Where is MIPS heading?

With the RISC-V division out of the same door, a like the day, one no one to mention, MIPS over the years have experienced what? Once able to arm, X86 side to shoulder MIPS, how to sunset The West Mountain, repeatedly “sold”, and finally by the descendants of income under the umbrella?     But the increasingly marginalized MIPS has stalled again after the open source program?     The future of this RISC pioneer has become a focus for many practitioners.

(Comments) Where is the MIPS road?

See him rise flat, see him sunset West Mountain

As one of the industry’s most classic streamlined instruction set architectures, MIPS is one of the earliest commercial RISC architecture chips, and in the past there was a long time to match the Arm and X86 architectures in various fields.

In 1980, the streamlined instruction set RISC was born, and in 1984, John LeRoy Hennessy, a former president of Stanford University, founded MIPS with his team. Their business model is to license the chip design to other vendors so that they can easily produce high-performance CPUs.

MIPS also produced its own processors in the early days of its existence, and they were very forward-looking in their design, introducing the first processor design R2000 in its second year of existence, and the R3000 three years later, of which the R3000 was its first popular product on the market, selling more than a million units. The subsequent R3000A set a record for sales of more than 100 million sales; even in 1991, the 64bit design R4000 was introduced, not to mention that its rival Arm only widely promoted the 64bit processor design in 2012.

Looking back, MIPS was once brilliant in the 1990s, with semiconductor companies such as Pacemips, IDT and Toshiba using MIPS designs to make chips, and the chips they produce were used by end devices such as Sony and Nintendo’s consoles, Cisco routers and SGI supercomputers. Of the more than 16 billion microprocessors produced each year, especially in the home router market, 99% are RISC processors. In the past, it has been used as an efficient computing architecture on an overcomputing platform.

But missing out on the smartphone era, and Arm’s attack, led to a “lost decade” for MIPS (2007-2017).

Since MIPS products have been designed to target Intel’s X86 from the very beginning, the main focus is high performance, and Arm, on the other hand, has been targeting embedded low-power areas since its inception. The company, founded by the old snobbeh, was not sensitive to business, and when Arm began working with Qualcomm, Apple, MediaTek and others to build mobile processor chips for the smartphone market, MIPS remained immersed in niche products such as high-definition boxes and printers. In the final analysis, the difference between the academy and the business school is the case.

Finally, in the 21st century, the smartphone market burst, Arm suddenly took to the center of the stage, and MIPS because of the focus on the middle and high end of the power advantage, limited its ability to expand on the smartphone.

The slowness of MIPS has cost them their most critical decade.

Another factor that beats MIPS is the licensing method, where IP authorization is more expensive than instruction set authorization and allows for the addition of instructions, which allows the big wigs to design their own MIPS core, add instructions, publish development tools, and fragment it. Arm, on the other hand, has a command set authorization that is far more expensive than ANI authorization, controlling fragmentation. At that time, Arm also very eye-catching design of the world’s best and cheapest USB debugging tools, attracted a group of code farmers, thus building a huge Arm open source software library.

Of course, another factor contributing to its decline is the licensing model and costs. Because Arm has been focusing on the low-power embedded space, it has been very flexible in its licensing approach, mainly IP licensing, because the architecture authorization is much more expensive, very good control of fragmentation, and in terms of price advantage, and therefore attracted more vendors (TI, LSI, etc.) as well as students’ attention and participation, Arm’s ecosystem is well improved; and MIPS is primarily a structural authorization, and allows for the addition of instructions, code farmers have designed their own, resulting in serious fragmentation, coupled with the late MIPS management is not good, the structure of the change slowly, and failed to match the mainstream operating system and equipment ecological optimization trend, This loses the IP licensing has the introduction speed, software high compatibility characteristics.

So MIPS was gradually abandoned by the market, had to see Arm step by step to success. In the end, because arm system on the embedded field and mobile device market is not a segment of the encroachment, coupled with MIPS year-on-year poor management, the last once-beautiful enterprises helpless to go bankrupt.

MIPS has changed hands a few times, the fate of the road is difficult

MIP was acquired by Arm and Imagination at the end of 2012, and why did Imagination and Arm buy MIPS at Sunset West?

For Arm, MIPS patents are valuable, especially for 64-bit and multithreaded patents. Because Arm’s 64-bit architecture is 70-80% similar to MIPS 64, if you don’t participate in the acquisition at this time, you’re likely to be caught up in a long-running patent battle with MIPS patent owners, and Only $350 million to solve this potential concern, Arm is happy to do so. So Arm acquired nearly 500 of its patents.

The acquisition of MIPS, Imagination is to strengthen its CPU business, and fancy MIPS’s strong product set, as well as Android architecture support and licensing to China. As a result, Imagination acquired MIPS entity and 82 core patents related to the core architecture of MIPS processors. At the same time Intel as Imagination’s largest shareholder, the acquisition of MIPS from the side can also be able to restrain The Development of Arm.

But when Apple started with its own GPU, Imagination’s business was in a hurry, and in order to focus on the GPU while reducing the load, Imagination had to sell MIPS. However, Imagination sold itself and was first acquired by Canyon Bridge, and MIPS was acquired by Tallwood Venture Capital.

Wave Computin, the founder, took over MIPS from Tallwood, and Wave Computing’s founding team was many of them from MIPS, such as CEO Derek Meyer, who was a former vice president of MIPS.

Although most of MIPS patents were acquired by Arm, there are still more than 350 patents. However, MIPS has been acquired several times over the years, shifting between companies, affecting the strength of its support and the sustainability of its development. As a result, there is a view in the industry that “the recognition of the MIPS architecture has not decreased, but the impact has decreased significantly”.

One of the main reasons for Arm’s success in the early years was the combination of both CPUs and GPUs, which made them popular in the mobile age. Only the CPU’s MIPS and the original GPU-only Imagination, both exposed the integration advantages of the market today. The popularity of the wafer foundry industry, the rise of IP providers, the integration of vendors to solve most of the problems, to provide more simple solutions, which was important in the past and even in the future of the Internet of Things era. Wave Computing, the dynamic AI field, whose youthful and fresh DPU architecture, combined with MIPS, will likely bring a complete computing architecture to the market in addition to layout edge computing.

Yet the fire of RISC-V, another member of the RISC family, has once again ruthlessly cracked down on MIPS.

Arm then welcomed RISC-V, open source is also a flash in the pan

For decades, Intel as the representative of the CISC architecture, Arm architecture, MIPS architecture you fight me, good or not wonderful, who would have thought that RISC-V’s outburst, with its very flexible and open characteristics to disrupt the entire market, and the momentum of development, so that the major architectural groups have no qualms.

Just as the outside world is almost forgetting the former superstar of MIPS, at the end of 2018, bold new owner ComputWaveing announced open source for the latest R6 instruction set, aiming to accelerate the popularity of the MIPS instruction set architecture and help get the already marginalized MIPS instruction set architecture back on track.

“The MIPS-based solution developed under MIPS Open will complement our existing and future MIPS IP core, and Wave will continue to create and license MIPS IPcore sons worldwide,” Lee Flanagin, Wave’s senior vice president and chief commercial officer, said in a statement. As part of our portfolio, solutions and IP portfolio. This will ensure that current and new MIPS customers have a wide range of solutions from which they can choose from their SoC designs and also have access to a vibrant MIPS development community and ecosystem.

Art Swift, president of MIPS licensing, said 8.5 billion MIPS-based core-based chips had been shipped since 2000, and a wide range of customers were sticking to MIPS, including Microchip, Mobileye, MediaTek and Japan’s leading Denso Denso. The 32/64-bit R6 of MIPS is inherited from MIPS for more than 30 years of technical accumulation, can be described as a tree root deep.

Looking at the entire industry’s architecture situation, for domestic companies, the X86 architecture has no advantage, Arm also authorized to buy, and now open source CPU instruction set has RISC-V and MIPS two choices. But “open source is not equal to free”, in fact, the real significance of MIPS OPEN is to strengthen the promotion of the MIPS kernel and ecosystem, open ISA instruction set. The industry sees MIPS Open as a balance between development and revenue generation because it is a win-win business model for expanding users and ecosystems.

Rupert Baines, chief executive of UltraSoC, said: “Given the momentum of RISC-V, MIPS open source is an interesting and shrewd move. “MIPS already has a lot of quality tools and software environments. It’s an intelligent way to amplify MIPS’s own advantages, and it doesn’t lose much. ”

THE GREATEST ADVANTAGE OF MIPS IS THAT IT IS A CLASSIC ARCHITECTURE THAT HAS BEEN PROVEN OVER THE YEARS AND SHIPPED IN LARGE QUANTITIES, AND IS STILL RESPECTED BY MANY ELECTRONICENGINEERS, WITH A LARGE NUMBER OF MATERIALS AND REFERENCE BOOKS TO LEARN. In addition, MIPS provides patent protection and central authorization to avoid ISA fragmentation, both of which are lacking in RISC-V.

For MIPS itself, the future success of MIPS depends to a large extent on ecological construction and community development. But compared to now such as the day-to-day Arm and the rising star RISC-V, whether to respond to the needs of users in a timely manner, gather enough “popularity”, whether the Jedi survival also let the industry pinch a sweat for it.

However, there are now rumors that because of the impact of the progress of wave computing AI chip, MIPS also suffered a disaster, earlier also revealed that the company stopped MIPS open source-related projects, this “life-long” enterprise once again went to the fork.

What else does MIPS’s legacy have?

Of course, because of MIPS’s unshakable “historical status” or save a number of loyal users. To date, there are still some companies that are using the MIPS architecture to design processors and SoCs. Chinese people are more familiar with such as Dragon Core, Junzheng, Zhuhai Torch, and so on, as well as Shanghai Core Joint This year also obtained MIPS exclusive business rights in China. It can be said that the future development of MIPS is closely related to China.

Chinese Academy of Sciences computing from 2001 began to develop dragon core series processor, 2002 Dragon Core 1 stream success, 2003 Dragon Core 2B stream success, 2004 Dragon Core 2C stream success, 2006 Dragon Core 2E stream success, 2007 Dragon Core 2F stream chip success, Dragon 2F for Dragon Core first product chip, In 2009, Dragon Core 3A stream was successful. In order to industrialization of the research and development results of Dragon Core processor, in 2010 by the Chinese Academy of Sciences and the Beijing Municipal Government jointly led the funding, the formal establishment of Dragon Core Technology Co., Ltd. More than a decade later, Dragon Core has also evolved new processors such as Dragon Core 2H, 2J, 3A3000, 3B3000 and more.

Dragon core since its inception is to do an independent CPU, in the case of arm architecture does not allow changes in the design, X86 architecture is almost non-sale, so MIPS is the best choice. So Longcore bought the MIPS license, based on MIPS to do its own design, to create their own instruction set. Early dragon cores were basically MIPS architectures, with little change. The early model of Dragon Core 3 is based on MIPS64 R3, and later formed the LoongISA instruction set now used by Dragon Core through continuous expansion of the instruction set, so we can now say that Dragon Core 3 is no longer AN MIPS architecture, but Dragon Core 3 is still compatible with MIPS64 R3 instruction set. However, Dragon Core’s LoongISA only supports MIPS64 R2/R3, which is not compatible with MIPS 64 R5/R6.

Now the main frequency of dragon core 3A3000 has reached 1.5GHz, the performance has also made rapid progress, if the architecture alone said that the Dragon Core architecture has been basically very advanced. At the end of this month, Dragon Core will release 3A4000 and 3B4000 processor chips.

The same is true of Beijing Junzheng, Beijing Junzheng’s team from Ark Technology, and continue the development direction of Ark Technology, has been emphasizing the independent research and development of CPU core. Ark Science and Technology was founded with the help of Yan Guangnan, and the older generation of science and technology workers, led by Yan Guangnan, were inspired by the spirit of self-reliance in mao zedong’s time, and had a passion for independent research and development, so that China could have its own chip. So Jun zheng is also the use of MIPS architecture, the early Jun is in the mp3/mp4 era development is good, but to the tablet era, the software and ecosystem dependence increased, and Jun Zheng adopted the MIPS architecture at that time has no advantage, but also a few years of light.

In December 2018, a dedicated independent and controlled innovative IP supplier and IC design services company Shanghai Core was established, and in early May this year, Shanghai Core announced that it had acquired exclusive commercial management rights from Wave Computing in China and all of MIPS’s technology, including infrastructure, nearly 100 32-bit/64-bit CPU cores and related CPU authorization tools, compiler/verification kits, and new software optimization of the existing Fab core core core.

After receiving MIPS’s exclusive commercial management rights and technology, the company can either develop its own new CPU kernel, or the CPU core can be authorized, and Chinese customers can authorize the development of a complete and autonomous CPU kernel.

“As one of the founding engineering team members of Silicon Valley MIPS, in the 1990s, I took the lead from Silicon Valley to open up the ground for MIPS to the Chinese market, and now the core industry is once again in the robe, with full confidence in the performance and effectiveness of rISC architecture and processors.” MIPS’ Multi-thread, virtualization, low power consumption and small chip size continue to dominate modern and future CPU processors. We extend these benefits to ASIC services to help our customers become more competitive in order to take advantage of the market. China has been developed over 25 years and has nearly 50 authorized customers. Now China can develop autonomous and controllable CPUs based on an open RISC architecture and core core synth synths. “

Add a Comment

Your email address will not be published. Required fields are marked *