NVIDIA GeForce RTX Explored: What You Need To Know About Turing

NVIDIA has a number of other optimizations and upgrades coming with Turing and the GeForce RTX 2000 series cards as well. The video engine has been upgraded, the SLI connectors have been nixed in favor of a higher-bandwidth, more flexible interconnect dubbed NVLink, and overclocking should be more robust too, thanks to the better power delivery characteristics on GeForce RTX cards. In addition, we're now being introduced to NVIDIA Scanner, a one-click overclocking tool that will eventually be coming to other cards in NVIDIA’s line-up as well…


The new video engine in Turing offers HEVC 8K30 HDR encoding in real time, along with VP9 10/12b HDR and HEVC 444 10/12b HDR decoding. The video engine is also more efficient and results in lower CPU utilization than previous-gen architectures for streaming to sites like YouTube and Twitch, which is ideal for streamers who may be gaming, as well as processing and managing their stream from a single system.

GeForce RTX 2080 and RTX 2080 Ti cards are also outfitted with a new inter-GPU interconnect dubbed NVLink that replaces the older SLI connectors on previous-gen cards. Whereas SLI essentially acts a display bridge and offers limited bandwidth, NVLink is both a memory and display bridge with up to 100GB/s of bi-directional bandwidth at its disposal.

The GeForce RTX 2080 has a single-channel NVLink bridge with 50GB/s of bandwidth. The RTX 2080 Ti, however, has a dual-channel setup that pushes available bandwidth to the aforementioned 100GB/s. NVLink is still compatible with SLI as we know it, but we’re told that fatter pipe between the GPUs lays the groundwork for some future multi-GPU goodness that’s yet to be talked about and enables much higher resolution, multi-monitor surround configurations at higher refresh rates as well. The DP 1.4a ports on the GeForce RTX series cards support 8K resolutions at 60Hz – with NVLink, 8K displays can actually be used in a surround configuration.

NVLink bridges will be available for about $79 directly from NVIDIA or through OEM partners for two-way GPU configurations, with three-slot or four-slot spacing to accommodate various motherboards and cooling configurations.


NVIDIA has given overclocking some attention as well. We’ve already mentioned the beefier VRMs on GeForce RTX cards versus their predecessors. In addition to being able to handle higher loads, the circuitry has been optimized to smooth and clean-up power delivery across the entire voltage and frequency curve. That cleaner power should help with overclocking, but NVIDIA is also making the process itself much easier with its new Scanner tool and API.

The NVIDIA Scanner is a one-click overclocking tool with an intelligent testing algorithm and specialized workload that helps users find the maximum, stable overclock on their cards without having to resort to trial and error. The NVIDIA Scanner will try higher and higher frequencies at a given voltage step and then test for stability along the way. The entire process takes around 20 minutes, but when it’s done, the Scanner will have found the maximum stable overclock throughout the entire frequency and voltage curve for a given card.


The NVIDIA Scanner is already built-into an upcoming version of EVGA’s Precision utility, but will be coming to others as well. It will be offered on some previous-gen GPUs in the future too, so it’s not limited to just Turing.

And if you’re wondering, NVIDIA tell us 2.1GHz+ overclocks should be possible with some GeForce RTX cards, though we haven’t completed our own testing just yet – stay tuned for that, it's coming soon enough.