Intel may use a Large L3 Cache in its Xe DG2 GPUs, Similar to AMD’s Infinity Cache

According to a source, Intel may yet again take a leaf from AMD’s rule-book in designing its Xe-HPG (High-Performance Gaming) graphics cards, leveraging a large L3 cache quite similar to the Infinity Cache used in the recently launched Big Navi GPUs.

The L3 cache on the HPG GPUs is likely going to be up to 32MB considering that these GPUs will target the mainstream market, rather than compete with the likes of the RTX 3080 or Big Navi.

The reason for this decision is the target audience. Intel is going to target the RTX 3060-3070 level GPUs which are mainly used for gaming at 1080p and 1440p and are the most sold products as well. As shown by the performance of the Big Navi GPUs, the Infinity Cache is much more effective at lower resolutions where the cache hit rates are close to 70%.

Intel also believes that upscaling will become quite a standard even in PC gaming with the launch of AMD’s Super Resolution technology next year. Both the Xbox Series X and the RDNA 2 GPUs will use a DirectML based upscaling technology with open standards as part of AMD’s FidelityFX suite.

If AMD can bring it on par with DLSS 2.0 in terms of quality and performance, it’ll certainly be the go-to solution for most developers as it’s not only Open Source and doesn’t require any physical hardware or license.

Low pixel counts paired with AI-based upscaling techniques should allow Intel’s HPG graphics cards to be competitive in the midrange market despite having a weak network of game optimizations, all thanks to the raw compute performance and high hit rates.


Computer hardware enthusiast, PC gamer, and almost an engineer. Former co-founder of Techquila (2017-2019), a fairly successful tech outlet. Been working on Hardware Times since 2019, an outlet dedicated to computer hardware and its applications.
Back to top button