NVIDIA unveils new RTX 4070 Ti, available January 5

The GPU is priced at $799 (USD) and is 'faster than the RTX 3090 Ti at almost half the power' according to NVIDIA.

14

NVIDIA recently unveiled its newest mid-range GPU in the form of the RTX 4070 Ti with a price point starting at $799 (USD), a bit more reasonable than its RTX 4090 graphics cards priced closer to $1,600. The RTX 4070 Ti is also noticeably cheaper than the RTX 3090 Ti and according to NVIDIA is “faster than the RTX 3090 Ti at almost half the power” thanks to innovations like NVIDIA’s Ada Lovelace architecture and DLSS 3.

Rather than target 4K resolutions like the RTX 4080 and 4090 Ti, NVIDIA’s new RTX 4070 Ti focuses instead on 2K resolution (1440p). For example, NVIDIA notes that the RTX 4070 Ti will “max out your 1440p monitor” in addition to delivering over 120 FPS in games like A Plague Tale: Requiem.

A graph from NVIDIA showing how the RTX 4070 Ti is faster than the RTX 3090 Ti in games like Cyberpunk 2077
© NVIDIA

Elaborating further on the power of the RTX 4070 Ti, NVIDIA touches on how 1440p is a “sweet spot for many enthusiast gamers” and how the card is perfect for these PC gamers in particular thanks to features such as its 7680 CUDA Cores and 12 GB of GDDR6X memory. Additionally, the RTX 4070 Ti is described as a great upgrade for those currently using the GeForce GTX 1080 Ti or RTX 2080.

Not only does the RTX 4070 Ti use DLSS 3 technology, it also reportedly delivers “an incredible 12x relative performance upgrade over the legendary GeForce GTX 1080 Ti” as well. If all of this information has you itching to get your hands on a RTX 4070 Ti, the cards will be available starting January 5. To learn more about NVIDIA’s new RTX 4070 Ti GPUs, be sure to read through the full announcement post from NVIDIA.

And for more in the realm of gaming technology, check out some of our previous coverage as well including how the latest Ford F-150 Lightning EV price hike drives the starting cost up 40% since launch, and how China is preparing a $143 billion fiscal package aimed at self-sufficiency in chip production.

Senior Editor

Morgan is a writer from the frozen wastelands of Maine who enjoys metal music, kpop, horror, and indie games. They're also a Tetris fanatic who's fiercely competitive in games like Tetris 99... and all games in general. But mostly Tetris. You can follow Morgan on Twitter @Author_MShaver.

From The Chatty
  • reply
    January 3, 2023 11:30 AM

    Morgan Shaver posted a new article, NVIDIA unveils new RTX 4070 Ti, available January 5

    • reply
      January 3, 2023 11:34 AM

      Targeting 1440p which is available on loads of $250 144hz gaming monitors - and the card is $800 lol.

    • reply
      January 3, 2023 12:48 PM

      In 2010, the Geforce 570 was released at an MSRP of $350. 6 years later, the Geforce 1070Ti was released at an MSRP of $399. And then another 6 years later, the Geforce 4070Ti is released at an MSRP of $799.

      The FPS gains between the 570 and the 1070Ti are also similar or greater than the projected gains between the 1070Ti and the 4070Ti.

      Really feels like they're taking us for a MAXPROFITS-fuck-the-consumer ride. I guess this is what happens with a monopoly/duopoly.

      • reply
        January 3, 2023 12:58 PM

        Agreed on price inflation for sure, but the transistor count is very nonlinear too and is the primary cost driver (along with R&D). If it were cheap to make the damn things AMD or Intel would be making market ground on nvidia in the discrete space, but they're not. So buckle up prices are not going down soon.

        GTX 570- 3B
        GTX 1070 Ti - 7.2B
        RTX 4070 Ti - 35.8B

        For reference:
        PS4 Pro GPU Transistors - 5.7B
        PS5 GPU Transistors - 10B

        pixels are expensive these days.

        • reply
          January 3, 2023 1:01 PM

          Eh, I don't think that matters much. They were making money on a $700 3080. Now they're making A LOT of money on a $1400 4080.

          • reply
            January 3, 2023 2:40 PM

            And they made a 12gb 3080 to make even more money.

        • reply
          January 3, 2023 4:00 PM

          If what you say is fair. Then maybe let's compare price per transistor. I'll do dollars per billion transistors just to keep the numbers readable.

          GTX 570: $116
          1070ti: $55.4
          4070ti: $22.3

          Math checks out. Would be interesting to see the historical trend of price per transistor

          • reply
            January 3, 2023 4:41 PM

            Its a super mix of how much TSMC charges for a wafer for each size process shrink which makes it more expensive, but when it shrinks you fit more on a wafer which makes it cheaper, but you also get worse yields on smaller process which makes it more expensive.

        • reply
          January 3, 2023 8:21 PM

          Transistor counts for Apple Silicon has been steadily increasing YoY on more cutting edge process by at least a year compared to Nvidia and AMD, yet you don't see price inflation to that same degree with iPhones, iPad, or their laptops, even as they add other things like better cameras, LIDAR, etc.

          Current GPU prices are affected by inflation and increased manufacturing costs for sure, but they're also driven (by Jen-Hsun's own statement to investors way back last summer) by a glut of 2020/2021 inventory that needs to be cleared out.

          The primary driver is oversupply of old stock, lest they crash the GPU market and completely fuck over their AIBs if they were priced in line without that as a factor.

          • reply
            January 3, 2023 8:49 PM

            True, but Apple also does 10x the volume of nvidia and has more pull with TSMC.

            Also phones have been creeping up in cost every year as well, and you’ll find lots of people here complaining about it.

            • reply
              January 3, 2023 9:24 PM

              For sure, and I hope I was clear that price inflation also happens with their phones and tablets. It just happened with the M2 Macbook Air. The degree is very different though.

              The issue with Nvidia is that the oversupply of Ampere stock created by massive overproduction in late 2021 running into a collapse in crypto. This had a massively distorting effect on new GPU prices.

              It bears repeating that Jen-Hsun himself told investors months before the official announcement of Lovelace GPUs that they would be priced in order to move Ampere stock without collapsing their prices, and that he didn't expect Ampere stock to be cleared enough from channel until around January 2023. Its also worth noting in the marketing slides when they announced 4090, 4080 16GB, and 4080 12GB (lol), that Ampere cards were on the same slide with their prices moving down accordingly. Its also worth noting that the price per frame was mostly linear across Ampere and Lovelace, pretty unprecedented in the GPU space.

              It would be as if Apple released the iPhone 14 Pro at $1400 instead of $1000 while continuing to sell the iPhone 13 at $1000 because they somehow allowed retail inventory to get that out of their control.

        • reply
          January 3, 2023 9:28 PM

          what about die size & nm lithography size? more chips per wafer = profit.

    • reply
      January 3, 2023 12:49 PM

      Hmm does this "faster than a 3090 Ti" include frame generation? Because if so... meh. If it is faster without frame generation at the same native res, then it would certainly be a much better deal than the 4080.

      https://static.techspot.com/articles-info/2569/bench/CP2077_1440p-p.webp

      • reply
        January 3, 2023 2:42 PM

        Whenever Nvidia says "our 40 series GPU is faster than X" it's in DLSS 3. I think it's about the same as a 3090 Ti in raster, which is pretty good for the price/size.

        • reply
          January 3, 2023 4:42 PM

          If it is on par with 3090 Ti raster for 40% of the price that is pretty good. Of course the 3090 TI was a TERRIBLE value.

          • reply
            January 3, 2023 4:45 PM

            3090 is an absolute hall of fame trash card, at least as something sold to gamers

          • reply
            January 3, 2023 9:01 PM

            Another way to look at it is 125% of the performance of the 3080 at 115% of the price.

            So pretty sub-par for generation-to-generation price/perf improvement

            • reply
              January 3, 2023 9:26 PM

              Yup. Bears repeating that roughly a 1:1 price per frame is pretty unprecedented. I don't think Turing, underwhelming as that was, was as bad.

              The difference here ofc is that they actually went fucking crazy with the 4090 and delivered a beast of a card, even if its price per frame is in line with 2020 GPUs.

        • reply
          January 3, 2023 7:42 PM

          DLSS 3 means frame generation? Blah!

          • reply
            January 3, 2023 7:56 PM

            Probably depends on the game

          • reply
            January 3, 2023 8:28 PM

            I'm kind of surprised they don't even have it as an asterisk or fine print here,
            https://www.nvidia.com/en-us/geforce/news/geforce-rtx-4070-ti/

            I find that pretty misleading.

            With the 40XX series cards, whenever they say "DLSS 3" when referring to performance I assume they're using frame interpolation.

            • reply
              January 3, 2023 10:20 PM

              yeah gamersnexus calls that out in their vid. also agreed, no asterisk is sussy.

Hello, Meet Lola