Rumor: Nvidia to answer Radeon RX 550 with GeForce GT 1030

Nvidia and AMD have traditionally gone back and forth creating new products and tweaking existing models to position one card or another at just about every conceivable price point. AMD has been absent in the high-end space despite Nvidia's launches of the all-conquering GTX 1080 and GTX 1080 Ti, but the red team has been striking first at lower price points of late. To wit, AMD drew first blood in the more budget-friendly end of the market last year with the RX 480 and the related RX 470, while the RX 460 made it to market before the GTX 1050 and GTX 1050 Ti competition. With this background, it should come as no surprise that whispers about a possible GeForce GT 1030 begin to circulate just as cards based on AMD's teeny-tiny Radeon RX 550 begin to hit the market.

Speculators can't seem to agree on whether the rumored graphics card will rely on a further cut-down version of the GP107 silicon found in the existing GeForce GTX 1050 and GTX 1050 Ti, or on a new chip possibly called GP108. Whatever the chip, the sources assume that buyers will get 512 stream processors exchange for their money. As for the memory packaged along with such a video card, 1 GB or 2 GB seem like the most logical amounts.

Video memory speed has been a sticky point for bottom-rung cards for decades, and we see little reason to expect a high-end configuration for the rumored GT 1030. Nvidia's current offering in the low-end territory, the GeForce GT 730, is offered in configurations with GDDR5 RAM on a 64-bit bus, as well as plain DDR3 on 64- or 128-bit buses. That's not a typo—some of those cards use the same chips present in system memory. The GeForce GT 730 cards with DDR3 memory on a 128-bit bus are based on the 40-nm GF108 "Fermi" chips originally used in 2010's GeForce GT 430.

AMD's Radeon RX 550 and RX 540 appear to be aimed at buyers who want the least cost and power consumption possible while enjoying some of the performance of a modern discrete graphics card. If Nvidia does release a sub-GTX 1050 desktop graphics card based on its Pascal architecture, the existence of mobile variants seems like a foregone conclusion. The green team has been particularly aggressive about endowing its mobile chips with similar performance to their desktop brethren, so we'd expect a mobile GT 1030 to hew closely to the desktop model.

Comments closed
    • jihadjoe
    • 2 years ago

    Seems to be confirmed as a GP108-300. 512 CUDA cores, 30W.

    [url<]https://www.techpowerup.com/232789/nvidias-pascal-gp108-300-gpu-pictured-benchmarked-powers-upcoming-gt-1030[/url<]

    • DavidC1
    • 2 years ago

    Technically it’s GDDR3, which isn’t the same as DDR3 system memory. Next step is GDDR5 not DDR4.

      • MathMan
      • 2 years ago

      Checkout Nvidia’s GT730 product page:
      [url<]http://www.geforce.com/hardware/desktop-gpus/geforce-gt-730/specifications[/url<] They mention GDDR5, the mention DDR3. The GDDR5 runs at 5Gbps. The DDR3 runs at 1.8Gbps. It's real DDR3, not GDDR3 (which may very well not even be in mass production anymore.)

    • kalelovil
    • 2 years ago

    If it is a new chip, a wonder whether we’ll see the first use of DDR4 on a graphics card.

      • NoOne ButMe
      • 2 years ago

      GDDR5 might be cheaper, as already is bought in bulk.

      I wouldn’t want to be buying DDR4 right now.

      • ImSpartacus
      • 2 years ago

      I was actually wondering this with respect to the 550’s curiously wide 128-bit bus. It isn’t necessary when you have 8-9 Gbps gddr5, but if you only have 2-2.5 Gbps ddr4? That helps.

    • Chrispy_
    • 2 years ago

    Yeah, the fact that Nvidia is willingly selling the re-re-re-badged GT430 in 2017 is kind of sick. They’re just milking the gullible at this point….

    At least the RX 540 should be capable of running AAA games at your typical 1366×768 budget laptop resolution. The same cannot be said of the GT730.

      • morphine
      • 2 years ago

      “Speculators can’t seem to agree on whether the rumored graphics card will rely on a further cut-down version of the GP107 silicon found in the existing GeForce GTX 1050 and GTX 1050 Ti, or on a new chip possibly called GP108”

      There’s no mention of a re-badge of the GT430, that was just an example to talk about potential memory configs.

        • CaptainObvious
        • 2 years ago

        He was talking about selling a rebadge of the GT430 as one of the GT730 available to consumers. As in this part:

        “The GeForce GT 730 cards with DDR3 memory on a 128-bit bus are based on the 40-nm GF108 “Fermi” chips originally used in 2010’s GeForce GT 430.”

          • jensend
          • 2 years ago

          username checks out

        • Shobai
        • 2 years ago

        The GF108 was used in the GT4[2,3,4]0, GT530, GT6[2,3]0 [DDR3 and GDDR5] and GT730 – he got the right number of prefixes, as far as I can see.

      • Ninjitsu
      • 2 years ago

      I dunno, a GT740M can manage 768p fairly well…

        • Shobai
        • 2 years ago

        You might have the wrong tree for barking: looks like the GT740M used either the GK107 or GK208, not the GF108.

          • Anonymous Coward
          • 2 years ago

          Quite a pile of chips and product names. But going through the pages at notebookcheck.net it looks like nVidia has kept both 64- and 128-bit interface 384-shader mobile GPUs for some years, Fermi, Kepler and Maxwell.

            • NoOne ButMe
            • 2 years ago

            Would you like a 640, 640 or 640? (Kepler G5, Fermi G3 and Fermi G5 I think)
            Or maybe it was the 630 that had 3 models.

            • ozzuneoj
            • 2 years ago

            I think its the 630. I have a GT 430 and a GT 630 in my stockpile of misc cards. This particular 630 (a PNY I believe) is much more modern looking and has a big fancy cooler but its actually nearly identical to the Fermi-based 430 except it has even lower memory clocks. GENIUS!

            Kepler was such an awesome advancement over Fermi… its pretty lame that anything pre-Kepler is still available on the market brand new. Hopefully Pascal will be the new Fermi and they’ll completely fill the lineup and finally replace all of those old chips.

            … hah, I know… that’s about as likely as any GPU under $100 being priced according to performance\features. As long as some uninformed derps keep buying GTX 750s and GT 740s for $100+ when 1050’s are available, the low end market will never make sense.

            • NoOne ButMe
            • 2 years ago

            I think Fermi may finally phase out for whichever GK108/208 or GM108 is smaller. 28nm ought to be cheaper than 40nm by now. And Fermi still doesn’t have promised DX12 support…

            • mczak
            • 2 years ago

            GM108 can’t do this job since it’s not a “full” graphic chip. Specifically, it lacks any kind of display outputs (and video decode/encode too), this is strictly a chip for optimus switchable graphics. That’s ok for mobile, not so much for desktop graphic cards…
            gk208 easily could have replaced gf108 in all products though obviously already.
            Albeit it’s not really that convincing anymore as a low-end graphic card for most uses – not only does it fail to outperform an ordinary HD 630 intel igp, it’s video decode (while actually decent – it can do 4k h.264, something which can’t be said about previous AMD low-end chips) of course isn’t quite up to modern standards neither.
            I fail to see though why nvidia would rebadge gk208 again – they didn’t with the gtx 900 series on the desktop, there are 900M series gk208 (and yes, gf117…) based mobile parts but I don’t think I’ve seen them in any products for ages (they can’t compete with GT2 intel HD graphics since Skylake really – only gm108 still sort of can albeit you really need the gddr5 version to outperform a HD620 with dual-channel memory).

            • Anonymous Coward
            • 2 years ago

            I’d love to see a low-end deathmatch between Fermi, Kepler and Maxwell. How much progress has been made in performance and efficiency? At the high end it seems like they cram in more and more hardware, but at the bottom they keep it fairly constant.

            • NoOne ButMe
            • 2 years ago

            Coolest would be higher end Fermi also. Like a 1-SM/SMX/SMM match, and a 384CC match.

            Although a 2-SM/SMX/SMM maybe easier to find all parts

      • NovusBogus
      • 2 years ago

      I tried running the Star Citizen alpha on a GT730 a few months ago. It was…interesting, to say the least.

      Hopefully the rumors are true, and hopefully it comes in a passive cooled half-height form factor. There aren’t currently very many good graphics options for those running SFF boxes.

Pin It on Pinterest

Share This