Intel’s XeSS is comparable to Nvidia’s DLSS 2.3
Intel’s XeSS, as an AI-driven super-sampling, is accelerated by XMX AI in the Xe-HPG architecture Xe Core (Xe Core), using ML to reconstruct low-resolution frames into high-resolution frames, increasing the number of running frames. XeSS is fully DirectX 12 compatible and, thanks to DP4a instructions, provides AI-based supersampling on a variety of hardware including integrated graphics.
Recently, Digital Foundry obtained the exclusive right to use Intel’s XeSS technology and conducted its first test with the Tomb Raider game, which is one of the first games to support XeSS technology. In fact, at the beginning of this year, Raja Koduri, Intel’s senior vice president posted a picture on social platforms, using the system built by the NUC 11 Extreme codenamed Beast Canyon to run “Tomb Raider” and enable XeSS technology.
Digital Foundry conducted a comparative test using XeSS 1.187, DLSS 2.3, and FSR 2.0 at different resolutions and presets, using the graphics cards of ARC A770 and NVIDIA’s GeForce RTX 3070. Among them, XeSS has four preset modes, namely Ultra Quality, Quality, Balanced, and Performance.
Image: Digital Foundry
Image: Digital Foundry
Image: Digital Foundry
The test results show that if Ultra Quality is selected, there are 23% and 16% improvements in 4K and 2K resolutions respectively. If you choose Performance, the increase in 4K and 2K resolutions will be increased to 88% and 52%. If you select Performance and turn on ray tracing, the improvement at 4K resolution will reach 177%.
Digital Foundry’s test also compared the image quality between XeSS 1.187 and DLSS 2.3 and provided comparison screenshots.
Image: Digital Foundry
Overall, Intel’s XeSS is comparable to Nvidia’s DLSS 2.3, although there are still some rendering issues with XeSS that hopefully will be resolved before XeSS is officially released.
