A few days ago, NVIDIA finally launched RTX IO. The suite of GPU-based decompression technologies was first announced in 2020, alongside the RTX 30 Series graphics cards, but it only debuted this week with the surprise release of Portal: Prelude RTX.
With RTX IO, games can offload 'dozens of CPU cores' worth of work to the GPU, which is much faster at this kind of work. This is part of a larger change in how the various hardware components access game data.
In the traditional Input/Output, the assets are sent from the hard drive to the CPU, decompressed through system memory, and only then fed to the GPU. However, the increase in asset size of recent games and bottlenecks between the CPU and the RAM make this system less than ideal, especially now that computers are equipped with NVMe SSDs.
RTX IO increases the I/O bandwidth by delivering compressed data directly to the GPU with only 'minimal staging' in the system memory. The graphics card decompresses the data by using the GDeflate open compression standard. This process is achieved at high throughput and frees up the CPU to take care of other processes.
As part of its announcement, NVIDIA put the focus on the faster texture load times and reduced disk space enabled by RTX IO. For example, in the slide below, NVIDIA highlighted RTX IO's 5X faster texture load times and 44% lower disk space in Portal: Prelude RTX.
However, given what we mentioned above regarding the CPU being freed of this job, we guessed RTX IO could help reduce stuttering. Indeed, when we asked about that, NVIDIA shared this statement:
RTX IO can be an aiding technology to improve stuttering by reducing the dependence on the CPU in loading textures and geometries and freeing it up to work on other tasks.
A
Read more on wccftech.com