Nvidia GeForce GTX 1050 3GB Review: (Mostly) Faster Than 1050 2GB

Results: Power, Clock Rates, and GPU Temperature

Slowly but surely, we’re spinning up multiple Tom’s Hardware labs with Cybenetics’ Powenetics hardware/software solution for accurately measuring power consumption.

In brief, Powenetics utilizes Tinkerforge Master Bricks, to which Voltage/Current bricklets are attached. The bricklets are installed between the load and power supply, and they monitor consumption through each of the modified PSU’s auxiliary power connectors and through the PCIe slot by way of a PCIe riser. Custom software logs the readings, allowing us to dial in a sampling rate, pull that data into Excel, and very accurately chart everything from average power across a benchmark run to instantaneous spikes.

The software is set up to log the power consumption of graphics cards, storage devices, and CPUs. However, we’re only using the bricklets relevant to graphics card testing. Nvidia’s GeForce GTX 1050s get all of their power from the PCIe slot, while our Asus Radeon RX 560 requires a single six-pin auxiliary connector.

Idle

Sitting on the Windows desktop, driving a single QHD monitor, Nvidia’s GeForce GTX 1050 3GB averaged less than 7W, spiking about 1W higher during our 10-minute data collection window.

The card’s fan barely spun up (there is no semi-passive mode) and its GPU only got a little warmer than our ambient environment.

Gaming

The built-in benchmark of Metro: Last Light Redux gives us a consistent sequence that can be looped as many times as needed to generate good data.

An almost-65W average landed well below the 1050 3GB's 75W total board power rating (though our peak power reading gets quite a bit closer at 72W).

Interestingly, although we observed an average current draw of 5.39A on the 12V rail through three runs of the Metro benchmark, there were some spikes that exceeded the PCI-SIG’s 5.5A ceiling. For comparison, check out the same data next to current draw from GeForce GTX 1050 Ti 4GB and GeForce GTX 1050 2GB:

Those two notable dips are indicative of brief pauses between test runs. Otherwise, the data clearly shows our GeForce GTX 1050 3GB sample pulling more current from the PCIe slot than it should.

The card’s fan speed does ramp up in response to increased thermal load, even though Nvidia’s GP107 processor never heats up much more than 65°C.

FurMark

A more consistent workload is less susceptible to power consumption spikes (even if the task is as taxing as FurMark). GeForce GTX 1050 3GB doesn’t exceed 70W.

This time, average current draw of 5.58A does exceed the official 5.5A ceiling.

Again, putting all three GeForce GTX 1050s on one graph shows current draw from the 1050 3GB higher than it should be.

Voltage was pulled back to avoid violating Nvidia’s power limit, and core clock rate consequently dropped to about 1595 MHz. For comparison, the GPU voltage during our Metro run remained above 1.0V, and the core clock rate stuck right around 1771 MHz.

Comparing All Of The Cards

AMD’s Radeon RX 560 was easily the most power-hungry card in our gaming and stress test workloads. Between the GeForce GTX 1050s, though, Nvidia’s new 3GB card came closest to the company’s 75W rating (as you might have guessed from our current draw data).

GeForce GTX 1050 3GB also ran the warmest across both of our benchmarks.

Nvidia pushes the GeForce GTX 1050 3GB’s clock rate a bit higher than what we got out of GeForce GTX 1050 2GB through our Metro: Last Light Redux workload.

Meanwhile, GeForce GTX 1050 Ti 4GB, with its 768 CUDA cores and 128-bit memory bus, doesn’t need to operate at aggressive frequencies to lay down better performance than the other two cards. It took a bigger clock rate hit in gaming scenarios to avoid violating its power limit.

FurMark makes each card’s respective ceiling much more obvious. Given their power limits, GeForce GTX 1050 3GB sustained roughly 1.6 GHz, the 1050 2GB was allowed right around 1.4 GHz, and 1050 Ti 4GB topped out around 1.3 GHz.

MORE: Best Graphics Cards

MORE: Desktop GPU Performance Hierarchy Table

MORE: All Graphics Content

This thread is closed for comments
21 comments
    Your comment
  • caledbwlch
    Any chance we can get a comparison of the 1050 3GB with the memory overclocked in comparison to the 1050 Ti?
  • dudmont
    These 1050s are begging for some OEM to add a 6pin PCIE connector and rework the power limits on the cards. With good cooling, I'm betting 2GHZ isn't out of the question, with a good GPU.
  • closs.sebastien
    would be good if compared with a 1060
  • King_V
    bit_user

    Well, our somewhat opposing speculations when the 3GB version announced pretty much are BOTH confirmed with this test.

    Sometimes it achieves parity with the 1050Ti, and sometimes it dips below 1050 performance.

    Depends on the game.

    One thing neither of us considered in our conversation - the need for higher clocks pushes up power consumption to the point where spikes occasionally exceed the specs for the PCIe slot.

    Still, the results are interesting - and my curmudgeonly side somewhat objects to the idea of cutting memory bandwidth and compensating for it by cranking up the power.
  • Giroro
    Since this is basically an overclocked 1050 ti with one memory chip missing, I figure it should be no problem to overclock any 1050 Ti to the same rate as the 1050 3GB.
  • redgarl
    Why this product is even released...?
  • BulkZerker
    @redgarl
    Second paragraph of the article.
    "According to our sources, it really doesn’t. Slowly but surely, GeForce GTX 1050 3GB cards will start replacing 2GB boards, particularly as the 512MB memory chips used on those 2GB implementations become harder to source. "
  • littleleo
    Better to test the real product when it shows up. Things must be slow at Tom's.
  • cangelini
    2646065 said:
    Any chance we can get a comparison of the 1050 3GB with the memory overclocked in comparison to the 1050 Ti?


    I've been playing with this a bit. Wanted to get the GDDR5 fast enough to give 112 GB/s but realistically it's going to take down-clocking the 1050 2GB, overclocking the 3GB card, and then seeing what difference the missing ROP partition/L2 makes. Will continue trying to come up with a good comparison.
  • cangelini
    242296 said:
    Better to test the real product when it shows up. Things must be slow at Tom's.


    It *is* a real product. There's a model number and everything :) I've been itching to do something with graphics for months!
  • stdragon
    1919666 said:
    These 1050s are begging for some OEM to add a 6pin PCIE connector and rework the power limits on the cards. With good cooling, I'm betting 2GHZ isn't out of the question, with a good GPU.


    The entire point of the 1030 and 1050 series is so that you're not required to have a PSU connector. While I get bus powered GPUs are not high-end gamer market products, they do make for good low to mid-range upgrade; specifically for OEM computers such as Dell and HP units. Though honestly most of those budget machines don't have a 300W PSU which is what nVidia is requiring as minimum for for a 1050 card.
  • Tom_207
    So what I read was a defective GP107 running higher clock speeds draws more power, makes perfect sense.
  • leonhartfvii8
    Why is the GTX 1050 Ti below 1050 from 2GB in Starcraft 2? WTF position error?
  • Trevor98
    251426 said:
    Why this product is even released...?


    Answer: product/market segmentation.
  • Stardude82
    Disappointing. The GTX 950 I bought 3 years ago for $100 trades blows with this card and even about the same power consumption.
  • 10tacle
    54974 said:
    Disappointing. The GTX 950 I bought 3 years ago for $100 trades blows with this card and even about the same power consumption.


    Well at such a low level GPU hierarchy, you are not going to see as big of a leap as you would say in higher level 9xx vs. 10xx GPUs. Not only that, but 3GB vs. 2GB doesn't mean much if you only have a 128-bit memory bandwidth bus and clock speeds won't help that much more either. It's like trying to push more traffic on a clogged freeway.

    And if I could only afford this level of a GPU for PC gaming in this day in age with Gen-8 console gaming graphics quality for 1080p and optimized games for utilizing that, I'd just buy a console (no offense intended as one who owns a retro gaming Pentium IV PC running a GTX 275).
  • cangelini
    2691467 said:
    Why is the GTX 1050 Ti below 1050 from 2GB in Starcraft 2? WTF position error?


    From the analysis on that page: "Nvidia dominated in StarCraft II, with all three GeForce cards hitting a platform-imposed bottleneck." Because that combination of settings is limited by something other than graphics, the variation in performance isn't indicative of those three cards' capabilities.
  • ET3D
    Worth mentioning is that NVIDIA's PlayReady 3.0 implementation requires a card with at least 3GB, so people who want a cheap HTPC card for Netflix in 4K (and a bit of 1080p gaming on the side) will now have another option.
  • 80-watt Hamster
    54974 said:
    Disappointing. The GTX 950 I bought 3 years ago for $100 trades blows with this card and even about the same power consumption.


    Right? This article inspired me to look at how my R9 380 (now with one dead fan) stacks up in the current market. It's fascinating that there are, at the moment, exactly zero cards available that outperform or even equal it for the $180 I paid in January 2016. (Substitute GTX 960 if Team Green.) One can easily spend that or much more on a 1050 ti for less performance.
  • King_V
    2691467 said:
    Why is the GTX 1050 Ti below 1050 from 2GB in Starcraft 2? WTF position error?


    I wondered about that as well, but immediately below, it's stated:

    Quote:
    Nvidia dominated in StarCraft II, with all three GeForce cards hitting a platform-imposed bottleneck.


    So, my assumption would be that the slight variation between their FPS ratings are probably within a margin of error, and possibly that ordering is not consistently reproducible.

    That's just a guess on my part, though.

    EDIT: ah, right, cangelini already answered this question.
  • littleleo
    134065 said:
    242296 said:
    Better to test the real product when it shows up. Things must be slow at Tom's.
    It *is* a real product. There's a model number and everything :) I've been itching to do something with graphics for months!
    You know you can put Aloe on the itch.

    VGA cards are still way over priced and stock is still an issue. Why not modify the gaming cards so they don't do the hash for crypto mining well.