AMD is expanding its processor portfolio beyond the x86 architecture with its first ARM-based APU, internally known as “Sound Wave.” The chip’s existence was uncovered through customs import records, confirming several details about its design and purpose. Built with a BGA-1074 package measuring 32 mm × 27 mm, the processor fits within standard mobile SoC dimensions, making it suitable for thin and light computing platforms. It employs a 0.8 mm pitch and FF5 interface, replacing the FF3 socket previously used in Valve’s Steam handheld devices, further hinting at a new generation of compact AMD-powered hardware.
↫ Hilbert Hagedoorn at The Guru of 3D
It only makes sense for AMD to enter the market for ARM SoCs, as it’s a whole section of the processor market they’re not tapping into. Even if they don’t manage to compete with the best ARM processors out there, they can still serve the mid and lower end just fine.
And it looks like American companies are leaving the entire RISV-C field to Chinese ones.
Yes, ARM is great, and currently offers the best power efficiency as an ISA without compromising performance. However it is still a closed system. (More closed to than x86, ironically)
While, StarFive, SiFive, and others are producing SoCs used by Banana, or Milk to produce mini (Raspberry Pi form factor) or desktop (mini-ITX) sized boards.
Sorry for jumping in like this, but my first thought was “this is great, but where is more diversity?”
Tenstorrent Ascalon — RVA23 – already supported in GCC and Clang
https://static.sched.com/hosted_files/riscvsummit2025/e2/Unleash%20your%20RISC-V%20Future%20with%20Tenstorrent%E2%80%99s%20High%20Performance%20Ascalon%20RISC-V%20Processor%20-%20Now%20Available!%20-%20Troy%20Jones%2C%20Tenstorrent.pdf
But yes, China is doing more…
https://www.theregister.com/2025/03/05/china_alibaba_risc_v_c930/
https://www.techpowerup.com/335026/chinas-rivai-technologies-introduces-lingyu-risc-v-server-processor
https://riscv.org/blog/three-high-performance-risc-v-processors-to-watch-in-h2-2025-ultrarisc-ur-dp1000-zhihe-a210-and-spacemit-k3/
LeFantome,
Thanks, I did not know about them. It is nice to have some domestic interest.
But apparently there are no current boards we can try, and Chinese manufacturers are leading the way to adopt them 🙂
https://www.reddit.com/r/RISCV/comments/1l4f90z/when_are_we_likely_to_actually_see_rva23/
sukru,
Diversity has never been the point, I believe. Money is. I believe that RISC-V is kept open just because there’s the only way newcomers (Chinese ones mostly) can make money and can leap forward uncumbered by patents and licensing, which is important if they ever aim to seriously compete outside the Chinese domestic market.
We’ve had diversity in the past. At home, I have 2x PA-RISC machines, 2x PowerPC machines, 3 ARM devices operating like a normal computer (running Linux, doing compuntery stuff rather than phone stuff), a MIPS machine, a few amd64 machines. And yet? Now the “serious” market is reduced to amd64 and ARM. It just stopped becoming economically feasible to keep maintaining multiple architectures when performance started being dependent much more on manufacturing process rather than how smart your ISA is.
I find it really hard to see super awesome competitive RISC-V devices unless there’s a strong market incentive for anyone to put RISC-V devices of serious performance and the big industrly players to then start porting money-making software for it.
AMD isn’t oblivious to RISC-V. ARM is the low hanging fruit and is popular NOW. AMD can take their time and develop RISC-V on the backburner.
I’m here for it. The less dependent the world is on Intel and Qualcomm in this space, the better off we all are. I’m honestly surprised AMD waited this long to enter the field.
Competing with Qualcomm, Samsung and now Apple ain’t easy feat. AMD had first to defeat first on the x86 ground before exploring new horizons and fight new quests. On the GPU side though, Nvidia is still strong opponent.
Qualcomm and MediaTek are probably the two real competitors.
I haven’t seen a Samsung Exynos design around in a while (US), which is surprising. I thought they would be a big Arm player by now in the SBC and desktop spaces.
Apple is Apple. They aren’t selling their procs to others, and AMD isn’t selling their procs to Apple.
AMD isn’t taking out Nvidia, but they do have valuable graphics tech in the IoT space. Nvidia seems to be leaving that space, or at least they don’t feel the need to update Tegra.
Will it boot mainline Linux? I’m getting tired of closed systems.
Much higher chance of having linux support coming from AMD, this isn’t because AMD is inherently more altruistic than Qualcomm or Apple, but because the graphics portion is RDNA 3.5 which already has Linux support. I can see AMD having some strong push for Linux support to be able to sell this in Steamdeck type devices to expand the market viability.
Indeed. AMD has a better corporate culture around upstreaming support for their chips.
Apple doesn’t have to think about anyone aside from Apple, and Qualcomm hates their customers.
That’s the least of your worries for having systems like this being usable.
NVidia the most.
Does it? AMD is nowadays dominant in the x86 market, which will be around for quite a while, even as WoA becomes more popular and mainstream (In particular, servers are still big – ARM servers only make sense in hyperscale right now. As is gaming.). In the ARM market, they’d be competing against several companies, like Nvidia, Qualcomm, Mediatek, etc. who have carved out various niches and market segments to themselves. The Radeon IP isn’t much a useful differentiator there, as Samsung is proving.
@DefineDecision
> the x86 market, which will be around for quite a while
Agreed
> ARM servers only make sense in hyperscale right now. As is gaming
I must misunderstand you. ARM is so compelling that all the cloud companies are making their own ARM silicon. AWS says that 90% of their EC2 (virtual machine) customers use Graviton (AWS ARM silicon).
For this in-house projects, I think that RISC-V makes even more sense and we should start to see that over the next 12-24 months.
But for AMD, who has to sell to customers, ARM makes the most sense right now.
LeFantome,
Yes, many ARM cores are not directly accessible to end users, but are ubiquitous everywhere. Not only cloud, every smart devices, switches, Internet infrastructure, and many other “below surface” places use ARM.
Still… yes, RISC-V would offer more benefits, especially when ARM ownership itself is still not fully decided. (Did SoftBank finally find a buyer)?
ARM is a publicly traded company now. The IPO was about 3 years ago.
https://www.google.com/finance/quote/ARM:NASDAQ
The 5-10W range is where AMD currently isn’t very strong. The Intel N150, N250, and, lesser, N355 kind of run this segment. This could be a pretty good fit if AMD is going to use the off the shelf Arm cores then add their graphics, network, and special sauce.
I’m also curious to know if MS is blaming Qualcomm for the failure of Windows Arm, which is probably fair.
It’s late 2025. This was probably something they were playing around with. Maybe for a game console, like the Xbox, but this sounds like an experiment. AMD has had an Arm license for a while, and they have been shipping Arm cores for their security processor.
It wold be neat to see, especially if AMD upstreams it’s Arm offering like its x86 offerings. People are kind of annoyed about how little support the current Arm vendors offer, and Arm needs better vendor support if it wants to make solid gains against x86. Support in a major distro like Debian or Fedora on day of launch would be a solid reason to pick these chips.
I remember back in the days, Intel also had entered the Arm game with the StrongARM bought from DEC in 1997, matured into the XScale (PXA, IXP, …) mobile processors running many PocketPC and early mobile phones, then sold to Marvell in 2006.
Has RISC-V catch up to Arm in performance or power usage? Until them cloud companies couldn’t care less.
@jgfenix
Well, for some classes of chip, the message seems to be that RISC-V has more performance from less power.
https://www.allaboutcircuits.com/news/upbeat-technologys-risc-v-mcu-takes-flight-with-near-threshold-computing/
But, at the higher end, the short answer to your question is no.
There are a bunch of RISC-V chips targeting servers that are “expected” to be about as fast as the fastest chips designed by ARM themselves. The Chinese claim to have a chip, Lingyu, which has already “reached the level of mainstream server chips from international brands like Intel and AMD”.
There are chips like the Spacemit K3 and UltraRISC UR-DP1000 that may be available this year that claim “desktop level performance”. That may mean better than a Raspberry Pi but nothing like Qualcomm or Apple Silicon.
Not there yet but not long now.
The front facing ISA is actually less relevant than the underlying microarchitecture. Most x86 processors are CISC to the programmer but RISC internally.
There is of course ISA specific optimisations that can be made to the underlying microarchitecture, but largely power savings are to be made by the implementer and are not inherent to the ISA. ARM is particularly good because it’s ISA is inherently frugal, but even x86 chips can be made to sip power whilst being quite fast. The aformentioned N100 series is a great example, but also historically the ISA agnostic Transmeta is another good example, as is the massive performance improvement and power savings gained by Intel moving to the Core2 microarchitecture.
For the love of god can it please support ACPI abd UEFI so that we can get a reasonable boot and chip support situation. Its pretty disgusting that in 2025 i cant just install a system onto Arm chips in a normal fashionor run a live session should i so wish
Shugo,
I agree with you on this, but I’ve pretty much given up all hope. 🙁 I wish RISCV would fix this, but the realist in me expects the same manufacturers to screw us over just like they did with ARM. The manufactures may benefit from RISCV’s openness and royalty-free licensing while continuing the tradition of restricting owner access to their own hardware.
@Alfman
You do not have to give up hope completely. There are good options out there:
https://milkv.io/titan
StarFive and SciFive even have UEFI firmware you can build yourself:
https://forum.rvspace.org/t/unlocking-new-possibilities-starfive-visionfive-2-sbc-now-supports-tianocore-edk-ii-uefi/2779/1
https://github.com/riscv-admin/riscv-uefi-edk2-docs
Many RISC=V SBC devices rely on device trees and lack a “BIOS” that provides things like keyboard drivers (even when they use OpenSBI). That said, many of them upstream the device trees to mainline Linux, you are not going to add random hardware to these units, and not most of the use cases for this kind of hardware are headless (no keyboards and mice). As long as future Linux kernels will work on these devices, I do not need to see ACPI and UEFI. For “desktop” or server use cases these are much more important of course.
For devices not needing full UEFI, there is OpenSBI:
https://github.com/riscv-software-src/opensbi
That’s more so a Linux problem and how they have decided to approach things.