Source: Content
Excerpted from
"
New Electronics
",
author Lin Zonghui
, thank you.
Ever since the ZTE incident stimulated the Chinese industrial sector, everyone from the government to the industry has been talking about "owning their own architecture", whether through acquisitions or self-development, hoping to get rid of the dilemma of relying on foreign suppliers for key technologies and materials.
Therefore, when RISC-V, an architecture with high flexibility, no licensing fees, and great commercial opportunities, emerged, the entire market also invested in research and development to promote the development of the related ecosystem.
RISC-V impacts the existing processor IP licensing ecosystem
As mentioned earlier, in terms of core processor architecture, Arm, MIPS and X86 are still the mainstream processor solutions in China, and before 2017, they accounted for nearly 100% of the market in total.
However, in 2018, the situation began to change. RISC-V officially entered the Chinese market. Although some Chinese chip manufacturers used the RISC-V instruction set to design products before that, it was not widely used because the market was not familiar with the architecture.
With the establishment of the China RISC-V Alliance and SiFive's cooperation with Andes to enter the Chinese market, SiFive also established SiFive China, an independent subsidiary in China, to lead the development of its business in China, which set off a widespread open source architecture trend.
SiFive or Andes sells pre-designed IPs, and using their designed architectures requires a licensing fee, but SiFive and other RISC-V IP suppliers only charge a one-time licensing fee, and future chip production does not require additional royalties based on sales.
If the manufacturer has sufficient technical capabilities, it can also directly use the RISC-V instruction set to develop its own architecture. The RISC-V instruction set also defines an extended instruction set. As long as it complies with regulations, manufacturers can define the content and form of the relevant instruction set to form a highly customized solution.
What’s important is that processor products designed using this method do not require any licensing fees or royalties. In contrast, if the same licensing method is used and Arm’s instruction set is used to customize one’s own processor, taking the highest-end solution as an example, the entry fee and licensing fee may be as high as tens of millions of dollars, not to mention the subsequent collection of royalties.
However, although the instruction set is free, it still requires the cooperation of other external costs to form a product, and most Chinese chip design companies do not have the ability to use the instruction set to design the processor architecture themselves, so they can only buy ready-made IP in the end. But no matter what, with the inherent advantage of lower licensing fees and no need for royalties, the concept of open source architecture has become instantly popular in China. More than 300 technology companies including Huawei HiSilicon and Alibaba have joined related alliances or invested in architecture development.
Arm is the first to bear the brunt of the response strategy, and the effect remains to be seen
Seeing the "free" business opportunity, existing processor IP companies have also become nervous. Arm has long dominated the processor IP supply market, and when free instruction sets dominate, it is naturally the most affected.
For customers, as chip sales grow, the royalties that follow them will become a profit killer. On the other hand, customers are actually unwilling to let Arm know their sales performance by collecting royalties. The
adoption of a royalty-free architecture can save costs and ensure that market performance is not known to Arm.
In order to respond to the challenges, Arm has also made certain adjustments and changes to its licensing strategy. For example, in the past, when customers chose a specific chip design, they had to pay a certain amount of licensing fees in advance, and the price could range from tens of thousands of dollars to millions of dollars. After the chip went into production, licensing fees and royalties would be charged based on the number of chips.
But for chip companies, the final design plan ready to be sent to the chip foundry may take half a year to a year to perfect. Customers pay large advance payments long before the chips actually begin to be delivered, which puts pressure on operating costs. Medium and large companies may have other business income support, but small companies may not be able to afford it at all.
To this end, Arm announced a new charging model, where chip manufacturers can obtain about three-quarters of the company's technology and chip design portfolio for a low one-time fee.
Subsequently, only when the chip is ready for production and starts shipping, will it be necessary to pay license fees and patent fees to Arm.
As the big brother in the IP licensing industry, Arm's support in ecology, development tools and manufacturing cannot be compared with free architecture. However, if it sits back and watches competitors eat into the market, its ecological advantage may eventually be reversed. The strategic shift for this purpose seems to be just a change in the charging order. Most IC design companies may not feel much, but it should play a certain role in small and medium-sized companies. However, the effect remains to be seen.
MIPS and IBM follow up with free license
Because the free instruction set has sparked widespread discussion, the industry believes that the commercial value it can create is no less than the traditional Arm licensing method. Even MIPS and Power Architecture, which used to follow the Arm licensing model, have joined the ranks of free instruction sets.
MIPS is an architecture with a very long history. In the past, it was very common in high-performance computing, network communication equipment and various embedded architecture products. In certain technical fields, such as single-core multi-threaded design, it is also better than Arm.
However, under the invasion of Arm architecture, the market has been lost, and coupled with the failure of business strategy, it has only been able to retreat to a few application fields, and is still being eroded by Arm architecture.
After encountering market setbacks, it was once acquired by Imagination, but Imagination did not develop a wider range of applications in the MIPS ecosystem. In addition, the cooperation with Apple will be terminated soon. Imagination, which may lose a large source of income, had to cut off its tail to survive. It sold MIPS to the US startup Wave Computing, and then sold itself to a Chinese company.
MIPS proposed a free instruction set licensing plan earlier this year, exempting it from collecting royalties, and also authorized the operating rights of the Chinese market to Xinlianxin. This approach is actually similar to SiFive's establishment of an independent company in China and Arm's establishment of Arm mini China in China.
In August, IBM also announced that it would open Power Architecture through the OpenPower Foundation, also using a licensing model similar to RISC-V.
In addition to free instruction set licensing, both companies also provide existing IP licensing models, but at the same time exempt from the collection of royalties, hoping to emulate the craze set off by RISC-V and thereby gain a larger market space.
The Power architecture has always held an important position in the field of high-performance computing, but there are too few market participants, and IBM is basically playing on its own. Although it can compete with Intel in many technical features, it is difficult for a good cook to cook without rice. So far, its market share in the relevant market is only about 1%.
In the past, the Power architecture has been promoted by the OpenPower Foundation, which provides a licensing method similar to the Arm architecture and has always charged lower fees than the Arm architecture. However, due to its unpopular application and lack of extensive software support, it has very few market users. Its efforts in the Chinese market in recent years have also been overshadowed by other open architectures with louder voices such as RISC-V. In order to avoid further marginalization, IBM has decided to follow in the footsteps of RISC-V and MIPS and open its instruction set.
The so-called openness this time is to follow the example of RISC-V and completely exempt the use of royalties at the instruction set level.
On the other hand, in order to emphasize the integrity of the Power architecture in the peripheral IP and to allow the instruction set license to be converted into commercial products more quickly, in addition to the open source Power architecture instruction set, IBM will also provide a variety of other technologies, including the softcore implementation of the Power architecture, the architecture-independent open coherent acceleration processor interface (OpenCAPI), and the reference design based on the open memory interface (OMI).
RISC-V has the fastest development, but MIPS/IBM still needs to be observed
At present, the RISC-V architecture has established an alliance in China, and there are many related IP suppliers entering the Chinese market. Taiwan's Andes, VeriSilicon, and the most popular SiFive have all actively laid out related markets and put forward many solutions.
Not long ago, Alibaba's Pingtou Ge released a high-performance IP based on RISC-V, Xuantie 910, which can reach a 16-core configuration and can operate at a clock speed of 2.5GHz. Its IP and customized extended instruction set will be completely open. This also means that China's RISC-V, in addition to existing IP suppliers, is also actively laying out completely self-owned IP, which is also a challenge for companies like SiFive.
Although Pingtou Ge's architecture is still obviously inferior to the high-performance architectures of IBM or Arm, it is already a rare high-performance architecture in RISC-V.
However, the architecture is still in the verification and testing stage and is unlikely to be commercialized in the short term, but this also represents the determination of Chinese manufacturers to deploy RISC-V for wider applications.
Judging from the overall situation, RISC-V is indeed very powerful at present, but its actual commercialization is still limited to small-scale, low-power products. Its overall market performance and related output value are still far behind Arm. The reason why MIPS and IBM launched the open instruction set licensing method is that it is intended to intercept.
However, considering that RISC-V has already gathered first-tier semiconductor manufacturers such as Google, Qualcomm, NVIDIA, Samsung, and WD, the potential R&D energy behind it is extremely huge. If it is a step slower, even high-performance computing may be replaced by RISC-V in the near future. This is also the biggest reason why IBM decided to jump into the open instruction set architecture processor licensing market after MIPS.
Arm is under great pressure and forced to adjust its business model
As for Arm, which is under attack from all sides, although it has actively launched a marketing war with its mature ecosystem, its existing licensing model and lack of instruction customization flexibility have caused it to gradually fall behind in the face of these open source architectures. Its main customers have gradually turned to RISC-V.
Although major computing cores, such as the high-end Cortex-A series, still dominate the mobile computing field, the low-end M series and even the R series for real-time computing are facing extremely severe challenges. Under the pressure of RISC-V, the boasts made by SoftBank when it acquired Arm may repeat the embarrassment of its original server market share declaration:
Arm claimed a few years ago that it would occupy 25% of the server market in 2021, but it has not exceeded 1% so far.
In order to maintain its advantage in the IoT market, Arm may propose more aggressive licensing plans in the future, not only to enter a larger market, but also to prevent free architecture from continuing to gain ground and affecting Arm's market layout.
In the Chinese market, Arm mini China mainly targets the IoT market. In addition to having its own architecture R&D team, its licensing conditions are also better than other Arm markets around the world. However, although the Chinese market is large, if its licensing preferences for the Chinese market affect customers in other countries, it may not be worth the loss for Arm. Therefore, perhaps Arm should consider changing the form of licensing and transferring the expected revenue source to other service areas.
In fact, Arm has announced limited support for customized instruction sets, which is similar to RISC-V.
Using Arm's strong design capabilities to fight against other IP competitors of RISC-V will certainly be a big challenge for the conservative Arm.
*Disclaimer: This article is originally written by the author. The content of the article is the author's personal opinion. Semiconductor Industry Observer reprints it only to convey a different point of view. It does not mean that Semiconductor Industry Observer agrees or supports this point of view. If you have any objections, please contact Semiconductor Industry Observer.
Today is the 2128th issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.
Semiconductor Industry Observation
"
The first vertical media in semiconductor industry
"
Real-time professional original depth
Scan the QR code
, reply to the keywords below, and read more
AI|
Wafer
|TSMC
|RF
|Huawei
|
IC
|
TWS headphones|Xiaomi
Reply
Submit your article
and read "How to become a member of "Semiconductor Industry Observer""
Reply
Search
and you can easily find other articles that interest you!