Nvidia unveils its GTX 1050 and GTX 1050 Ti for laptops

Boy, that mobile GeForce GTX 1070 is something, ain’t it? The power-efficient puissance of Nvidia’s Pascal architecture has certainly brought portable performance to a new plateau. It might not match the desktop GTX 1070 blow-for-blow, but it truly is faster than the GTX 1060 6GB—an impressively fast graphics card in its own right.

The thing is, did you see that on-battery test in our review of Gigabyte’s P57X v6? For those who skipped our earlier review, the relevant frametime graph is reproduced above. Long-time TR readers will already know as much, but for those who are just joining us, those frame times would make for a miserable experience. Even though the FPS average of those frames works out to 30 FPS, that plot indicates an intolerably jerky experience all the way through, at least in that test. Folks who want to game on the go are buying machines that really need to be plugged in to produce the best performance.

Moving between power outlets and truly being on the move is the difference between “portable” and “mobile” in a nutshell. Some folks really do want proper mobile performance, though. The laptop version of the GTX 1060 is a better compromise than the mobile GTX 1070, but the GTX 1060’s GP106 chip is still pretty thirsty for today’s slim & light laptops. Drawing upwards of 100W from a battery means that even the largest laptop batteries probably won’t survive much past an hour or so.

Folks who have wanted real PC-gaming-on-the-go until now have had to settle for the GTX 960M. That chip is based on the now-venerable GM107 part we first saw in the GTX 750 Ti almost three years ago. As impressive as it was then, it’s time for an update. So just as the GTX 750 Ti was replaced in the sub-75W, no-PCIe-power-connector-required space by the GTX 1050 Ti, the 960M is now being replaced by the dynamic duo of the GTX 1050 and the GTX 1050 Ti.

Yes, just like with the higher-powered Pascals, the GTX 1050 and 1050 Ti are headed into laptops in more or less unaltered form. I say “more or less” because there have been some changes, of course. Contrary to what you might expect, though, the changes are mostly in favor of higher performance. Check out this fancy chart full of specs:

  ROP

pixels/

clock

Texels

filtered/

clock

Stream

processors

Base clock speed Boost clock speed Memory

interface

width 

Memory transfer rate
GTX 1050 32 40 640 1354 MHz 1455 MHz 128-bit 7 GT/sec
GTX 1050 (mobile) 16 40 640 1354 MHz 1493 MHz 128-bit 7 GT/sec
GTX 1050 Ti  32 48 768 1290 MHz 1392 MHz 128-bit 7 GT/sec
GTX 1050 Ti (mobile) 32 48 768 1493 MHz 1620 MHz 128-bit 7 GT/sec
GTX 960M 16 40 640 1097 MHz 1176 MHz 128-bit 5 GT/sec

The items of particular note above are the core clock rates. While the laptop GTX 1050 hews fairly close to its desktop counterpart, the 1050 Ti takes a leap forward, with a 200 MHz bump to its base clock and a 220 MHz increase in its boost clock range. That’s a big jump, and an unusual move for a mobile version of a desktop GPU. 

Curiously, Nvidia’s documentation also lists the laptop GTX 1050 as having just 16 ROP units—half that of the desktop card. At a glance, that change seems pretty major. We’ll have to see if we can get our hands on a laptop equipped with one of these chips and put it to the test against the desktop GTX 1050 to see whether there’s a major difference in performance for real-world use. Despite the halved ROP count, the mobile GTX 1050 retains a 128-bit path to memory.

Besides the clock rates, the laptop GTX 1050 will apparently be available in a configuration with 4GB of local memory. That’s in contrast to the desktop part, which is only available with 2GB. Both laptop cards will be using the same 128-bit path to 7 GT/sec GDDR5 memory as their desktop counterparts.

 

BatteryBoost learns new frame-pacing tricks

As we saw from the frame-time chart at the beginning of this piece, smooth frame delivery is a major obstacle to enjoying a game on battery power. To ensure smooth frame delivery while saving battery power through lower frame rates, the mobile GTX 1050 and GTX 1050 Ti will be the first chips to benefit from a major update to Nvidia’s BatteryBoost technology.

BatteryBoost isn’t actually new, of course. The tech debuted with the mobile GeForce 800M-series. Nvidia says the version integrated into the new mobile Pascal chips is vastly improved over the original, though. The new version doesn’t just limit frame rates with a user-defined cap—it paces frames in hardware to ensure a smooth experience.

We know from our original Inside the Second investigations that Nvidia has been integrating some form of frame-pacing technology in its GPUs for several generations now. That technology has obviously had the most attention (and probably the biggest impact) when applied to multi-GPU micro-stutter.

The new BatteryBoost puts this tech to work in a novel way. Nvidia says the frame-metering done to limit the frame rate with BatteryBoost is done “right in the GPU.” Specifically,  the company says BatteryBoost can “control frame rate variance directly at flip time.” We suspect it makes use of the same secret sauce the company uses to smooth the hills and valleys of SLI frame times. According to the green team, the hardware-based nature of the new BatteryBoost algorithm means smoother gameplay compared to software-based framerate limiters.

The goal of BatteryBoost is mainly to achieve longer playtime while on battery power. By default, BatteryBoost will limit frame rates to 30 FPS. That may seem like a bad time waiting to happen, but long-time TR readers will know that consistent frame times and a smooth gaming experience are just as important as a high average frame rate.

Games might not be especially fluid-looking at 30 FPS, but if they’re as smooth-running as Nvidia promises, the experience could still prove immersive and undistracting. Given the current state of affairs with gaming on battery, we’re happy for the improvement. Folks who are willing to trade battery life for fluidity will be able to set higher caps, as well. To further reduce GPU load, GeForce Experience also has a new option to use different visual settings in games when running on battery power.

The combination of the two features should result in a pretty significant drop in power consumption while gaming. That, in turn, should definitely allow laptops to run significantly longer on battery. Nvidia surprisingly doesn’t make any bombastic claims about BatteryBoost this time around—the original release claimed double the runtime compared to laptops without the feature—but it does say that Pascal-powered laptops should be able to run 30% longer than Maxwell-equipped models.

 

A quick look at performance expectations

Of course, running longer is all fine and well as long as you’re having a good time. Fortunately, the new GeForces should be able to do a decent job with whatever games you want to throw at them. Unfortunately, because we don’t have a laptop handy with one of the new graphics chips on board, we’ll have to make do with some numbers provided by the green team. All of the benchmarks were performed at a 1920×1080 screen resolution. Our apologies for the watermarked stats tables.

According to Nvidia’s numbers, the laptop versions of the GTX 1050 and 1050 Ti run strikingly close to the desktop cards’ performance. The company claims Grand Theft Auto V‘s built-in benchmark reports 70 FPS on the GTX 1050 and 87 FPS on the GTX 1050 Ti. That’s in on “very high” settings, albeit with MSAA disabled. The Division‘s built-in benchmark is more demanding, but the new cards still pulled out 36 FPS on the GTX 1050 and 46 FPS on the GTX 1050 Ti with the “high” preset.

Things shake out the same way in basically all of the company’s tests—including Fallout 4, Crysis 3, Rise of the Tomb Raider, Far Cry 4, Hitman, and The Witcher 3 Middle Earth: Shadow of Mordor is the major exception. Nvidia tested the open-world title on its “high” preset, and it strongly favored the GTX 1050 Ti. The faster card put out 74 average FPS, versus 43 on the GTX 1050. All in all, a strong showing for both of the petite Pascals.

Of course, these are Nvidia’s numbers, and we haven’t verified them ourselves. Also, all of the company’s numbers are just average FPS measurements, so they don’t tell us much about the smoothness of the gameplay experience. Moreover, beyond the model names (MSI GP62 and GE62), Nvidia didn’t actually specify exactly what hardware was in the laptops it used for testing. MSI has used those names for many, many laptops with a dizzying array of hardware configurations, so it’s difficult to really put these numbers in context.

I think it’s also worth noting that you won’t see this kind of performance if you want to make good on those extended battery life claims. To stop sucking power and start sipping juice, users will need to enable BatteryBoost, and that means lower framerates. Whether that trade-off is worth it is up to the individual user, of course, but we’re eager to give the next generation of BatteryBoost a try to see whether the guaranteed smoothness outweighs those lower frame rates.

Nvidia says “every major laptop manufacturer” will have laptops available with the GTX 1050 and GTX 1050 Ti inside soon. As mainstream graphics parts, the 1050 and 1050 Ti are sure to be found in gaming laptops of every size and shape. While the performance of the GTX 1060, GTX 1070, and GTX 1080 is very impressive, most folks gaming on laptops probably aren’t so hardcore. Paired with a G-Sync laptop display, the GTX 1050 and 1050 Ti could let you get in a couple satisfying rounds of Overwatch or time with other demanding titles on your lunch break. Stay tuned to our news page if you’re looking for a laptop with one of these inside—we’re sure to see plenty of products during CES.

Comments closed
    • chµck
    • 3 years ago

    any idea on this vs an amd 460?

      • tipoo
      • 3 years ago

      You looking at the Radeon Pro 460 in a certain laptop perchance? The 460 is closer to the 960 (above it) than the 965, so with the 1050 being the replacement of the 965 with a small boost over it, it should beat the 460 as well.

    • deruberhanyok
    • 3 years ago

    1050’s specs are a bit odd; GPU’s 640/40/16 matches the 750 ti, which is okay, and Pascal is going to be quicker than Maxwell, also okay, and the memory speed is much higher, also okay.

    But it seems like a set of cuts possibly designed to match performance differences in current GM107 mobile workstation GPUs – maybe the 1050 Pascal supplants the Quadro M1000M, a GM107 512/32/16 part, while the full on 1050ti variant with 768/48/32 would be used in a replacement for the M2000M’s 640/40/32 (TPU’s GPUDB lists the M2000M with 32 ROPs, just going off of that, I thought GM107 was a 16 ROP part).

    Was anything said about the form factor for these? The previous-gen 960M, which found its way into a lot of systems, was just rebadged 850M/860M, based on GM107 and afaik typically integrated on the motherboard, whereas the 965M was GM204/GM206-based and on an MXM module.

    Either way it’s an upgrade, but if it’s BGA type then that might mean we’ve finally moved on from GM107 as the baseline for NVIDIA “gaming” graphics on laptops, THREE YEARS since the launch of the 860M. It’s about time. It was getting a bit long in the tooth.

    (Workstation variants will probably be MXM regardless, but that’s a whole other thing).

    • ImSpartacus
    • 3 years ago

    I never thought I’d see a tech report article that actually clearly and plainly explains the differences between deceptively named Nvidia GPUs. I’m happy to be wrong about that.

    • Redocbew
    • 3 years ago

    [quote<]The power-efficient puissance of Nvidia's Pascal architecture has certainly brought portable performance to a new plateau.[/quote<] It's too much! I can't take it! The preponderance of alliteration is pushing me perilously close to the precipice.

    • Chrispy_
    • 3 years ago

    \o/ Awesome \o/

    Did I miss it or were there no TDP numbers for these new chips?

    I’m presuming that the mobile 1050Ti is 75W since it’s higher clocked than the desktop version. I was kind of hoping that the 1050 might be a sub-50W part. I really miss the endurance of my 45W Kepler 650m (GK107) but it’s too slow to be more than a very casual gamer option now, at something like Geforce 920m levels of performance.

      • tipoo
      • 3 years ago

      Nvidia doesn’t post them outright anymore for some reason, maybe because they’re configurable by the laptop maker and dependant on the cooling. But the 1060 in laptops was around 80W I think, the 1050TI should be around 50W.

      • ImSpartacus
      • 3 years ago

      Most mobile gpus don’t publicize tdps because the oems have flexibility to play with clocks and other levers/pulleys to get a particular thermal output. The same thing can happen with CPUs (Intel actually embraced this with their m series stuff a while back).

      And yes, that means that laptops can perform differently despite comparable spec sheets. Pretty sketchy, but what are you gonna do.

        • Voldenuit
        • 3 years ago

        [quote<]And yes, that means that laptops can perform differently despite comparable spec sheets. Pretty sketchy, but what are you gonna do.[/quote<] Makes sense. A desktop i7 with a water cooler will boost more consistently than one with the dinky stock intel HSF, so the situation is even more pronounced in notebooks, where CPU and GPU parts will run into thermal and power constraints sooner. I don't think it's shady per se, just giving OEMs the flexibility to prioritize weight, size, cost, power draw, cooling, noise, performance etc.

    • Kretschmer
    • 3 years ago

    This is exciting tech!

    • tipoo
    • 3 years ago

    “Nvidia says “every major laptop manufacturer” will have laptops available with the GTX 1050 and GTX 1050 Ti inside soon.”

    Shots fired at Apple, lol
    Unless…

    Anywho, I expect Nvidia is being coy on wattages again? These would probably be ~50W or 35W parts right?

      • RAGEPRO
      • 3 years ago

      No specs as usual (as you expected.) Your guess on the wattages sounds about right, maybe 50 and 40W.

        • Chrispy_
        • 3 years ago

        Oh, ignore my earlier comment about TDP values above then 😛

        I’d expect the 1050Ti to be a drop-in replacement for the 75W 960M. I’d obviously be happy if it were lower, but that’s what I’m expecting.

        The cut ROPs on the vanilla 1050 don’t seem to have hurt 1080p performance based on Nvidia’s own cooked numbers, but I wonder if that is specifically a power-saving measure. If I understand it right, having more ROPs typically only matters when you raise the resolution and it’s fair to say that anyone buying a lower cost laptop with “only” a 1050 in it is likely to be running 1366×768 and probably 1080p at the very most.

          • synthtel2
          • 3 years ago

          ROP load increases with resolution approximately the same amount as shader load does.

          Is it just me, or was Nvidia more reluctant than usual to use MSAA in their performance data? That could be related to a reduced ROP count.

            • Chrispy_
            • 3 years ago

            possibly, but MSAA is almost dead now anyway with so few rendering methods that actually support it. FXAA, SMAA and temporal methods are what do work (and are cheaper than MSAA) in modern rendering engines.

            • synthtel2
            • 3 years ago

            It looks like 3 of the 5 games in question do support it, though, and they used 2x/FX/unknown (and disabled reflection MSAA in GTA V). Part of it is that I’m used to them using heavy MSAA anywhere it’s available because it makes them look better than AMD, and this could just be because it’s a low-end bench in which they don’t directly compare themselves to the competition.

            * I think FO4 is still a purely forward renderer (should be easy to use MSAA with), but they didn’t use it unless it’s a part of ultra presets (which it might be – maybe some gerbil who owns it could check?).

            * Crysis 3 is light pre-pass (relatively easy MSAA), and they chose 2x MSAA over SMAA or 4x.

            * GTA V is deferred, but seems to be arranged such that MSAA isn’t [url=http://www.geforce.com/sites/default/files-world/attachments/grand-theft-auto-v-pc-anti-aliasing-performance.png<]too costly[/url<] (not like DX:MD at least). Reflection MSAA is [url=http://international.download.nvidia.com/geforce-com/international/images/grand-theft-auto-v/grand-theft-auto-v-reflection-msaa-performance.png<]pretty cheap[/url<] (that part is forward and tiny) and it seems weird that they explicitly disabled it. MSAA isn't dead as long as forward renderers aren't, and forward+ / forward clustered renderers aren't necessarily so bad as all that (see the latest Doom for an example). MSAA will also remain the gold standard anywhere specular aliasing is otherwise controlled (which it can be and IMO should be).

Pin It on Pinterest

Share This