ideally i would like Arm to make a better core than oryon and make it available to soc vendors like rockchip, mediatek, nxp, ti etc
how did nuvia out-Arm Arm anyway?
@mntmn Narrower focus is my guess 🤔
@epilys @mntmn not just that, but also taking care of system bottleneck analysis and properly sizing your memory hierarchy according to the system you're building. This is something you can do more easily when you control all aspects of the design top to bottom.
@mntmn by doing the same thing apple did: disregard ARM's stock designs and building their own. I never understood why ARM designs are so low-performance comparing to licensee ones. Maybe they just don't care, they got paid anyway!

@mntmn ARM maintains the ISA, but they have two license models.

The cheaper one gives a company access to particular predefined cores and other components to use in their chips. These interchangeable blocks have forced the blocks to be low-wattage jack-of-all-trades.

The architecture license means the produced chips have to conform to the ISA, but the ISA -> uOPs can be completely custom. Apple left out all 32-bit ISA, and added custom accelerators and tuning. Same engineers.

@min1123 ok, so you mean it's not a big deal but ARM just wasn't interested in maximum performance for laptop/desktop usecases?

@mntmn Unfortunately after Acorn, desktops/laptops were a niche market for them. They make money from licensing for selling billions of chips per year, and they kept getting bought and sold. New designs are still coming to market using the Cortex A53, announced in 2012.

They've never even intervened by applying a standard for reset vector, so all the bootloaders are different per chip. A last hard stance they took was HardFP and NEON in aarch64. Cortex A9 was a mess without it (see Tegra 2).