That was my thought when I read it as well, and I have to scroll all the way down to find this comment.

AMD said part of the reason why they partner with Samsung was because they dont compete in Mobile Phone. Which is a valid point. But now Samsung is competing directly in Laptop and PC using these tech. As a tech enthusiast of course I want to see this ARM / AMD Graphics combination. But as an investor I am thinking If we were mislead in a way.

And if this true does it mean AMD is open to license AMD graphics IP?

> And if this true does it mean AMD is open to license AMD graphics IP?

This is not the first time this happens. There was an Intel CPU with AMD's GPU.

https://ark.intel.com/content/www/us/en/ark/products/130411/...

AMD didn't license their IP to Intel. It was a collaboration with AMD GPU fabbed on TSMC and linked together via Intel EMIB.

It could be a good market for AMD. There a tons of companies designing their own ARM processors, but who don't have the resources to also create a GPU. Nvidia is completely absent from that market, at least currently, so it might be a easy way for AMD to push their GPU design.

From a business POV yes. I just dont like companies that lie. Of course at the moment this is still a rumour.

Hopefully this means ARM chips will finally have a good GPU option that has high quality open drivers

It's kind of amazing that Nvidia is so widely reviled for frustrating agonizing drivers filled with blobs, while ARM has been around providing no help, no docs, no assistance, no drivers, what-so-ever for over a decade. No one's been thrilled, but basically doing nothing has been more effective PR than doing a so-so job.

The situation has changed somewhat recently, but basically only after the reverse engineering efforts succeeded & were fully operational.

ARM has plenty of docs, sources, etc. for its Mali GPU: https://developer.arm.com/ip-products/graphics-and-multimedi... Their GPU drivers are even distributed by Debian: https://wiki.debian.org/MaliGraphics

You are probably conflating vendors like Broadcom who license the ARM ISA and processor IP but add their own (closed) GPU to SoCs they produce, like those used in the Raspberry Pi.

Remember ARM is just a company that license and sell an ISA and chip design. It's up to system integrators to turn those designs into physical chips, and sadly we have seen many of them don't value making their designs available to open source, etc. The solution to this is to not support those vendors, not to complain at ARM.

I've spent about 10 minutes clicking through looking for anything even mildly technical. The first link you sent me, "plenty of docs", is a 2 page PDF product brief & 5 web pages describing some features at a high level. There's a ton of docs... for game developers, application developers, &c. Talking up features & capabilities. But for making drivers? For describing how to use the hardware? There's some links to embedded development / DDK development sites for arm, but none of them are open accecss. ARM wont help.

The Debian drivers you cite are the Panfrost Lima drivers, which are the reverse engineering works everyone else did in the dark with no help from ARM, until very very recently.

I am not conflating anything. ARM has provided no documentation no support to help folks make drivers. I stand by my original post entirely, and disagree deeply. ARM's gpu chip design is stamped out into countless chips, verbatim or close to verbatim. Even if individual chip makers did want to support their chips, I suspect it is in fact ARM who would block them from releasing open source drivers, from documenting the chips they make: ARM's desire has been to protect it's IP, and to help the various embedded support partners that do know how to work with their chips and their IP in a secure fashion. ARM has been a bad player in the Linux world, has worked against allowing Linux support. The problem is absolutely ARM, not those vendors.

Can someone explain to me how ARM expect to protect its IP if it doesn't publish anything about it?

I was under the impression that in order to protect IP, one has to either patent it or copyright it, thus some form of publication must be made. If ARM does publish technical documentation (1), ARM has a way to prove infrigement or prior art whenever someone else tries to use their IP without their approval. If ARM doesn't, how can they prove anything?

(1) Obviously, if they publish the tech. docs, they might as well help driver implementors, might they not?

Mali is every bit as proprietary as Nvidia.

> Mali is every bit as proprietary as Nvidia.

I don't think anyone was claiming it isn't. I think GP's (and his GP) point is that ARM /usually/ makes the information needed to develop high quality open source drivers available where as Nvidia tends to do the opposite.

Whether these technologies are proprietary or not is orthogonal to this issue.

But it absolutely doesn't, that is the point. There is no register level documentation or documentation for the instruction set the chips use. There is no pertinent architecture documentation at all. Having a list of performance counters does not equal documentation.

Android is a hellscape of proprietary drivers but desktop Linux is basically down to one: Nvidia. It stands out more because it's the last holdout.

Desktop Linux on ARM is in a similar situation to Nvidia, with the proprietary Mali drivers from ARM vs the incomplete reverse-engineered Panfrost / Lima open source drivers.

Qualcomm is pretty widely disliked for doing the same. Google even called them out for creating security problems by making updates hard (although they weren't quite as harsh as Linus was.)

Amazing what they were able to do in such a short period of time after completely stopping R&D on both their custom CPU and GPU in 2019, opting to instead form a strategic partnership with AMD for the new exynos chips.