Apple’s success with its self-developed M-series chips for PCs has ignited a new global wave of interest in Arm PC chips.
According to the latest reports, AMD’s next-generation Sound Wave APU is a new Arm-based chip equipped with RDNA 3.5 GPU cores, MALL cache, and is set to compete with Qualcomm, Intel, and the upcoming NVIDIA offerings in the Windows on Arm market by 2026.
In a new video by leaker “Moore’s Law is Dead,” it is revealed that AMD’s next-generation Sound Wave APU will feature a CPU with 2 performance cores and 4 efficiency cores based on Arm architecture. While it won’t offer extreme performance, it promises ultra-long battery life similar to Windows on Arm laptops.
As a leading x86 processor supplier, this isn’t the first time AMD has been rumored to enter the Arm PC chip market. With more details emerging, this chip giant might be on the brink of starting another growth trajectory.
AMD’s First Arm PC Chip #
It is reported that AMD’s Sound Wave APU will be manufactured on TSMC’s 3nm node, targeting a 5-10W power envelope. It will feature two performance cores, four efficiency cores (six cores in total), and 4MB of L3 cache. Crucially, it is said to have four RDNA 3.5 Compute Units (CUs) with “enhanced ML performance,” informally dubbed “RDNA 3.5+” to differentiate it from the current RDNA 3.5 CUs.
The chip’s AI computing power appears to be a major focus, evident from the so-called “fourth-generation AI engine,” supported by 16GB of 128-bit LPDDR5X-9600 RAM, making it suitable for “standard” systems. The leaker notes that it’s unclear whether the newer RDNA 3.5 CUs will enable FSR 4. However, gaming seems to be a secondary focus for this chip.
An intriguing development with the Sound Wave APU is the inclusion of 16MB of Memory Access Last Level (MALL) cache (known as Infinity Cache to AMD users), which is uncommon for a 5-10W APU. Notably, “Strix Point” lacks this feature.
The leaker further highlighted how the AI engine, CPU, and GPU will work on the same controller, making the additional bandwidth shared between them through the 16MB Infinity Cache particularly beneficial, even for a modest chip like this.
The leaker added: “This APU gets MALL cache, but for power reasons, it makes sense to limit the CU count while providing extra bandwidth to the AI engine.” He continued: “Considering this small APU has MALL cache, I would be surprised if Medusa Point lacks it. I heard Zen 6 Olympic Ridge has an 8CU RDNA 4 iGPU, so if Sound Wave’s iGPU still uses only the RDNA 3.5 variant, that would be quite interesting.”
While Sound Wave is still in its early stages, all the rumors suggest that this chipset is primarily aimed at AI workloads while avoiding the typically high power consumption of competitor hardware. The 16MB Infinity Cache might not sound remarkable at first, but that perspective changes when considering the chip’s rated TDP of just 5-10W.
Moreover, the RDNA 3.5+ CUs could be a significant driver for AI workloads, depending on their optimization for machine learning. As AMD claims a 2026 release, official details are unlikely to surface soon. Still, we now know it might be an optimized Arm-based APU rather than the Zen 6-leading chip previously speculated.
Back when Sound Wave first surfaced last year, rumors indicated it would be a Zen 6-based APU built on TSMC’s new 3nm process node. Expectations were for AMD’s true next-generation Sound Wave APU to adopt the new Zen 6 CPU architecture and the latest RDNA 4 graphics architecture.
However, things seem to be shifting now, with multiple reports suggesting AMD is moving Sound Wave into the Arm CPU world.
This could mean AMD is genuinely preparing to enter the Arm APU market. If true, it would mark another significant decision for the company after conquering the x86 CPU market.
AMD’s Growing Strength in the x86 Market #
AMD had a strong performance in 2024, making notable strides in the x86 CPU market across consumer and server segments. According to the latest data from Mercury Research, AMD is steadily chipping away at Intel’s dominance, particularly in desktop processors. Although AMD’s growth in the server market slowed slightly, its revenue share grew significantly, highlighting its success in high-margin sectors.
In Q4 2024, Intel maintained a 75.4% shipment share in the consumer PC processor market. However, with more users opting for Ryzen processors, AMD made considerable progress, increasing its shipment and revenue shares.
By the end of 2024, AMD’s consumer CPU revenue share climbed to 24.6%, a 4.5% YoY and 0.7% QoQ increase. Its shipment share also rose by 4%, signaling steady growth. AMD’s rising market share in the high-end segment and increased Average Selling Price (ASP) drove this revenue growth.
The desktop processor market was a bright spot for AMD. Its shipment share soared to 27.1%, a 7.4% YoY increase, despite a 1.6% QoQ decline. Intel had mixed results this quarter, losing ground due to Raptor Lake CPU stability issues and inventory adjustments. However, with the launch of Arrow Lake-S desktop processors, Intel recaptured some territory, posting a 1.6% QoQ increase in shipment share.
Simultaneously, AMD’s revenue share in the desktop segment reached 27.3%, a 12.1% YoY increase, driven by the growing popularity of its high-end Ryzen 9000 series CPUs.
In mobile processors, Intel continued to dominate, holding triple AMD’s shipment share. Nevertheless, AMD made steady gains, with its mobile shipment share reaching 23.7% in Q4 2024, marking a 3.4% YoY and 1.4% QoQ increase. Its revenue share rose to 21.6%, a 6.5% YoY and 2.4% QoQ increase.
Even in servers, AMD continued its steady climb, reaching its highest market share ever by Q4 2024. The company’s shipment share rose to 25.1%, up 2% YoY and 0.9% QoQ. AMD’s revenue share surged to 35.5%, reflecting a 3.7 percentage point YoY and 1.6 percentage point QoQ increase, signaling its growing dominance in high-performance, high-margin server markets.
While Intel remains the market leader, AMD’s revenue growth, rising ASPs, and increasing popularity in high-end segments indicate its strategy is working. With continued momentum across desktop, mobile, and server sectors, AMD has proven itself a formidable challenger in the ongoing CPU wars.
However, the rise of Arm PCs means AMD cannot afford to be complacent.
Arm PC Chips: A Rocky Road Ahead #
It is undeniable, as mentioned at the beginning of the article, that the success of Apple’s M-series has sparked interest in Arm PC chips. This has also driven Qualcomm to heavily invest in its Snapdragon X Elite series chips. Even GPU giant NVIDIA is rumored to be teaming up with MTK to develop a high-end Arm PC chip called the N1X, with a more mainstream N1 version set to launch in 2026. Both chips are reportedly being built on TSMC’s N3 node and co-designed with MediaTek. Additionally, a domestic startup, CiXin, is another player in this field. In July last year, they released their first heterogeneous high-efficiency SoC, the CiXin P1, built using a 6nm process.
According to data released by TechInsights in October 2024, the current global laptop market share between x86 and Arm is approximately 82/18. By 2025, after a full year of the new generation of Windows-on-Arm “AI PCs” being on the market, this ratio is not expected to change significantly, with TechInsights projecting an x86/Arm market share of 80/20.
TechInsights further predicts that by the end of 2029, Arm’s share of the laptop market will double, exceeding 40% (resulting in an x86/Arm ratio of around 60/40). Interestingly, by then, Arm’s revenue share in the laptop market is expected to climb to 52%.
In a June 2024 interview, Arm CEO stated, “I believe Arm’s market share in Windows could exceed 50% within the next five years.”
However, judging by Qualcomm’s struggles over the past few years, the road for Arm PCs is not a simple one. The rumored failure of NVIDIA’s Arm PC chip further reinforces this notion.
According to foreign media citing Geekbench scores for a chip registered as N1X, NVIDIA’s chip performed poorly, scoring only 1,169 in single-core tests on Geekbench 6. In comparison, Apple’s current M4 chip (also Arm-based) scored 3,831. The multi-core score was even less impressive at just 2,417, while the M4 scored 15,044. And that’s only the base M4 chip — MacBooks equipped with the M4 Max chip can exceed 25,000, delivering over 10 times the performance.
That said, from my perspective, hardware challenges are only a matter of time. The real hurdle lies in the ecosystem, with Windows on Arm being the biggest issue.
For decades, the Windows + Intel combo has built a rich, reliable, and familiar ecosystem for PCs. In contrast, progress on Windows on Arm has been slow, which explains why Qualcomm and its partners have launched multiple Arm PC chips over the past few years with little success. Apple, on the other hand, managed the transition seamlessly thanks to its tightly controlled, closed ecosystem.
Therefore, for Arm PCs to gain widespread consumer acceptance, a dedicated version of Windows optimized for Arm chips, along with native applications, is crucial.
Over the past decade, app emulation has been a major issue for Windows on Arm. Microsoft did introduce x64 app emulation for Windows 11 about three years ago, which helped somewhat.
Still, as Microsoft points out, while Windows on Arm can run native Arm apps and many unmodified x86 and x64 apps, for optimal performance and battery life, apps should be built natively for Arm. This requires collaboration between Arm PC manufacturers, chipmakers, and software developers to push for a more robust ecosystem.
As foreign media outlet The Register sees it, for Windows on Arm to become viable, chipmakers and OEMs need to focus on the following:
- Properly investing in native Windows software for Arm — emulation isn’t enough.
- Urging Microsoft to build Windows for Arm from the ground up.
- Redefining Windows on Arm as a complementary device, rather than a direct x86 replacement.
Many people hope Windows on Arm succeeds, recognizing the potential benefits Arm architecture could bring to everyday devices, especially in areas where Intel and AMD’s current x86 designs are constrained, such as power efficiency. However, unless software compatibility issues are resolved, Windows on Arm will remain a platform that is merely “good enough” — not a true competitor.
In Conclusion #
This isn’t AMD’s first foray into Arm chips.
As early as February 2014, AMD introduced a server CPU platform based on 64-bit ARM technology, marking the first collaboration of its kind between the two companies. At the time, the Opteron A1100 series chip, codenamed “Seattle,” was manufactured using a 28nm process.
However, after the launch of this chip, AMD’s products were slow to hit the market, only making a belated arrival in 2026. It’s understood that this delay was due to a significant challenge AMD faced — some building blocks, such as fine-tuning ARM CPUs with ACPI and PCI Express support, were not initially adapted for the server domain. AMD had to wait for these modules to emerge before the A1100 could have a proper chance.
In the end, this product may never have truly reached the market, likely due to ecosystem-related reasons. But looking at it from another perspective, it took less than a decade for Arm server chips to rise from AMD’s initial product announcement to widespread adoption. Standing on the shoulders of giants, the true rise of Arm PCs seems just around the corner.
Will AMD become the winner in this race? Only time will tell.