Sometimes, when companies think it's time for the world to move on to a new standard or format, they ruthlessly kill off older products or services and force people to change. The most famous example of this is of course Apple, with its relentless deprecation of floppy and optical drives, all ports other than USB Type-C, and most frustratingly, the 3.5mm headphones socket. On the other hand we have examples of about-face reversals, such as Microsoft's universally hated “Modern” Windows 8 interface, its abysmal adoption rate as people refused to leave Windows 7, and the company's resulting campaign of apologies and free Windows 10 upgrades.
Nvidia seems to have found itself in the latter camp. The company proudly unveiled its GeForce RTX series late last year, hoping that the allure of ray tracing would create a rush of demand for new, expensive graphics cards. That hasn't been the case. The GeForce RTX 2080 Ti, 2080, and 2070 have been praised for their performance, but even now, barely any games support ray tracing and those that do have very limited implementations for specific effects such as shadows in Shadow of the Tomb Raider and reflections in Battlefield V.
DLSS (Deep Learning Super-Sampling), the second major feature of the GeForce RTX series, has also been received poorly. This feature uses a neural network to help intelligently upscale graphics rendered at low resolutions, but image quality suffers in real-world tests, and again, very few games actually support it.
Nvidia has found itself in a sort-of chicken-and-egg situation — more games should support these features in the future, but that isn't going to motivate people to upgrade now, and the GeForce RTX series cards are much too expensive to be bought as impulsive upgrades. Now, production and distribution of the GeForce GTX 10-series is reportedly winding down, leaving a gap in the market.
Enter the brand new GeForce GTX 1660 Ti, presumably the first of several models in a new series. This oddly-named specimen uses the same Turing architecture as the GeForce RTX 20-series, minus the ray tracing capabilities. This GPU can be considered a cut-down version of the recently launched GeForce RTX 2060 and is aimed at the mid-range market, typically gamers with a GeForce GTX 970 or GeForce GTX 960 graphics card who are ready to upgrade.
The high prices of GeForce RTX series cards has left them without a viable option — till now, that is. We have an Asus ROG Strix GeForce GTX 1660 Ti, and if you've been waiting to upgrade or build a new gaming PC, we'll tell you if this is the answer to all your problems.
Nvidia GeForce GTX 1660 Ti specifications and features
First of all, let's talk about this name — the GeForce GTX 1660 Ti has no precedent. As the GeForce GTX 10-series was due for a refresh, it was heavily rumoured that the next generation would be called the 11-series. Now that we have a 20-series instead, Nvidia has decided that its cut-down models will jump to 16xx.
Gadgets 360 asked an Nvidia representative about this in a pre-launch meeting, and we were told “16 is closer to 20 than 10”. That's really all there is to it, and so much for continuity. At least that leaves room for a few more non-RTX generations. More interestingly, the use of the ‘Ti' suffix predicates the eventual introduction of a hopefully even more affordable GeForce GTX 1660, which is already being rumoured.
The GeForce GTX 1660 Ti might also be the most heavily leaked PC component in recent history. We've seen everything about it from specifications to benchmark scores, renders, and even packaging from multiple partner brands. If you aren't caught up, here's what you need to know.
The GPU itself is codenamed TU116, and is not just an existing GeForce RTX GPU with sections disabled. It's been fabricated from scratch without RT cores or Tensor cores. For an overview of the Turing architecture and what these components do, check out our complete overview of the architecture.
This makes sense for the price segment that Nvidia has targeted — even the high-end GeForce RTX models reportedly struggle when ray tracing is enabled, which means that a low-end GPU would not be able to handle that load anyway.
The TU116 GPU has 1,536 compute units called CUDA cores, arranged into 24 clusters called Streaming Multiprocessors. It has a base clock speed of 1500MHz and a maximum boost speed of 1770MHz, but board partners are free to tweak these values. You get 6GB of GDDR6 RAM running at 12Gbps on a 192-bit bus for up to 288GBps of peak memory bandwidth. The total TDP is a very manageable 120W
