Monday, October 13th 2025
AMD "Sound Wave" Arm-Powered APU Appears in Shipping Manifests
Despite AMD's assertion that the Arm ISA doesn't provide any inherent efficiency advantage and that power savings are primarily dependent on the package and design, it seems AMD is developing an Arm-based SoC codenamed "Sound Wave." Recent shipping manifests, noted by X user @Olrak29_, indicate that the Sound Wave APU has resurfaced after a period of silence. It comes in a BGA 1074 package, featuring 1074 pins, and is intended for embedded systems without a removable socket. Measuring 32 × 27 mm, its compact size is suitable for mobile designs like handhelds and laptops. It utilizes the FF5 socket, replacing the FF3 used in Valve's Steam Deck SoC, and has a 0.8 mm pitch.
As an Arm design, it's likely to incorporate a big.LITTLE architecture, with some models featuring two larger P-Cores and four smaller E-Cores. This configuration would result in a six-core SoC paired with RDNA 3.5, with up to four CUs in the top model. This computing power is designed to operate within a 10 W TDP power envelope, meeting the demand for extended gaming sessions on battery power alone. There will likely be options to adjust the TDP for specific customers, allowing for higher or lower configurations based on their needs. The market this SoC will compete in is still unknown, but the competition with Arm-based solutions is heating up as Qualcomm and NVIDIA ready next-generation designs. Estimating the time of launch and pricing point for this APU is impossible, as we have little details on when and how AMD plans to implement this in third-party designs.
Sources:
Olrak29, via VideoCardz
As an Arm design, it's likely to incorporate a big.LITTLE architecture, with some models featuring two larger P-Cores and four smaller E-Cores. This configuration would result in a six-core SoC paired with RDNA 3.5, with up to four CUs in the top model. This computing power is designed to operate within a 10 W TDP power envelope, meeting the demand for extended gaming sessions on battery power alone. There will likely be options to adjust the TDP for specific customers, allowing for higher or lower configurations based on their needs. The market this SoC will compete in is still unknown, but the competition with Arm-based solutions is heating up as Qualcomm and NVIDIA ready next-generation designs. Estimating the time of launch and pricing point for this APU is impossible, as we have little details on when and how AMD plans to implement this in third-party designs.


32 Comments on AMD "Sound Wave" Arm-Powered APU Appears in Shipping Manifests
Valve helped accelerate Windows to Linux translation with Proton for the Steam Deck. This would provide the impetus to increase the efficiency of x86 to ARM translation as well, and increase the market for ARM-native software and games.
And no, they will not run faster just because it's ARM.
If it is, this is basically an ARM chip with Zen DNA.
Unless you have specific needs, x86 is fine.
This is very exciting news. AMD has the knowledge, experience and technology to make APU's, albeit x86. I'd like to see ARM version in smartphones one day. Could be unrealistic, considering Qualcomm, MediaTek and Exynos, BUT! It definitely makes things interesting.
They are certainly waiting for FSR4 APU, ML upscaling in a handheld is super useful, the switch 2 is really benefitting from DLSS, greatly.
AMD in my view is aiming to make a product that they can pitch to Nintendo for the NS3, and if they can demonstrate that they have a product in production that is hitting that 10w power budget, then in AMD's view they are increasing their credibility with Nintendo that they can deliver that product.
Compatibility is a software problem more then a hardware problem. Linux already has a competent ARM translation layer, and far better performance will follow widespread adoption. As it stands the market is so small there's no push for it.
But it isnt impossible. Apple showed that. The M series x86 code capability is surprisingly good.
Unity/Unreal can be easily compiled for ARM, and shaders can use FP16 instead of FP32 which offers 2x the performance in RDNA architecture.