AMD Unveils Their Embedded+ Architecture, Ryzen Embedded with Versal Together | Hacker News Reader

AMD Unveils Their Embedded+ Architecture, Ryzen Embedded with Versal Together

anandtech.com

69 pointsteleforce14 comments

Note that this is an expansion to the existing Ryzen Embedded R2000 series from 2022 - no new SoCs announced here.

I'm a bit confused why this doesn't (also) target Ryzen Embedded V3000. R2000 taps out at 4C/8T while V3000 has options up to 8C/16T.

Interesting, but is x86_64 as an arch used much in any embedded applications?

Anything I've worked on for the last 10+ years has been ARM variants (with one MIPS device I prefer to forget)

Teslas are X86, and so are other car infotainment systems.

> and so are other car infotainment systems

[[citation needed]]

Worked on a number of them - it was basically always ARM. 90% of them are covered by nvidia + qualcomm

Yeah, a lot of small deployments(<100, <1000) use amd64 because they can buy them in small quantities and they don't need an embedded developer to hack something together that ships on the hardware. I personally think arm64 is often a better choice for these kinds of systems but it's also hard to find people who are comfortable with non-Windows systems in these spaces and you have to go that direction if you go arm64.

So you'll see a lot of cheap, technically "embedded," Windows systems that are sitting in places driving displays or providing kiosk functionality because the economics work a lot better than trying to find an embedded Linux developer and build an arm64 system around the same thing.

Interesting, thanks. Last time I encountered x86 in an embedded fashion was ~25 years ago (PC/104) at a job I worked at.

>Interesting, but is x86_64 as an arch used much in any embedded applications?

A lot. ATMs, MRIs, digital signage, etc.

Higher end oscilloscope, signal generators and similar. A lot of the world runs on Windows Xp Embedded or very quirky Linux distros.

Many industrial applications are x86

I heard one time someone make the the distinction between 'big E' Embedded and 'little e' embedded. Little e is classic embedded where you need some computation next to inputs. That's generally where you see talk of real time. Then there's big E where you have devices that you aren't really supposed to interact with via a keyboard, screen, or shell prompt for the most part, but have closer to server compute, like storage appliances or high end car head units.

Pretty much. If you can put x86 on your (big E) embedded thing, you can shortcut so many classic embedded time- and money-pits. Look up VPX for example for rugged PC platforms. But same here, if you can do without (industrialized PC, 'edge' stuff or even COTS server...) you can shave so much cost, in sometimes very competitive markets.

Edit: of course this works for low-volume. If you're getting millions of objects out the door, it's another story...

That's a helpful distinction. Another to add to that is that both classes of E(e)mbedded system are designed to be used with a single application (i.e. the user isn't running arbitrary software).

I have seen this in some NAS systems. It's a pretty good fit, especially for the ones that can be upgraded to 64GB RAM and run VMs or docker.

Generally plenty of RAM plus many fast PCIe lanes is not something most ARM chips offer.

Afair recent analysis of russian cruise missile revealed its running Ryzen based pc inside.

Reply on news.ycombinator.com