AMD has recently filed a patent to spread the rendering load across several GPU chips. A game scene is split into separate blocks and distributed across chips to optimize the use of shaders in games. Two-level chiplet binning is used for this.
AMD issues patent for GPU chips implementation to make better use of shader technology
The new patent, published by AMD, provides further insight into what the company plans to do with next-level GPU and CPU technology in the coming years. At the end of June, it turned out that fifty-four patent applications had been submitted for publication. It is not known which of the more than fifty published patents will be used in AMD’s plans. The practices discussed in the patents detail the company’s approaches in subsequent years.
An app spotted on the website by community member @ETI1120 Computer Basepatent number US20220207827discusses critical image data in two stages for efficiently passing loads from a GPU across many chips. CPU first applied it to the US Patent Office late last year.
When image data on a GPU is rasterized in standard ways, the shader unit, also known as the ALU, performs the same task and assigns a color name to individual pixels. In contrast, textured polygons located at a particular pixel in a particular game scene are mapped directly to the pixel. Finally, the formulated task will retain the atypical principles and differ only through other textures located in different pixels. This method is called SIMD or Single Command – Multiple Data.
For most current games, shadowing isn’t the only task a GPU does. But instead, several post-processing elements are included after the initial shading. The actions the GPU will add will be, for example, anti-aliasing, shadowing, and turning off the game environment. However, ray tracing happens along with shading, creating a new computational method.
Speaking of the GPU that controls the graphics in current games, the load created by the computer is exponentially increasing to thousands of computational units.
In games on GPUs, this computing load ideally scales up to a few thousand compute units. This differs from processors in that applications have to be specially written to add more cores. The CPU scheduler generates this action, dividing the work on the GPU into more digestible tasks handled by the compute units, also called bundling. The image from the game is created and then divided into individual blocks containing a certain amount of pixels. The block is computed by a subunit of the GPU, where it is synthesized and rendered. After this process, pixels waiting to be computed are included in a block until the graphics card’s subunit is finally used. The shaders’ computing power, memory bandwidth, and cache sizes are taken into account.
In the patent, AMD explains that splitting and combining requires a complete and complete data connection between all elements of the GPU, which poses a problem. Data connections that are not on the pattern have a high level of latency, resulting in slower processing.
CPUs have made this transition to chips effortlessly due to their ability to send the task over several cores, making them accessible to the chips. GPUs don’t offer the same flexibility, making their timers comparable to a beginner dual-core processor.
AMD is aware of the need and tries to address these issues by changing the rasterization pipeline and sending tasks between several GPU chips, similar to CPUs. This requires advanced splitting technology, which the company has introduced a “two-level split”, also known as a “hybrid split”.
In hypergrouping, splitting is handled in two separate stages, rather than processing it directly in pixel-by-pixel blocks. The first step is to calculate the equation, take a 3-dimensional medium and create a two-dimensional image from the original. The stage is called vertex shading and is completed before rasterization and processing is pretty minimal on the GPU’s first chip. Once complete, the game scene begins to be grouped, developed into rough boxes, and rendered into a single GPU chip. Then routine tasks such as rasterization and post-processing can begin.
It is unknown when AMD will start using this new process or whether it will be approved. However, it gives us a glimpse of the future of more efficient GPU processing.
News Sources: Computer Base, Free Patents Online
