AMD Trademarks “Infinity Cache” – A Radeon GPU Memory Breakthrough?

Last month, a YouTuber called RedGamingTech started reporting on a new Radeon technology called “Infinity Cache”, an innovative feature which enables higher levels of effective memory bandwidth on the company’s RDNA 2 graphics products. 

Thanks to @momomo_us on Twitter, we now know that AMD has trademarked the term “AMD Infinity Cache”, as can be seen on the Justia Trademarks website. This trademark adds an air of legitimacy to RedGamingTech’s prior reporting on the technology. That said, the patent mentions both CPUs and GPUs, a factor which could make Infinity Cache a generic AMD term that could be used to discuss multiple technologies. 

Recent rumours have suggested that AMD’s flagship RDNA 2 RX 6900 graphics card would feature 16GB of GDDR6 memory over a 256-bit memory bus, giving Radeon significantly less raw memory bandwidth than their Geforce counterparts. Nvidia’s RTX 3080 currently uses 10GB of faster GDDR6X memory over a wider 320-bit memory bus. 

If rumours regarding AMD’s Infinity Cache are correct, AMD has discovered an innovative technique which allows Radeon graphics cards to offer gamers competitive levels of memory bandwidth while using a smaller memory bus and slower DRAM than their rivals. If this is true, AMD’s RDNA 2 PC graphics architecture is genuinely groundbreaking. 

AMD’s Infinity Cache is a unique component of AMD’s revamped Radeon memory subsystem, enabling increases to raw memory bandwidth by acting as a faster cache for important data. In theory, this case reduces the need for AMD’s RDNA 2 graphics cards to utilise its GDDR6 memory, reducing the strain on the GDDR6 memory system’s limited bandwidth. At this time, no concrete information about AMD’s Infinity Cache is available.