Due to the hot sales of PCs integrated with Apple's self-developed M series chips, a new round of Arm PC chip wave has been set off around the world.

Apple integrates self-developed M series chips

According to the latest news, AMD's next-generation Sound Wave APU is a new Arm-based chip that will be equipped with RDNA 3.5 GPU cores, MALL cache, and will compete with Qualcomm, Intel, and the upcoming NVIDIA in the Windows on Arm market in 2026.

In a new video released by the leaker "Moore's Law is Dead", they pointed out that AMD's next-generation Sound Wave APU will be equipped with 2 P-cores and 4 E-cores Arm-based CPUs. Although it will not provide powerful performance, it will have super long battery life like Windows on Arm laptops.

As a leading X86 processor supplier, this is not the first time that AMD has been rumored to enter the Arm PC chip market. But as more and more news is revealed, the chip giant may start another growth curve.

AMD's first Arm PC chip

It has been revealed that AMD's Sound Wave APU will be manufactured on TSMC's 3nm node, targeting a 5-10W form factor, with two performance cores, four efficiency cores (six total), and 4MB of L3 cache. Crucially, there will allegedly be four RDNA 3.5 CUs with "improved ML performance", informally named "RDNA 3.5+" to indicate the difference from the current RDNA 3.5 CUs available on the chip.

The chip's AI computing capabilities appear to be the main focus, as can be seen from the so-called "fourth-generation AI engine", which is supported by 16GB of 128-bit LPDDR5X-9600 RAM and is said to be suitable for "standard" systems. The whistleblower said it is not known whether the newer RDNA 3.5 CUs allow for FSR 4. However, gaming seems to be a secondary focus for this chip.

An interesting development with the Sound Wave APU is that it will allegedly feature 16MB of Last Level Memory Access (MALL) Cache (AMD folks call it Infinity Cache), which is unusual for a 5-10W APU. This is odd, as the "Strix Point" doesn't have this.

The leaker then points out how the AI Engine, CPU, and GPU will work on the same controller, so even on such a common chip, the extra bandwidth of the 16MB Infinity Cache shared between them could be beneficial.

The leaker further states: "This APU gets the MALL Cache, but for power reasons it makes sense to limit the number of CUs while providing the extra bandwidth to the AI Engine". He continues: "Given that this small APU has the MALL Cache, it would really surprise me if the Medusa Point didn't have it. I've heard that the Zen 6 Olympic Ridge has an 8CU RDNA 4 iGPU, so it would be interesting if the Sound Wave's iGPU still only uses the RDNA 3.5 variant”.

While it's still early days for Sound Wave, all the rumors we've heard about the chipset suggest it's primarily aimed at AI workloads without the (usually) high power consumption of competing hardware. The 16MB Infinity Cache doesn't seem too impressive at first, however, that's not even taking into account that the chip is rated for a TDP of just 5-10W.

In addition, the RDNA 3.5+ CUs could be a big driver for AI workloads, depending on how well they’re optimized for machine learning. With AMD claiming a 2026 release, it's unlikely we'll hear anything official from AMD anytime soon, but we now know it's likely to be an optimized Arm-based APU, rather than the Zen 6-leading chip we previously thought it would be.

As mentioned above, back when we first heard about Sound Wave last year, there were reports that it was a Zen 6-based APU. At the time, the project was revealed to be manufactured using TSMC's new 3nm process node. We expect AMD's true next-generation Sound Wave APUs to feature the new Zen 6 CPU architecture and the latest RDNA 4 graphics architecture.

But now, the situation seems to be changing, as multiple sources show that AMD is moving Sound Wave to the Arm-based CPU world.

This also means that AMD may really be ready to enter the Arm APU market. If this is true, it is another important decision made by the company after winning the X86 CPU market.

AMD is gaining momentum in the X86 market

AMD is set to have a strong 2024, with significant gains in the x86 CPU market in both the consumer and server sectors. According to the latest data provided by Mercury Research, AMD is chipping away at Intel's dominance, especially in the desktop PC processor space. While AMD's growth in the server market slowed slightly, its revenue share grew significantly, highlighting its success in high-margin segments.

In the fourth quarter of 2024, Intel maintained its lead in the consumer PC processor market with a 75.4% shipment share. However, AMD made significant progress, improving both its shipment and revenue share as more users opted for Ryzen processors.

AMD's consumer CPU revenue share will climb to 24.6% by the end of 2024, up 4.5% year-over-year and 0.7% quarter-over-quarter. The company's shipment share also grew 4%, indicating that AMD's CPUs are making steady progress. This revenue growth was driven by AMD's chips' growing share in the high-end market, as well as rising average selling prices (ASPs).

The desktop processor market was a bright spot for AMD. 

The company's shipment share surged to 27.1%, up 7.4% year-over-year, though down 1.6% from the previous quarter. Intel, on the other hand, had a mixed quarter, losing share due to Raptor Lake CPU stability issues and inventory adjustments. However, the company managed to recover some lost ground with the launch of its Arrow Lake-S desktop processors, with its shipment share increasing by 1.6% quarter-over-quarter.

Meanwhile, AMD's revenue share in the desktop space reached 27.3%, up 12.1% year-over-year, as its high-end Ryzen 9000 series CPUs became increasingly popular. Despite Intel's efforts with Arrow Lake-S, it still struggled to recover revenue share in the high-end desktop space.

In the mobile processor space, Intel continued to dominate, with a shipment share three times that of AMD. However, AMD made steady progress, with its mobile shipment share reaching 23.7% in the fourth quarter of 2024. This marks a 3.4% year-over-year increase and a 1.4% increase from the previous quarter. The company's revenue share also rose to 21.6%, up 6.5% year-over-year and 2.4% from the previous quarter.

While Intel's Arrow Lake and Lunar Lake laptop processors remain fiercely competitive, AMD's continued growth shows that the adoption of Ryzen-powered laptops is increasing. Improved performance, energy efficiency, and brand awareness have helped AMD attract more users, making this segment a key battleground for future expansion. Laptops sell three times more than desktops, and AMD's 3.4% year-over-year growth represents a solid step forward.

Even in servers, AMD continues its steady climb, reaching an all-time high in market share by the fourth quarter of 2024. The company's share of shipments rose to 25.1%, up 2% year-over-year and 0.9% from the previous quarter. While Intel's Xeon processors still power three-quarters of servers, maintaining that lead is becoming increasingly difficult, forcing Intel to offer deeper discounts to compete.

The really eye-catching data is that AMD's server revenue share soared to 35.5%. This is an increase of 3.7 percentage points year-on-year and 1.6 percentage points month-on-month, indicating that AMD not only sold more chips, but also occupied a larger share of the high-performance, high-profit server market. In fact, AMD's data center business performed better than Intel's data center and AI divisions in the fourth quarter, which was Intel's worst quarter for data center CPU sales in 13 years.

AMD's strong performance in 2024 highlights its growing influence in the consumer and server CPU markets. While Intel is still in the lead, AMD's revenue growth, improved average selling prices, and increasing popularity in the high-end market show that its strategy is paying off. With its continued momentum in the desktop, mobile, and server fields, AMD has proven itself to be a strong challenger in the ongoing CPU war.

But the rise of Arm PCs has forced AMD to defend itself.

Arm PC chips won't have a smooth journey

It is undeniable that, as mentioned at the beginning of the article, it was the success of Apple's M series that attracted everyone's interest in Arm PC chips. This also prompted Qualcomm to make a big push for its Snapdragon X Elite series of chips. Even GPU giant Nvidia is rumored to work with MTK to create a high-end chip Arm PC chip called N1X, and will launch a more mainstream N1 version in 2026. Both chips are said to be manufactured based on TSMC's N3 node and designed in cooperation with MediaTek. . Domestic startup Cixin is also a player in this track. In July last year, they released the first heterogeneous high-efficiency SoC Cixin P1 built using a 6nm process.

According to data released by TechInsights in October 2024, the x86/Arm market share of the global notebook computer market is currently about 82/18. By 2025, when the new generation of Windows-on-Arm "AI PCs" have been on the market for a full year, this number will not change significantly, and TechInsights expects the x86/Arm market share to be 80/20.

TechInsights predicts that by the end of 2029, Arm's share of the notebook market will double to more than 40% (with an x86/Arm ratio of about 60/40). Interestingly, Arm's notebook market revenue share will climb to 52% by then.

Arm CEO said in an interview in June 2024: "I think Arm's market share in Windows could exceed 50% in the next five years."

But judging from Qualcomm's struggles in the past few years, Arm PC is not an easy road. The earlier rumored Nvidia Arm PC chip overturned, which also confirmed this statement.

According to foreign media citing Geekbench citing the chip running score registered as N1X, the score of the Nvidia chip looks terrible, and it only got 1169 single-core scores in Geekbench 6. In contrast, Apple's current M4 chip (also based on Arm) scored 3831. The multi-core score is even more insignificant, only 2417 points, while the M4 is 15044 points. This is just the basic M4 chip. The MacBook equipped with the M4 Max chip can break 25,000 points, with more than 10 times the performance of the latter.

However, in my opinion, hardware problems are only a matter of time for them, and ecological problems are the biggest challenge. Especially Windows on arm, it is a big problem that cannot be avoided.

Over the past few decades, Windows + Intel have created a rich, reliable and familiar ecosystem for PCs. But progress on Windows on arm has been slow, which is why Qualcomm and its partners have released a number of Arm PC chips in the past few years but have made almost no progress. The reason why Apple can easily transform is due to its completely closed ecosystem and strong control.

Therefore, if you want Arm PC to be accepted by most consumers, you first need a version of Windows built specifically for Arm chips and optimized versions of applications.

Application simulation has been a big problem for Windows on Arm in the past decade, but Microsoft did provide x64 application simulation for Windows 11 about three years ago. This helps to ensure

However, as Microsoft said, although Windows on Arm can run native Arm applications and many unmodified x86 and x64 applications, for the best performance and battery life, applications should be built as Arm native as much as possible. Therefore, this requires Arm PC manufacturers and chip manufacturers to jointly encourage software manufacturers to build an ecosystem and switch.

Therefore, in the view of foreign media theregister, to make Windows on Arm feasible, chip manufacturers and OEMs need to do the following:

1. Appropriate investment in native Windows software on Arm - emulation is not enough

2. Push Microsoft to build Windows for Arm from scratch

3. Redefine Windows on Arm as a supporting device, not an x86 replacement

Many are hopeful that Windows on Arm will succeed, recognizing the potential benefits that the Arm architecture can bring to everyday devices, especially in areas where current x86 designs from Intel and AMD are limited, such as power efficiency. However, unless software compatibility issues are addressed, Windows on Arm will remain a merely "good enough" platform rather than a true competitor.

This isn't AMD's first foray into Arm chips

As early as February 2014, AMD launched a server CPU platform based on ARM 64-bit technology, which was the first time the two companies had carried out such cooperation. According to the introduction at that time, this Opteron A1100 series chip, code-named Seattle, was manufactured using 28-nanometer process technology. However, after the launch of this chip, AMD's product has not been released until 2026. It is understood that the reason for this is that AMD faces a big problem-some building blocks (such as fine-tuned ACPI and PCI Express support for ARM CPUs) were not initially adapted to the server field, and AMD needs to wait for the emergence of these modules before giving A1100 a chance. However, it is best that this product should not actually be launched, and the reason is probably related to the ecology behind it. But from another perspective, it took less than ten years for Arm server chips to become popular after AMD announced its first product. Standing on the shoulders of giants, I believe that the real rise of Arm PC is just around the corner. Will AMD be the winner of this track?

Source: Edited from Semiconductor Industry Observer

View more at EASELINK