Jesper Berg At Computex 2022, AMD revealed the name of its implementation of Microsoft’s Direct Storage technology – AMD Smart Access Storage, or SAS. This is AMD’s answer to Nvidia’s RTX IO, and just like its Nvidia counterpart, it is all about reducing loading times for games.
Faster loading times are achieved mainly through GPU-assisted decompression of, e.g., game textures in a much more efficient way than has previously been possible.
Using Smart Access Storage, the graphics card can communicate directly with an NVMe SSD to load game assets, while using the graphics card’s compute units (CUs) instead of the system’s processor (CPU). This in turn increases performance and also frees up conventional resources in the form of CPU load as well as reduced memory usage.
AMD Smart Access Storage looks set to become the fastest implementation of Microsoft’s Direct Storage API to date when combined with AMD’s recently unveiled Ryzen 7000 series. The next-gen platform will include 24 PCI Express 5.0 lanes, whereas Nvidia is expected to stay on PCI Express 4.0. Also, Intel’s Alder Lake platform is limited to 16 PCI Express 5.0 lanes.
Consequently, this will be a key selling point for AMD’s upcoming hardware by the end of the year, as it will then be possible to pair a Ryzen 7000-series CPU with a next-gen RDNA 3-based graphics card, to greatly enhance storage performance over PCI Express 5.0.
What Real-World Improvements to Expect?
Data on real-world loading time reductions using AMD’s SAS using the aforementioned hardware and a high-end PCIe/NVMe SSD is still scarce. What we do know is that it is based on Microsoft’s Direct Storage technology, which is already used in the latest Xbox consoles.
In one example showcasing the game Forspoken by the developer Luminous Productions, Direct Storage and an (unnamed) NVMe SSD appears to almost cut loading times in half compared to a SATA SSD – from 3.7 to 1.9 seconds in a specific scene. A mechanical hard drive, by contrast, uses 21.5 seconds to load the same scene.
