AMD’s FidelityFX Super Resolution 2.0 (FSR 2.0) expertise is coming later this yr, however earlier than it drops, AMD has outlined simply what makes its tremendous upscaling algorithm tick. A presentation at GDC outlined what’s improved with FSR 2.0 and in addition what has been optimised, and it is excellent news for RDNA 2 avid gamers in that regard.
AMD’s newest GPU structure hoovers up some thrilling new optimisations to run FSR 2.0.
There are heaps of enhancements coming with FSR 2.0—a full hour-long GDC discuss may solely nearly cowl it at a prime degree. However typically, you are improved picture high quality, velocity, antialiasing, and tons of optimisations.
These enhancements are available all styles and sizes: tonemap optimisations to release GPU assets, Cache Blocking to cut back the time spent wanting exterior of native reminiscence, and transforming the FidelityFX Single Go Downsampler to help in low-overhead Auto Publicity characteristic, to call just a few. There’s additionally a brand new characteristic to “lock” pixel large ridges in place to make sure skinny options do not flicker or disappear on-screen, similar to fences or distant objects.
All of which is able to work throughout AMD, Nvidia, and even Intel’s upcoming GPUs. FSR, and FidelityFX options typically, are open-source.
Although there are additionally particular RDNA-2 primarily based optimisations to enhance FSR 2.0.
For instance, there are particular RDNA-2 primarily based enhancements for Lanczos approximation, an underlying algorithm for FSR. RDNA 2 shaders are additionally in a position to make use of both Wave32 or Wave64 modes. Whereas FSR shaders run in Wave32 mode for essentially the most half, they’ll generally profit from Wave64, which implies RDNA 2 GPUs could possibly capitalise at instances.
RDNA 2’s Infinity Cache additionally sounds useful in some cases right here—FSR 2.0 can turn out to be fairly cache hungry with 4K frames. Even with a bigger cache, AMD says some spilling can happen. Therefore why it is utilizing Cache Blocking to assist alleviate congestion by splitting up workloads and operating all of them within the native space cache for elevated hit charges. That must also profit different GPUs with out a lot native reminiscence, too.
And simply because AMD is rolling out enhancements to RDNA 2-based GPUs, such because the RX 6800 XT, it is making certain that these enhancements will not hinder older Radeon GPUs or these from the competitors, it says.
To take action, there are numerous fallbacks in place. The optimisations that will hinder non-RDNA-2 playing cards are marked down and FSR 2.0, when configured accurately by AMD, ought to mechanically ignore them when operating on different GPUs. Wave measurement is the instance AMD provides, which is just set to make use of Wave64 when FSR efficiency will really enhance on appropriate GPUs in consequence.
FSR 2.0 feels like a serious development in AMD’s upscaling expertise, and I am hopeful it is akin to Nvidia’s leap in efficiency and high quality between DLSS 1.0 and DLSS 2.0. With that comes “vital” efficiency features over FSR 1.0, although additionally has its drawbacks. AMD says that FSR 2.0, not like FSR 1.0, is just really useful to be used on pretty latest and highly effective GPUs. The Nvidia GeForce GTX 1060, for instance, isn’t really useful by AMD to be used with FSR 2.0.
Whereas some avid gamers with older tech might not profit as a lot as others from FSR 2.0, AMD is bringing FSR 2.0 to Xbox with Microsoft in order that console avid gamers may give high-quality upscaling a go.
