTX 3080 Ti, or RTX 3090 first – with its less powerful siblings (RTX 3070 and 3060, perhaps?) coming later. The RTX 2080 and 2080 Ti, for example, launched in September 2018, with the RTX 2070 following a month later and the RTX 2060 bringing up the rear in January of 2019. We don’t know if those same dates will hold true this time around, but if you’re looking for a midrange card instead of the roughly $800-$1,300 flagship, it’s likely you’ll have to wait a tad longer than the big-spending early adopters.

Hardware

We expect the new RTX 30-series to use Nvidia’s new Ampere architecture, originally announced earlier this year for use in data centers. Those chips use a 7-nanometer process (down from Turing’s 12nm), allowing for more transistors in a tight space – and thus better performance. Many sources, however, point to Nvidia using Samsung’s 8nm process for its gaming GPUs – or at least certain models – which could be a setback if AMD keeps TSMC’s superior 7nm node, which is also present on the PS5 and Xbox Series X. We’re expecting a bump in memory, too, with the RTX 3080 rumored to have up to 20 GB of GDDR6X, compared to the RTX 2080 Ti’s 11GB. That’d give a nice boost to gamers running 4K, but keep in mind those sources are saying “up to,” so don't be surprised to see less on some or all variants – especially since other sources are sticking to a much lower 10GB number. (Aren’t rumors fun?) [widget path="global/article/imagegallery" parameters="albumSlug=best-4k-gaming-tv-for-ps5-and-xbox-series-x&captions=true"] Apart from the typical -80, -70, and -60 tiers, we may also see a flagship RTX 3090 card that sits above the RTX 3080 in performance, with up to 24GB of GDDR6X memory. Another possibility is some badass new cooling on the Founder’s Editions of these cards, if leaks are to be believed – check out the serious heat sinks under that shroud! That makes sense, given rumors that one or more of the cards will be so power-hungry that it’ll use a 12-pin power connector, which most current power supplies don’t even have. Another recent leak gave us a look at the RTX 3090’s overall size compared to an RTX 2080, and it’s a big boy.

Performance

If you’re looking for the best performance PC gaming has to offer – chasing 4K 144Hz, perhaps? – you’ll want to take any early numbers you hear with a grain of salt. Some are saying the RTX 3080 performs 20% better than an RTX 2080 Ti, with the RTX 3090 showing a 50% improvement over the RTX 2080 Ti in 3DMark Time Spy Extreme – a benchmark that may not even be showcasing the full potential of these cards. AMD’s Big Navi, a competing high-end card also rumored to release this September, may boast similar improvements, however, so we could be in for a compelling race this time around. And when AMD and Nvidia compete, gamers win. PC graphics cards won’t be the only competition for Nvidia, of course, given that new consoles are also on the horizon. While the PlayStation 5 and Xbox Series X’s performance are still shrouded in some mystery, the manufacturers are claiming they’ll be capable of 10.28 and 12.15 teraflops of compute power, respectively. That’s less than the RTX 2080 Ti’s 14.2 teraflops, and if price is no object in your pursuit of smooth and crisp gaming you'll want to keep an eye out: the GA102 chip we’re expecting to see in the RTX 3090 is rumored to run at 21 teraflops, fully doubling the PlayStation 5. [youtube clip_id="ZEzI02SKALY"] Of course, teraflops are an admittedly weak value for comparison since that metric doesn’t even take into account additional tensor and RT cores, differences in GPU architecture, or the fact that the consoles will likely clock down much of the time. We’ve seen plenty of times that raw teraflops don’t tell the whole story, so again: fill your pockets with big grains of salt. Regardless, it's still safe to say consoles are likely to be at a pretty deep performance deficit compared to Nvidia’s top-tier cards right from the start. Either way, we also hope to see a boost in ray tracing performance in the RTX 30-series, with some rumors pointing to a 3-to-4x improvement compared to last gen thanks to more FP32 units and tensor cores, not to mention advances in VRAM efficiency and features like DLSS 3.0. We’ve already seen ray tracing go from performance-killing to maybe-almost-okay depending on the game and hardware in question, so further improvement could make the tech much more compelling. Again, though, all these are average numbers based on rumors, so don’t put too much stock in them (or Nvidia’s marketing numbers) just yet. Real-world benchmarks will tell the full story once we have our hands on these bad boys.

Price

Of course, performance doesn’t mean much without a price attached, especially once you put it next to competing cards from other manufacturers. We don’t yet know what Nvidia’s new GPUs will cost, but we can look to past launches to make some flimsy predictions. When Nvidia released its RTX 20-series, the RTX 2080 was priced at $799, with the RTX 2080 Ti costing a whopping $1,199. We wouldn’t be surprised to see similar prices this time around (with the RTX 3070 and RTX 3060 cards closer to $599 and $349, respectively), but it’s always possible Nvidia could hike prices a little higher the way it did when those Turing cards launched, which were more expensive than their Pascal predecessors. Of course, competition from AMD could counter that if their RDNA 2 cards are powerful enough. As for that RTX 3090, current rumors suggest a holy-moly price tag of $1,399 – and that’s a starting price, which means it could be even higher for the Founder’s Edition (rumored to be $1,499). Other custom variants could cost even more, so if you want the best of the best...well, I hope you have a money tree growing in your backyard. [poilib element="poll" parameters="id=ed7dc712-0779-4b08-ac25-fc6614183eee"]