Bringing 2019 to a close in the GPU space, we have one final video card review for the year: NVIDIA’s GeForce GTX 1650 Super. The last of the company’s mid-generation kicker cards to refresh the lineup in the second-half of 2019, the GTX 1650 Super is designed to shore up NVIDIA’s position in the sub-$200 video card market, offering solid performance for 1080p gaming without breaking the bank. At the same time, it’s also NVIDIA’s response to AMD’s new Radeon RX 5500 XT series of cards, which having landed last week, significantly outperform the original GTX 1650.

Like the other Super cards, these refresh parts serve to shore up NVIDIA’s competitive positioning against AMD’s Radeon RX 5000 series cards, offering improved performance-per-dollar at every tier. NVIDIA has taken some flak for uncompetitive pricing, and this is not unearned. And it goes perhaps most for the original GTX 1650, which although is easily the best 75W card on the market, it’s always been surpassed in value by AMD’s cards; first the last-generation Polaris cards, and now the new Navi cards. So with the GTX 1650 Super, NVIDIA and its partners finally get a chance to rectify this with a more competitive part, that while no longer fitting into the original’s 75W niche, offers better performance all around.

NVIDIA GeForce Specification Comparison
  GTX 1660 GTX 1650 Super GTX 1650 GTX 1050 Ti
CUDA Cores 1408 1280 896 768
ROPs 48 32 32 32
Core Clock 1530MHz 1530MHz 1485MHz 1290MHz
Boost Clock 1785MHz 1725MHz 1665MHz 1392MHz
Memory Clock 8Gbps GDDR5 12Gbps GDDR6 8Gbps GDDR5 7Gbps GDDR5
Memory Bus Width 192-bit 128-bit 128-bit 128-bit
VRAM 6GB 4GB 4GB 4GB
Single Precision Perf. 5 TFLOPS 4.4 TFLOPS 3 TFLOPS 2.1 TFLOPS
TGP 120W 100W 75W 75W
GPU TU116
(284 mm2)
TU116
(284 mm2)
TU117
(200 mm2)
GP107
(132 mm2)
Transistor Count 6.6B 6.6B 4.7B 3.3B
Architecture Turing Turing Turing Pascal
Manufacturing Process TSMC 12nm "FFN" TSMC 12nm "FFN" TSMC 12nm "FFN" Samsung 14nm
Launch Date 03/14/2019 11/22/2019 04/23/2019 10/25/2016
Launch Price $219 $159 $149 $139

From a pure hardware perspective, perhaps the most interesting thing about the GTX 1650 Super is that, unlike the other Super cards, NVIDIA is giving the new Super card a much bigger jump in performance over its predecessor. With a of 46%, increase in GPU throughput and faster 12Gbps GDDR6 memory, the GTX 1650 Super is much farther ahead of the GTX 1650 than what we saw with October’s GTX 1660 Super launch, relatively speaking.

The single biggest change here is the GPU. While NVIDIA is calling the card a GTX 1650, in practice it’s more like a GTX 1660 LE; NVIDIA has brought in the larger, more powerful TU116 GPU from the GTX 1660 series to fill out this card. There are cost and power consequences to this – the 284mm2 is a very large chip to be selling for $159 – but the payoff is that it gives NVIDIA a lot more SMs and CUDA Cores to work with. Coupled with that is a small bump in clockspeeds, which pushes the on-paper shader/compute throughput numbers up by just over 46%.

Such a large jump in GPU throughput also requires a lot more memory bandwidth to feed the beast. As a result, just like the GTX 1660 Super, the GTX 1650 Super gets the GDDR6 treatment as well. Here NVIDIA is using slightly lower (and lower power) 12Gbps GDDR6, which is attached to the GPU via a neutered 128-bit memory bus. Still, this one change gives the GTX 1650 Super 50% more memory bandwidth than the vanilla GTX 1650, very close to its increase in shader throughput.

Do note, however, that not all aspects of the GPU are being scaled out to the same degree. In particular, the GTX 1650 Super still only has 32 ROPs, with the rest of TU116’s ROPs getting cut off along with its spare memory channels. This means that while the GTX 1650 Super has 46% more shader performance, only has 4% more ROP throughput for pushing pixels. Counterbalancing this to a degree will be the big jump in memory bandwidth, which helps to keep those 32 ROPs well-fed, but at the end of the day the GPU is getting an uneven increase in resources, and gaming performance gains do reflect this at times.

The drawback to all of this, then, is power consumption. While the original GTX 1650 is a 75 Watt card – making it the fastest thing that can be powered solely by a PCIe slot – the Super-sized card is officially a 100 Watt product. This gives up the original GTX 1650’s unique advantage, and it means builders looking for even faster 75W cards won’t get their wish, but it’s the power that pays the cost of the GTX 1650 Super’s higher performance. Traditionally, NVIDIA has held pretty steadfast at 75W for their xx50 cards, but then again at the end of the day, despite the name, this is closer to a lightweight GTX 1660 than it is a GTX 1650.

Speaking of hardware features, besides giving NVIDIA a good deal more in the way of GPU resources to play with, the switch from the TU117 GPU to the TU116 GPU also has one other major ramification that some users will want to pay attention to: video encoding. Unlike TU117, which got the last-generation NVENC Volta video encoder block for die space reasons, TU116 gets the full-fat Turing NVENC video encoder block. Turing’s video encode block has been turning a lot of heads for its level of quality – while not archival grade, it’s competitive with x264 Medium – which is important for streamers. This also led to TU117 and the GTX 1650 being a disappointment in some circles, as an otherwise solid video card was made far less useful for video encoding. So with the GTX 1650 Super, NVIDIA is resolving this in a roundabout way, thanks to the use of the more powerful TU116 GPU.

Product Positioning & The Competition

As is always the case in the lower-end segment of NVIDIA’s product stack, the GTX 1650 Super is a pure virtual launch. This means that NVIDIA hasn’t put together a retail reference design, and all of the cards are based on partner designs.

At this point, the partners have been shipping TU116 and TU117-based cards for over 8 months, so they have been able to hone their GTX 16-series designs. This means that they’ve been able to hit the ground running, with existing designs ready for quick modification or straight reuse right away. The net result is that the newest GTX 1650 Supers look like and are built like the GTX 1660 and GTX 1650 cards that have preceded them.

Within NVIDIA’s product stack then, the GTX 1650 Super is not a wholesale replacement for the GTX 1650 – that card is still sticking around – but the GTX 1650 Super is going to be the value option for this performance segment. For consumers and OEMs who need 75W cards (for cooling or power reasons), then the $149 GTX 1650 remains the best choice. For everyone else, the GTX 1650 Super offers a whole lot more performance for $10 more. And while it’s not going to be performing on the same level as NVIDIA’s $200+ GTX 1660 cards, the GTX 1650 Super packs enough horsepower that it’s not going to be too far behind.

NVIDIA GeForce 20/16 Series (Turing) Product Stack
RTX 20 Series GTX 16 Series
RTX 2080 Ti GTX 1660 Ti
RTX 2080 Super GTX 1660 Super
RTX 2070 Super GTX 1660
RTX 2060 Super GTX 1650 Super
RTX 2060 GTX 1650

It’s looking outside of NVIIDA’s product stack where we find the real competition for the GTX 1650 Super: AMD’s new Radeon RX 5500 XT series cards, particularly the $169 4GB model. While NVIDIA did not directly call out AMD when first revealing the GTX 1650 Super, the timing – between the RX 5500 series announcement and RX 5500 XT launch – leave no doubt in that respect. NVIDIA has been seemingly content to let AMD hold the $150-$200 market with their RX 580/570 cards, but with the latest AMD launch, that passive positioning has come to an end.

As I noted in last week’s RX 5500 XT review, there’s not quite a 1:1 match between Radeon and GeForce parts right now. The 4GB RX 5500 XT is $10 more expensive than the GeForce GTX 1650 Super, which as Ian Cutress made note of when we were discussing this article earlier this week, in the sub-$200 market customers are typically buying what they can afford. So even $10 matters in some cases. Still, it’s in NVIDIA’s best interests to meet or beat the RX 5500 XT 4GB on performance, to deny AMD that edge. And meanwhile NVIDIA generally has the edge on energy efficiency, though it’s no longer the one-sided fight it was against AMD’s Polaris-based RX 500 series cards.

Finally, the wild card factor here is once again going to be gaming bundles. NVIDIA doesn’t offer one, but AMD does. Along with a 3-month trial for Microsoft’s Xbox Games Pass program, the company is bundling the forthcoming “Master Edition” of Monster Hunter: Iceborne. We don’t often see game bundles with sub-$200 cards, so the inclusion of one can be a powerful factor in this segment of the market, since a game is a more significant fraction of the value of a card.

Though with most of Newegg’s video card stock being anything but in stock, just getting a card is a challenge right now. The first wave of GTX 1650 Super cards have done pretty well sales-wise, so the video card retailer has all of two models of GTX 1650 Super in stock, and Amazon is much the same.

Holiday 2019 GPU Pricing Comparison
AMD Price NVIDIA
Radeon RX 5700 $319 GeForce RTX 2060
  $279 GeForce GTX 1660 Ti
  $229 GeForce GTX 1660 Super
Radeon RX 5500 XT 8GB $199/$209 GeForce GTX 1660
Radeon RX 5500 XT 4GB $169/$159 GeForce GTX 1650 Super
  $149 GeForce GTX 1650
ZOTAC Gaming GeForce GTX 1650 Super
POST A COMMENT

67 Comments

View All Comments

  • guachi - Friday, December 20, 2019 - link

    The 1650S overall is the card to get if buying a new card. But I'd still recommend a used 570 or 580 (and maybe a new 570 if you get one in sale).

    Polaris will never die.

    I just wouldn't buy THIS 1650S. The noise. Ouch! 50dB?

    No.
    Reply
  • lmcd - Friday, December 20, 2019 - link

    It's a convenient size, enough so that airflow in my case will result in better overall acoustics in my case compared to a larger card. Agreed that there's better designs yet but this isn't as awful as you're implying. Reply
  • Spunjji - Monday, December 23, 2019 - link

    50db is terrible under any circumstance. Reply
  • lmcd - Friday, December 20, 2019 - link

    Gonna be honest I don't quite understand why the 1050 Ti and 1060 3GB both need to be in this graph set while the 1070 didn't make it in. Usually there aren't performance regressions from one generation to another, so it's more interesting to compare a higher card from the previous generation to a lower card from the current generation. Reply
  • Ryan Smith - Friday, December 20, 2019 - link

    The 1070 was a $350 card its entire life. Whereas the 1050 Ti and 1060 3GB were the cards closest to being the 1650 Super's predecessor (1050 Ti was positioned a bit lower, 1060 3GB a bit higher). So the latter two are typically the most useful for generational comparisons.

    At any rate, this is why we have Bench. So you can use that to make any card comparisons you'd like to see that aren't in this article itself. https://www.anandtech.com/bench/GPU19/2638
    Reply
  • catavalon21 - Saturday, December 21, 2019 - link

    In doing so, it paints an interesting light for those of us who do not upgrade every generation. While the 970 can't be compared directly to it in Bench, it's interesting to see how many benchmarks show it besting the 980 - which was a $550 card when it debuted. Maybe the RTX series cards are worthy of their criticisms for gen-over-gen improvement in performance per dollar, but not this guy. Yes, I know 980 was 2 generations ago, but still. The 980 takes some of the benchmarks, especially CUDA, but across the board, the 1650S competes very well. For a card to have 980-like performance for $160 at 100 watts, I'm impressed. Reply
  • The_Assimilator - Saturday, December 21, 2019 - link

    No you're wrong, according to forum keyboard warriors there's been no improvement in price/perf in the last half decade because they can't get top-tier performance for $100. ;) Reply
  • Spunjji - Monday, December 23, 2019 - link

    That we're only seeing a price/performance improvement over Pascal more than half-way into the Turing generation kinda proves those "keyboard warriors" correct, though. It's nice, but it was annoying when on release a large chunk of the press decided to sing songs about how new boundaries of performance were being pushed (true!) while downplaying how perf/$ remained still or regressed (equally true). Throwing up some straw men now doesn't change that. Reply
  • Spunjji - Monday, December 23, 2019 - link

    The 1060 6GB already beat out the 980 under most circumstances - at worst it was roughly equal. That was a very nice perf/$ improvement indeed for a single generation, and it's where we got most of the gains the 1650 Super is now building on.

    The 2060 is an instructive example of how the RTX series disappointed in that regard, as the cost increase roughly matched the performance increase and its RTX features are arguably useless.
    Reply
  • Kangal - Saturday, December 21, 2019 - link

    Thanks for the review Ryan.
    But I have to go against you on the mention of 4GB VRAM capacity for 2020. You have forgotten something very important. Timing.

    Sure, PC Gaming makes a lot more money than Console Gaming (and Mobile Gaming is even larger!!), but that is because the wealth is not distributed fairly, it's quite concentrated. Whereas the Console Market is more spread out, so publishers can make profits more universally and over a longer timeframe. On top of that, there's the marketing and the fear of piracy. Which is the reason why Game Publishers target the consoles first, then afterwards port their titles to PC... even though originally they developed them on PC!

    I needed to mention that above background first to give some clarification. Games for 2020 will primarily be made to target the PS4, and they might get ported to the PS5 or Xbox X. Or even those launch titles for the PS5/XbX, they will actually be made for the PS4 first, and had enhancements made. Remember the 2014 games which were still very much PS3/360 games?

    And it will take AT LEAST a full-year for the transition to occur. So games in Early 2022 will still target the PS4, which means their PC Ports will be fine for current day low-end PCs. I mean even with the PS5 release, the PS4 sales will continue, and that's a huge market base for the companies to simply ignore. And even in the PC Market, most gamers have something that's slower than a GTX 1660 Ti. Besides, low VRAM isn't too much of an issue, most of the time the game will only require 3GB RAM to run perfectly. If you have more available, say 8GB, then without any changes from your end, you will see it now start using say 6GB of VRAM. That's Double! And you didn't even change the settings! Why? Most games now use the VRAM to store assets it thinks it might use later on, so that it doesn't have to load them when required. This is analogous to how Mac/Linux uses System RAM, as opposed to say Windows does. If it does have to load them, performance will take a momentary dip, but perfectly playable.

    And even if the games now require more VRAM by default to be playable, in most cases that problem too can be solved. You can change individual settings one-by-one and see which has the most effect to the graphical fidelity, and how much it penalises your VRAM/RAM usage, and your framerates. I mean look at lowspecgamer, to see how far he pushes it. Though for a better idea, have a look at HardwareUnboxed on YouTube, and see how they optimise graphics for the recently released Red Dead Redemption 2 (PC) game. They fiddled with the graphics to get a negligible downgrade, but boosted their framerates by +66%.

    So I think 4GB VRAM will become the new 2GB VRAM (which itself replaced the 1GB VRAM), but that doesn't mean they're compromising on the longevity of the card. I think 4GB will be viable for the midrange upto 2022, then they're strictly just low-end. Asking gamers to get 8GB instead of 4GB for these low-midrange cards is not really sensible at the prices... it is exactly like asking the GTX 960 buyers to get the 4GB Variants instead of the 2GB Variants.
    Reply

Log in

Don't have an account? Sign up now