Just yesterday we told you that Samsung is going to end its contract with AMD because it wants to develop its own GPU, information that ended up being the result of a misinterpretation and which has just been completely dismissed. The agreement between the two companies will continue, as will the project of the former create a “custom GPU”.

I put quotes because actually Samsung it will use AMD graphics architecture to create said GPU, which tells us something very important, as I advanced in the title, that a South Korean wouldn’t have the experience, resources, or talent to create their own graphics architecture to shape into a fully fledged custom GPU.

Honestly, we have to keep in mind that in fact Apple also wasn’t able to build its own GPU completely from scratch. Although the topic is very much under wraps, it is clear that the core architecture that Apple used to develop its graphics cores, both in the A-series and M-series SoCs, was a GPU architecture. PowerVR by Imagination Technologies.

Starting from this foundation, Apple introduced a significant degree of customization and shaped the GPU we all know, and honestly I think it’s only natural that Samsung decided to do the same, since they ended up developing the GPU from scratch. It is something extremely complexboth at the level of software, patents (intellectual property), software, drivers and support.

It is expected sometime in October this year Samsung LSI Tech Dayan event in which Martin Ashton, Vice President of AMD, will be in attendance, which will offer more detailed information about this collaboration between the two giants. We’ll be keeping an eye out, as it’s very likely that this event will offer more details about the architecture Samsung will use, as well as the keys to its next custom GPU.

Personally, I think Samsung I should go for RDNA3because it is a more complete, efficient and interesting option due to the improvements it brings in terms of performance, and also because it comes with dedicated AI cores, which could make a significant difference, especially considering the weight of NPUs on mobile devices.

I remind you before I finish We should not confuse this custom GPU project with the Xclipse 940, which will be integrated into the Exynos 2400 SoC. The first is a much more ambitious project, while the second will be more of a “transitional solution” that will serve Samsung to further gain experience in developing a more advanced and powerful solution and with a greater degree of customization.