Nvidia is telling anyone who will listen that enabling DLSS (Deep Learning Super Sampling) on its new GeForce RTX series graphics cards will result in an "enormous leap in performance," and to drive the point home, the GPU maker has provided a fresh set of benchmarks. Unfortunately, we're still waiting on DLSS games to arrive to test the performance claims for ourselves.
That day will come. In the meantime, Nvidia is serving up benchmarks purporting to show how its GeForce RTX 2080 Ti, 2080, and 2070 cards compare to the previous generation, with DLSS enabled on the new parts and TAA (temporal anti-aliasing) on the older ones.
"In each and every series, the Turing GPU is twice the performance," Nvidia CEO Jensen Huang said. "This is a brand new way of doing computer graphics—it merges together traditional computer graphics and deep learning into a cohesive pipeline."
Assuming the data is accurate and applies directly to future DLSS games, Huang is correct in his assessment, at least in so far as the model numbers align. However, pricing throws a wrinkle into things. If aligning the cards by price, it's more accurate to compare the GeForce RTX 2080 Ti to the Titan, 2080 to the 1080 Ti, and so forth. Performance is still better across the board, according to this single benchmark, but not twice as good.
This is part of the promise of Turing in general, that the new GPU architecture will push gaming forward with deep learning and real-time ray tracing capabilities. It would be easier to accept the pricing premium if those features live up to the hype, when the games arrive. We just won't know until they do.
For what it's worth, there are more than two dozen games supporting DLSS on the horizon.