AMD ready to flex its ARMs | Hacker News Reader

AMD ready to flex its ARMs

amd-now.com

51 pointsbasilgohar15 comments

blogspam of: https://www.tomshardware.com/news/amd-we-stand-ready-to-make...

Via https://wccftech.com/x86-champion-amd-is-ready-to-make-arm-c... no less. I link both sources.

The main additional point I raised was linking the timing of the announcement to nVidia's struggles with their Arm acquisition, which I felt was relevant and hadn't found mentioned elsewhere.

That is indeed interesting. I didn't realize Nvidia's acquisition may be stalling due to the EU. I had rather hoped it would go through in order to create three viable companies offering both state-of-the-art CPU and GPU options.

The acquisition by nVidia is problematic from numerous fronts, not the least of which is the relevant concerns raised in the EU case that this would be leverage nVidia would use against its own competitors and monopolizing the ARM architecture for their own purposes.

Arm as an independent company is better for the broader semiconductor industry because of widely their technology is licensed, and the current ease at which it does so. These are risked by any kind of an acquisition, let alone nVidia who is not famous for their generosity of their technology (CUDA, G-Sync, Linux drivers, DLSS, etc. are all proprietary).

Contrast this with AMD's approach on such matters, where they offer theirs with liberal licensing and/or for free: FreeSync (which is just VESA Adaptive Sync),FSR, Linux drivers, their backing of OpenCL, and more.

Edit: Added AMD counterpoints to nVidia.

Hmm, plenty of other architectures to go to if ARM gets hard to license, IMO. In fact, there are already plenty of people using RISC-V who would have otherwise used ARM. Not sure if it's even beneficial for the semiconductor industry having ARM the way it is right now.

I work in embedded systems and ARM doesn't offer much benefit to me, so maybe I'm biased.

I'm not opposed to this. I'm all for a more open architecture. I think AMD's already got their foot in the door with regards to ARM, though. Moreover, ARM is already dominant in mobile- and low-power markets.

It'd be great if they also explored alternative (and more open/libre) architectures like OpenPOWER and RISC-V, but I feel like they're not headed in that direction at this time – not sure what that would buy them in comparison to ARM.

I think technical superiority has rarely been the deciding factor in architecture dominance – otherwise, x86 likely never would have dragged along this far.

Interesting. So, Qualcomm has forever basically relied on their GPU prowess to remain king of mass market mobile chips. Mali doesn't really compare.

As I remember (which may be wrong), AMD designed Adreno for Qualcomm. Also, they're apparently working with Samsung now to design a GPU for Exynos.

While not explicitly clear, wondering if this a potential play to compete directly with Qualcomm et al in that space, or purely server oriented.

AMD has demonstrated a strong desire and commitment to custom semiconductor designs beyond the traditional desktop, server, and mobile segments – this is best demonstrated in their multi-generational design wins in the recent few Playstation and XBox consoles. And there are more examples if we look a little further.

Qualcomm has mostly relied on its integrated modem to remain king. It is good in other things as well cpu/gpu/adsp/pmic etc. in general overall solution rather than the chip.

AMD did not designed adreno for Qualcomm. AMD acquired ATI and as part of acquisition it also got mobile gpu(adreno) which was designed for handhelds. It later decided it does not have much use of it and Qualcomm acquired the ip along with people at peanuts.

Any more info? From what I can tell AMD acquired Imageon in 2006, and sold to Qualcomm in 2009. Adreno is an anagram of radeon, afterall.

I wish they'd "stand ready" to make their own GPUs in sufficient quantities. Jokes aside, ARM chips are great but I'd also like to see what they could do with RISC-V.

AIUI "not x86" was assumed to be ARM.

I wouldn't be surprised if it was not.

It's a shame AMD never kept up making its own "Seattle" ARM chips. They looked very promising (and, in theory quite cheap!)

They even had relatively quick full upstream kernel support, absolutely unheard of in the ARM world

I don't know. The AMD A1100 series was underwhelming when compared to their competition.

I'd be interested in a high-clockspeed desktop chip based on ARM's X2 design.

X2 should be nearly as fast as Zen per clock. It could be a great dev platform if they could ramp the clocks up.

Reply on news.ycombinator.com