The GeForce GTX 1660 Ti is NVIDIA's answer for the highly competitive sub-$300 segment. The card is based on the all-new TU116 graphics processor that has been specifically designed to meet the demands of that market, which is mostly "price". Unlike other Turing GPUs, TU116 does not feature acceleration for RTX real-time raytracing or DLSS because the specialized hardware consumes a significant portion of the die area on other Turing GPUs, which increases manufacturing cost significantly. NVIDIA did keep the other improvements of Turing though, like GDDR6 memory, larger caches, concurrent execution of float and integer operations, and adaptive/variable rate shading.

As a result, when averaged over all our gaming benchmarks at 1080p, we see the GTX 1660 Ti beat the Pascal-based GTX 1070 and roughly match AMD's RX Vega 56, which is pretty impressive for a mid-range card. While we don't have a GTX 1660 Ti reference-design, we expect this card to perform within 2–3 % of one, which isn't much, but follows what we've seen on factory-overclocked cards for the other Turing GPUs. A reference-design card would give us more precise numbers on the performance increase the factory-overclocked MSI GTX 1660 Ti Gaming X offers. Compared to the RTX 2060, which is NVIDIA's next-fastest SKU, the MSI GTX 1660 Ti Gaming X is 13% behind. Compared to the GTX 1060 6 GB, which the GTX 1660 Ti replaces, the performance uplift is 40%, although at a higher price point. AMD just released the RX 590, on a 12 nanometer process no less, to address the growing requirements of the mid-range segment, and the GTX 1660 Ti makes short shrift of that, offering 30% more performance. With those performance results, the GTX 1660 Ti is a great choice for gamers running maximum details with a Full HD monitor. If you are willing to dial down detail settings a bit, it should be able to reach 60 FPS at 1440p in most titles, too.

MSI is using a large thermal solution on their flagship Gaming X card. The cooler, which seems very similar to the GTX 1060 Gaming X, does a good job at keeping the card cool. Temperatures only reach 67°C during gaming. At the same time, the cooler is very quiet with 32 dBA, although it could be quieter. On the GTX 1060 Gaming X, we saw 28 dBA with similar heat output and a similar cooler. MSI's Gaming X is the only card reviewed today that includes the highly popular idle-fan-stop feature, which turns off the fans completely during idle, Internet browsing, or light gaming.

Compared to Pascal, Turing improved power efficiency once more, and the GTX 1660 Ti is no exception. Actually, it is the most power-efficient GPU design we've ever tested. MSI has increased their board power limit to unlock additional performance, which does cost a little bit of efficiency, but their card is still leading our performance per watt charts. Unlike other models, you may manually increase the board power limit by an additional 8% to bring it up to 140 W.

Manual overclocking worked well, netting around 10% in real-life performance, most of which is due to the enormous overclock these GDDR6 memory chips can take—we went from 1500 MHz to 1900 MHz. GPU overclocking reached an average clock of 2045 MHz when gaming, which is the highest of all cards tested today.

The lack of ray-tracing and DLSS on the GTX 1660 Ti seems like a big deal at first, especially considering how much NVIDIA is pushing those technologies. While both are extremely promising, they are not the most important thing to have right now, especially when every dollar matters. While I have no doubt that RTX support will be growing vastly, only a few titles support it at this time, so I don't think anyone could be blamed for skipping the tech for now and waiting for it to mature more. Looking at silicon economics, including RTX/DLSS would have either driven die size (= cost) so high that reaching the targeted sub-$300 price point wouldn't be possible anymore, or the number of shader cores would have had to be reduced, which would have resulted in no performance improvement over AMD—NVIDIA went the other route. "RTX" is a great vehicle for NVIDIA to sell a feature that is more than just "higher FPS". Specifically, the GTX 1660 Ti targets people who are comfortable with 1080p 60 FPS and even willing to play at lower details if their GPU is affordable. If, however, they feel they miss out on visuals, something that DX12 couldn't achieve, they might be willing to spend more and go for the RTX 2060, for example. The next step for NVIDIA is transitioning to the 7 nm production process, which increases density and should allow smaller GPUs to have RTX, too. I also wouldn't be surprised to see NVIDIA haters on the forums who previously dismissed RTX as "useless" now ask "wut no RTX?"

Priced at $299, the GTX 1660 Ti Gaming X comes at a $20 premium over the NVIDIA MSRP, which I find to be reasonable since you get a better cooler, idle-fan-stop, RGB lighting, and a bigger factory overclock. Looking at the whole market, the GTX 1660 Ti does compete with the RX 590, which it beats with ease: 30% faster, almost half the power consumption, similar noise. The RTX 2060 starts at $350, is 15% faster, and offers NVIDIA's RTX features, like ray-tracing and DLSS, which definitely make it an option if you can save up more money and are looking to play at 1440p. Strong competition has come in the past few days from AMD Vega 56, which has seen its price dropped to $330 (€269 including VAT in some European countries). This of course makes a compelling argument for people looking for price/performance, but Vega 56 runs much hotter and noisier than GTX 1660 Ti. Looking at the board designs, I have no doubt that prices for the GTX 1660 Ti can go down much further, so I wouldn't be surprised to see it at or below $250 soon, at which point it would spark the upgrade-itch for many users, especially those with older cards, like the GeForce 900 series.