Stories
Slash Boxes
Comments

SoylentNews is people

posted by janrinok on Monday April 06 2020, @10:13PM   Printer-friendly
from the good-enough-for-some dept.

NVIDIA's GeForce GTX 1650 GDDR6 Released: GDDR6 Reaching Price Parity With GDDR5

Tucked inside NVIDIA's announcement of their spring refresh of their mobile GPU lineup, the company included a new low-end mobile part, the GeForce GTX 1650 GDDR6. Exactly as it says on the tin, this was a version of the company's GTX 1650 accelerator, except with newer GDDR6 instead of the GDDR5 it launched with. Now, in one of NVIDIA's more poorly kept secrets, their desktop product stack is getting a version of the card as well.

[...] The entry-level card is the cheapest (and the slowest) of the Turing family, offering as much performance as NVIDIA can pack into a 75 Watt TDP.

[...] Overall, this low-key release should mark a more important turning point in the state of GDDR memory. If NVIDIA and its partners are now willing to release GDDR6 versions of low-end cards, then this is a strong indicator that GDDR6 has finally lost most of its new technology price premium, and that memory prices have fallen by enough to be competitive with 8Gbps GDDR5. GDDR6 prices were a sticking point for the profit-sensitive NVIDIA during the original Turing product stack launch, so while it has taken an extra year, the company is finally offering a top-to-bottom GDDR6-based product stack.

Let's see more GPUs and APUs with HBM already.


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 2) by Runaway1956 on Tuesday April 07 2020, @01:07AM (12 children)

    by Runaway1956 (2926) Subscriber Badge on Tuesday April 07 2020, @01:07AM (#979833) Journal

    I purchased one of these. Zotac makes a low profile card, that doesn't need any extra power connections. Just plug it into any x16 slot, and away it goes. For $150 you get a low power consumption card, that outperforms the 980, and runs head-to-head with the 1080, both of which sold for near a thousand dollars.

    Had I known about the Super cards, I would have waited a few more weeks to make that purchase. Extra cores for almost the same price, basically. But, I haven't found the 1650 Super in low profile yet. Give it some time, I guess.

    No, it isn't going to run the latest and bestest games at the highest settings, but the card isn't aimed at gamers. For that, you want the 2080 cards.

    Starting Score:    1  point
    Karma-Bonus Modifier   +1  

    Total Score:   2  
  • (Score: 2) by Booga1 on Tuesday April 07 2020, @02:13AM

    by Booga1 (6333) on Tuesday April 07 2020, @02:13AM (#979847)

    Yeah, I have a friend who could use something like this as well. He bought an eMachines computer with a dinky 250 watt power supply in a custom form factor that isn't replaceable. He's currently on another 75 watt nVidia card, so this would be a perfect option to swap it out for.

  • (Score: 2) by takyon on Tuesday April 07 2020, @02:36AM (4 children)

    by takyon (881) <takyonNO@SPAMsoylentnews.org> on Tuesday April 07 2020, @02:36AM (#979849) Journal

    I am more interested to see what performance will fit in this kind of TDP in the future than in the monster cards.

    For comparison:

    RTX 2080 Ti: 13.45 TFLOPS
    RTX 2080: 10.07 TFLOPS
    GTX 1080: 8.9 TFLOPS
    GTX 980: 5 TFLOPS
    PlayStation 4 Pro: 4.2 TFLOPS
    GTX 1650 (GDDR5): 3 TFLOPS
    GTX 1650 (GDDR6): 2.85 TFLOPS
    PlayStation 4: 1.84 TFLOPS
    Ryzen 7 4800U: 1.79 TFLOPS

    This GDDR6 version offers no advantage over the one you have, which is probably why it was quietly launched.

    From the TFLOPS number, which does not mean everything, it's weaker than the 980 or 1080. This list [tomshardware.com] confirms that, although the performance gap is only about 25% between 1650 and 980.

    --
    [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
    • (Score: 0) by Anonymous Coward on Tuesday April 07 2020, @03:41PM (3 children)

      by Anonymous Coward on Tuesday April 07 2020, @03:41PM (#979978)

      VR isn't quiet portable yet.
      the 1080GTX is the "low bar" and not found in laptops and the 2080 is still to expensive when inside a laptop?

      • (Score: 2) by takyon on Tuesday April 07 2020, @06:53PM (2 children)

        by takyon (881) <takyonNO@SPAMsoylentnews.org> on Tuesday April 07 2020, @06:53PM (#980041) Journal

        VR is still in an early adopter hell. Any headset without eye tracking and foveated rendering is obsolete. Those could be used to push framerate and (maximum) resolution up, with the eventual target being 8K-16K and 200+ FPS. Software advancements may be almost as important as the GPU hardware advancements.

        Portability goes beyond laptops. All headsets will eventually become standalone, like today's Snapdragon 8XX headsets but with 100x the performance.

        Nvidia just refreshed its high-end laptop GPUs:

        NVIDIA’s 2020 Laptop Refresh: Launches GeForce RTX 2080 Super, 2070 Super, & GTX 1650 Ti [anandtech.com]

        First and foremost, like the desktop realm, the regular RTX 2080 is now gone from laptops as well. With the RTX 2080 Super taking up the flagship spot – and not being massively different from the original RTX 2080 – NVIDIA has excised the original entirely. The RTX 2070 Super is instead NVIDIA’s second-tier adapter for laptops.

        The RTX 2070, on the other hand, is still staying around. Instead, it’s getting pushed down the product stack to the third-tier position. NVIDIA now expects RTX 2070 to start showing up in laptops as cheap as $1199.

        The RTX 2060 is also along for the ride. And this one is a particularly notable shift, as the RTX 2060 will now be NVIDIA’s anchor SKU for $999 laptops. This spot was previously held by the GTX 1660 Ti, and while NVIDIA does not explicitly discuss laptop part pricing, reading between the lines it’s clear that the company has cut laptop adapter prices to make this new product stack happen. So, as NVIDIA likes to promote, RTX laptops now start at $999.

        If you want a "VR laptop", you'll have to pay out the ass. Unless you can find a laptop that "fell off the back of the truck".

        --
        [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
        • (Score: 1, Informative) by Anonymous Coward on Thursday April 09 2020, @01:05AM (1 child)

          by Anonymous Coward on Thursday April 09 2020, @01:05AM (#980458)

          I'm waiting for the trifecta of eye-tracking, foveated rendering, and dynamic focal length compensation
          (your eyes naturally want to focus on a close plane when aimed at something closer -- and providing just a fixed focal depth gives you a headache).

          Although for actually perfect rendering of movement, it's also necessary to provide per-eye 3d-- (a, b and c axis rotations centred on each eye -- eyes rotate about z, don't you know -- don't believe me? Look in to a mirror at your iris, now tilt your head -- your eye will probably stay level, requiring it to rotate in its socket.) --motion compensation to the image element.

          So that it can stay locked to the fovea despite eye movement. This is the only way to provide motion-blur effects that work properly when your eyes are tracking moving objects, as well as to eliminate the loss-of-detail that occurs when there's a big velocity difference between the two.

          Nice additional benefit of that, is that the actual render resolution (and actual display resolution) can then be adapted to the fovea itself, which allows it to be dropped surprisingly low -- less than a few Megapixel IIRC -- without any loss of visible detail. One can still have ~4 pixels per actual fovea cell without needing to truly render the massive dimensions we currently use, which will cause massive improvements to both dynamic performance (frame rate / latency) as well as energy efficiency.

          It makes a good case for a deferred HW renderer like PowerVR -- since the selection of the pixels required to really be rendered could be made by that kind of hardware quite late in the rendering process, although it would require some adaptation to deal with the non-linearity in the pixel spacing.

          But, yeah, 3d movement systems that have fast enough feedback to effectively neutralise the movements of two human eyes, whilst remaining lightweight, is a bit beyond us just yet it seems.

          'til then, I just don't think VR is serious. It's getting better, but it's just not quite there yet. It'll continue to give you a headache after a few hours of use, and require games to have workarounds to solve the issues with blur due to movement.

          • (Score: 2) by takyon on Thursday April 09 2020, @02:25AM

            by takyon (881) <takyonNO@SPAMsoylentnews.org> on Thursday April 09 2020, @02:25AM (#980479) Journal

            The focal length problem is being worked on:

            Half Dome Updates: FRL Explores More Comfortable, Compact VR Prototypes for Work [oculus.com]


            Half Dome was our first prototype to achieve two key milestones. First, using Fresnel lenses, it supports a 140° field of view. Second, by physically moving the screens based on eye tracking, it ensures that the image remains sharp, even when inspecting close objects. This was our first step toward more immersive and visually comfortable VR displays in the lab.

            [...] Overall, we’ve been able to improve the form factor substantially and reduce weight by a full 200 grams over Half Dome 1. While the field of view for Half Dome 2 is narrower than its predecessor, it’s still 20% wider than Oculus Quest.

            [...] The mechanical varifocal system relies on voice coil actuators and flexure hinge arrays. These reduce friction, minimize noise, and maximize the lifetime of the headset’s moving mechanical parts—all improvements on the original Half Dome prototype.

            We also unveiled an electronic version in the Half Dome series of prototypes—one that replaces all of the moving mechanical parts with a new type of liquid crystal lens made from a thin, alternating stack of two flat optical elements: polarization-dependent lenses (PDLs) and switchable half-wave plates.

            PDLs are special because their focal length changes based on their polarization state. By changing the voltage applied to the switchable plates, we can toggle between the two focal lengths. This could make for a great set of digital bifocals, but it doesn’t necessarily make for comfortable VR. By stacking a series of PDLs and switchable half-wave plates on top of each other, we’re able to achieve smooth varifocal that lets you comfortably and seamlessly adjust your focus in the headset.

            [...] As you can see, the liquid crystal lenses shift focus between different depths. To achieve smooth varifocal, we address the full stack of liquid crystal lenses, with each additional pair doubling the number of focal planes. In the example above, six liquid crystal lenses are driven to sweep through 64 focal planes, and you can see the focal depth smoothly changing at the right as we cycle through different sets of lens states.

            In addition to having no moving parts, this approach allows for a significantly better form factor compared to its predecessors in the Half Dome series of prototype. When we integrate the electronic module into a complete prototype headset, it defines a new state of the art for VR ergonomics.

            [...] This latest update to the Half Dome series of prototypes digitally switches between a large number of focal states—so many, in fact, that it appears continuous, just like the mechanical varifocal displays. For the first time, we’ve achieved a more ergonomic design with a significant reduction in weight. That opens the door to a future where we can be in VR for hours at a time, collaborating with others using any number of digital whiteboards spread throughout our ideal virtual workspace. While we’re still developing the Half Dome series of prototypes at FRL, we’re excited to see where our varifocal research takes us next.

            Yeah, that's another great reason why VR is in early adopter hell.

            As far as eye tracking in headsets goes, it should be standard soon. The performance left on the table by not using foveated rendering is too important.

            Eye tracking is the next phase for VR, ready or not [cnet.com]

            --
            [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
  • (Score: 2) by jasassin on Tuesday April 07 2020, @03:24AM (5 children)

    by jasassin (3566) <jasassin@gmail.com> on Tuesday April 07 2020, @03:24AM (#979867) Homepage Journal

    but the card isn't aimed at gamers

    Not being a jerk here. Who is it aimed at?

    --
    jasassin@gmail.com GPG Key ID: 0xE6462C68A9A3DB5A
    • (Score: 5, Interesting) by Runaway1956 on Tuesday April 07 2020, @04:18AM (2 children)

      by Runaway1956 (2926) Subscriber Badge on Tuesday April 07 2020, @04:18AM (#979884) Journal

      It's aimed primarily at "entry level" gamers, whatever the heck that means. You'll find that phrase on various sites. The low profile card is very specifically aimed at people who have small form factor boxes, which is exactly what I was looking for. It fits into a 2U server, and a bunch of those mini-ATX, micros, and whatever else. It will make an excellent card for enterprise, and for regular people who need something better than an embedded Intel, but don't need a gaming rig.

      Let me hit on that gaming thing again: the 1650 will play just about any game 5 years old very nicely, on higher settings, if not the highest settings.

      Games being released now and in the next year or so are actually waiting on optimised drivers for the RTX 2070 and 2080. My low-end 1650 isn't even under consideration by real gamers.

    • (Score: 2) by shortscreen on Tuesday April 07 2020, @08:12AM (1 child)

      by shortscreen (2252) on Tuesday April 07 2020, @08:12AM (#979915) Journal

      There are gamers and then there are gamers. For some, it seems that spending the most money and obtaining the most bragging rights is the primary game. For them, anything other than the flagship part is "low end" even though 99% of games would run fine on it. (Since when is the GTX moniker used for low end parts anyway?)

      I play plenty of games, but I've never once spent even $100 on a video card let alone $500. But it's nice that somebody buys that stuff, so I can buy it from them on ebay five years later for much less.

      • (Score: 2) by takyon on Tuesday April 07 2020, @11:40AM

        by takyon (881) <takyonNO@SPAMsoylentnews.org> on Tuesday April 07 2020, @11:40AM (#979937) Journal

        (Since when is the GTX moniker used for low end parts anyway?)

        Since the ultra expensive RTX 20-series cash grab was launched. Now "GTX" is used for a handful of cards [wikipedia.org] without any dedicated raytracing (or ML) cores. Turing might be the only generation to do this since Nvidia (as well as AMD) are going to want some minimum level of raytracing functionality on every card eventually.

        --
        [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]