NVIDIA announced its latest GPUs, the GeForce RTX 30-series, based on its ‘Ampere’ architecture. The company revealed the price and release date for the RTX 3070, 3080, and 3090, with the first of the cards launching in just two weeks.

In a pre-recorded presentation streamed today, Nvidia revealed the GeForce RTX 3070, 3080, and 3090 GPUs based on its latest ‘Ampere’ architecture. Before we dive into details, here’s the release date, price, and basic specs for the ‘Founder’s Edition’ of each card:

RTX 3090 RTX 3080 RTX 3070
Price $1,500 $700 $500
Release Date September 24th September 17th October
CUDA Cores 10,496 8,704 5,888
Boost Clock (GHz) 1.7 1.71 1.73
Memory 24GB (GDDR6X) 10GB (GDDR6X) 8GB (GDDR6)
Connectors 1x HDMI 2.1, 3x DisplayPort 1.4a

Succeeding the ‘Turing’ architecture of the prior generation, ‘Ampere’ aims to deliver more performance across three pillars: shading, ray tracing (RTX), and AI (DLSS). The company claims the new graphics cards represent the “greatest generational leap” in performance from one series of Nvidia GPUs to the next.

Image courtesy NVIDIA

For games that make use of those features, Nvidia claims up to 2x performance and 1.9x power efficiency between 20-series and 30-series cards. For games that don’t use features like raytracing, the gains don’t quite reach 2x performance, but still offer a nice jump, according to the company.

Nvidia also announced ‘RTX IO’ a new accelerated storage pipeline designed to offload game asset data decompression from the CPU to the GPU, accelerating performance by “up to 100x compared with hard drives and traditional storage APIs.”

“When used with Microsoft’s new DirectStorage for Windows API, RTX IO offloads dozens of CPU cores’ worth of work to your GeForce RTX GPU, improving frame rates, enabling near-instantaneous game loading, and opening the door to a new era of large, incredibly detailed open world games,” the company says.

Image courtesy NVIDIA

RTX IO will require an NVMe SSD and an RTX GPU (as far as we understand, this includes existing 20-series cards as well as the upcoming 30-series).

Image courtesy NVIDIA

While the first wave of the prior RTX 20-series GPUs included USB-C ports with support for the VirtualLink standard, so far it looks like the first wave of RTX 30-series will eschew a USB-C port. Whether the piping is there for third-party manufacturers to add USB-C ports (and therefore VirtualLink support), isn’t clear at this time but we’ve reached out to Nvidia for clarification.

The company also introduced a new technology called Nvidia Reflex which aims to reduce game latency. The feature needs to be specifically built into applications, but Nvidia says “by integrating directly with the game, Reflex Low Latency Mode aligns game engine work to complete just-in-time for rendering, eliminating the GPU render queue and reducing CPU back pressure in GPU intensive scenes. This delivers latency reductions above and beyond existing driver-only techniques, such as NVIDIA Ultra Low Latency Mode.”

While the company is highlighting Reflex’s ability to reduce latency for competitive non-VR esports games, it said in a comment on its site that “in theory, it could improve latency in VR games, however we have not tested it.” We’ve reached out for more details.

Newsletter graphic

This article may contain affiliate links. If you click an affiliate link and buy a product we may receive a small commission which helps support the publication. More information.

Ben is the world's most senior professional analyst solely dedicated to the XR industry, having founded Road to VR in 2011—a year before the Oculus Kickstarter sparked a resurgence that led to the modern XR landscape. He has authored more than 3,000 articles chronicling the evolution of the XR industry over more than a decade. With that unique perspective, Ben has been consistently recognized as one of the most influential voices in XR, giving keynotes and joining panel and podcast discussions at key industry events. He is a self-described "journalist and analyst, not evangelist."
  • TechPassion

    Who wants my 2080 RTX for 500 EUR?

    • Smelina

      I feel you pain. Good luck with selling the 2080 for at least 300 Eur.

    • Bob

      Some people are selling the 2080Ti at that price so good luck and godspeed buddy with that ridiculous second-hand price.


      • geogan

        You think that’s bad… in last two days loads of 2080Ti have gone up on second hand sites in my country and they are STILL looking for €700 ($829) to €1,300 ($1,540) for them. Crazy. People are replying laughing at them.

    • Mettanine

      Try selling it in Europe. I’m still seeing the same ridiculous prices across the board (> 800€) for used 2080 cards.

    • Andrew Jakobs

      haha, I guess you’re sarcastic.. You will not even get rid of a RTX2080ti with that price.. Wonder how stores who have several of those still in their inventory will get rid of them because nobody in their right mind is gonna buy those for the prices they went for last week..
      Personally I’m gonna wait and see what the RTX3060 is gonna be, it’s supposed to be around $399.. I’m still quite happy with the RTX2060Super I bought last year.

      • Darshan

        Andrew i am with you for RTX3060. herd its going to have 2070 super lever of power which is very similar to 2080, also its going to be RTX supported with new tensor cores. so its worth shot.

    • Loveless

      I bought it For 350 EUR , even it was Top of it’s line, Asus Rog Oc version.

  • Rekriux

    Not so sure, it seems like more features require specific implementation of nvidia sdk/library. It’s like they keep going against open standard and vulkan… Do they realy want restrictive features only available to then in games/app ? That’s not a commercial behaviour I would endorse.

    • Adrian Meredith

      Initially that’s true but most of rtx is standardised as part of direct X ultimate. Latest Vulcan had equivalents too. The fully path traced quake 2 rtx is open source and runs on vulkan. Dlss is proprietary though and that’s probably the best part.

    • Nopeload

      I agree, but isn’t directstorage a windows function?

    • kontis

      AMD promotes their GPU Open solutions.
      The fact they are usually inferior to what Nvidia offers is not a coincidence.
      There is a popular joke in IT that if your competitor is quicker to market and has better solution just make your alternative open source or less proprietary, because you don’t have much to lose and you need that advantage to attract devs and customers.

      Harsh reality is that Nvidia’s ecosystem is now light years ahead of AMD, so all AMD has now to be open.
      At least Nvidia thanked them for Mantle that became Vulkan…

      It’s great that nvidia still innovates, but if they become GPU monopoly for much longer there will be a time when they become complacent. Good thing Intel entered the industry, We need real competition.

  • Adderstone VR

    So VirtuaLink is dead? Valve Index was the only headset to even mention it, they just mentioned they dropped the adapter idea. So the VR industry conned Nvidia and all the partners to include this port on an entire generation of cards and not a single peripheral used it, I’m very disappointed.

    • TechPassion

      Yes. totally IDIOTIC, acutally should be a crime, what the manufacturers did. That VirtuaLink was a slim cable, fast transfer. It was such a good idea to do it this way and not steal HDMI/DP slots.

      • IcedForce

        Too bad it just didn’t work.
        AFAIK Valve was only one who promised a VirtualLink for Index as an option but they couldn’t get it work (too much noise in too weak signal and so not at all reliable). Most likely others ended up with same problems because VirtualLink was pushing the limits of the USB-C (almost nothing else uses all of the pins) and even then taking it to extremes with 3-5m cables and promising them to be passive (with perfect signal maximum length of passive USB cables are 2.0: 5m, 3.x: 3m and 4.x: 80cm), as in cheap compared to active solutions today (with boosting voltage and adding extra hardware behind for error correction and stuff you can get the signal go a lot farther but it gets expensive).

        Promised way too much, way too soon and got pushed out to consumers way too early and the end result is; No one can deliver it because the promises were too great.

        • TechPassion

          I see. Thanks for clarifying the topic.

          • Valve were timely in getting in touch with pre-order customers for the Index Virtualink cable, I got a prompt refund and $20 store credit as goodwill gesture which was a positive counter to a disappointing outcome.

        • Nopeload

          Will 4.0 help things?

          • IcedForce


            USB4.x is mostly update to harmonize the USB-C connector jungle (same connector but the HW behind can be quite different ie. HDMI vs. DP, Thunderbolt support and the dumpster full of USB-C headphones that work only with one manufacturers some phones) and to bring some features bind to the USB-C connector (fast charging aka. USB-PD, dual-link and probably something else that I don’t remember) so every device using USB4.x USB-C will have those features.

            The baked in dual-link is the one thing that would make VirtualLink possible because it doubles the data-rate of the USB3.x but that also means more possibilities for interference (more data moving through more wires -> more current moving and more magnetic stuff going on). The downside is that while that is the maximum datarate of the USB4.x, the maximum length of a passive cable for that is only 80cm because those interferences. When it’s really out and cables start coming out you probably will see longer than 80cm (1-1,5m is probably usual) USB4.x rated for the full “USB3.x dual-link” speed but most likely they will be expensive (either extremely well shielded) or bogus, the third ones will be the active cables which are already expensive AF (USB3.0 10m active cables go for +50€/$ easily, even 100 coins isn’t that rare) and with more data moving they will need to get better components (active means they have build-in signal boosting, error correction and all that fancy stuff that requires microchips and stuff) to support those datarates and that is going to be expensive.

            I bring the expensiveness just to point out that you really cannot just make 5-10m USB4.x active cable that supports the full datarate and just put it in a box. I can easily see something like 5m USB4.x full-datarate cable going for 150-200 coins, at first, so using less datarate and other options is going to be cheaper for a while. This actually is the reason why Quest Link Cable isn’t included in the box because it really costs that ~90 coins (89€ from Finland) because it is active cable (optic in this case which is even more expensive than just copper and boosting the voltage but far less prone for interference, which really shouldn’t be a problem in that 5m length yet but anyway) (also notice on this that for Quest use they recommend max. 3m cables or their own 5m which is active, 3m being the theoretical max. for USB3.x passive cables, and that they are doing a lot of stuff within the HMD instead of throwing it all to the PC through that same cable).

        • Octo

          Odd because the XTAL headset which is easily one of the most high-end headsets out there used it and the developers called it “reliable and convenient interface and communication protocol”

          • IcedForce

            The big questions are what was promised and the price. If Valve refunded $20 for not delivering VirtualLink they were really promising it to be passive and that is most likely the problem. The price is the reason why Quest doesn’t include the Link Cable, 70-90€/$ is realistic price for an active USB3.2 cable.

            XTAL doesn’t operate in consumer space (at least what I know) and so their HMD is probably priced so the 50-100 extra for a cable in it isn’t that noticeable and their customers probably don’t whine about it. Like +100€/$ to the price of Quest and general consumers would go mad because they wouldn’t understand that the cable is actually expensive, but +100€/$ to >1000€/$ HMD that is mainly sold to companies, no one even bats an eye because cables can be expensive and quite many of the are expensive.

        • Charles

          Too specific to VR for something as significant as a special port on the GPU, when VR is still a niche market.

          • Renard

            This could be used to connect displays, too, though

    • psuedonymous

      Valve promised the card-end of VirtualLink but did not implement it. Rift S implemented the HMD-end of VirtualLink (the connector inside the Rift S is an OCuLink connector) but never the card end unified connector. The VirtuallLink Consortium page vanished some weeks ago.

    • Sven Viking

      It’s sort of a chicken and egg problem unfortunately. The port needs to be available on the vast majority of supported cards for anyone to bother using it, but it never will be because nobody was using it. :(

    • Yes, it died the exact moment even Valve abandoned it

  • Nopeload

    I think the Reflex program will be less interesting in VR since you can’ already change your steam input bindings to do things like make the trigger actuate sooner. It matters outside of VR since they have gaming mice and just click. I’m more interested in directstorage, allowing assets to stream in straight from your SSD and saving CPU load on large scenes which would make higher framerates and more complex scenes in VR work a lot better. Not to mention just the cost viability now where I can get a 3080 and have a great experience with max super sampling and 144 in games without CPU limitations.

  • xyzs

    A 3080 is gonna be adopted soon :)

  • AwakeBopppppps

    So yeah 3070 will be 500$ in usa i guess but in czech will be for 700$ atleast

  • Fireking

    Should I get the 3070 I have a 2060 super ring now, just wondering if the 3070 is good upgrade or are they close in prominence

  • Outing Nazis since Sth Africa

    Way too expensive when people have so little money nowadays, at least in the collapsing US.