China Dominates the World TOP500 Supercomputers

[HannibalBarca]

China is the largest developing country, no Chinese think it is already "the first/best in the world", and certainly no Chinese declare that China is a Superpower or "Czar" in any field. So your statement is aimless.

China does have an unique combination of all needed fundamentals that move a country forward, as we have witnessed in the past 30 some years. This unique combination works wonder for China, and it also puzzles the rest of world. The conventional western wisdom has failed to understand and explain this phenomenon, hence countless failed predictions. Yours may fall in this category.

Anyway, China is still a developing country, and it is actually DEVELOPING, that is all that matters to Chinese.

Pls restrain from your agressivity... Neither will I ,neither you will see what will happen next.. since we will be dead when that time comes... So in the End Whatever it's my "prediction" or Yours, none of us will know. But as far I can see, neither I attacked, nor I will upon those countries. I don't care, it was a discussion, so stay in "that" discussion.
Best regards

 

[TaiShang]

China is the largest developing country, no Chinese think it is already "the first/best in the world", and certainly no Chinese declare that China is a Superpower or "Czar" in any field. So your statement is aimless.

China does have an unique combination of all needed fundamentals that move a country forward, as we have witnessed in the past 30 some years. This unique combination works wonder for China, and it also puzzles the rest of world. The conventional western wisdom has failed to understand and explain this phenomenon, hence countless failed predictions. Yours may fall in this category.

Anyway, China is still a developing country, and it is actually DEVELOPING, that is all that matters to Chinese.

Well said. What the above poster repeats is pedestrian, at best. Rise and fall of the nations, alright, we are all familiar with that argument.

I think what these people often miss is the long historical perspective. Like you said, China, as well as East Asia, in general, did not conform to a number of Western assumptions for good or for bad, including development patterns and regionalization movement.

Future will definitely not repeat the past in the exact same fashion.

 

[rott]

Made an account just to write this post. While everyone are right to be proud of this supercomputer(me included) there's few problems that I hope it gets addressed in the future.
1. inadequate memory: Tianhe-2 have 88GB or RAM per node at just over two TFLOP/s while the sunway have only 32GB per node that runs at 3TFLOP/s.
2. Memory isn't particularly fast either at 136.5GB/s which could be a issues since the CPU already lacks on board cache.
3. There also a mention of inefficient interconnect, didn't dig too much into that so I won't comment.

The above problem resulted in a pretty abysmal HPCG benchmark of 0.371PFLOP/s compared to Tianhe-2's 0.580PFLOP/s.

Don't flame me for pointing out the negatives, always look forward to what my home country can come up with next

Thanks! Please continue posting.

 

[onebyone]

NSA, DOE say China's supercomputing advances put U.S. at risk

China's computing efforts are a threat to U.S. national security and may undermine profitable parts of the U.S. economy, a new report warns

• 
• By Patrick Thibodeau
  Senior Editor, Computerworld | MAR 15, 2017 3:00 AM PT

Advanced computing experts at the National Security Agency and the Department of Energy are warning that China is "extremely likely" to take leadership in supercomputing as early as 2020, unless the U.S. acts quickly to increase spending.

China's supercomputing advances are not only putting national security at risk, but also U.S. leadership in high-tech manufacturing. If China succeeds, it may "undermine profitable parts of the U.S. economy," according to a report titled U.S. Leadership in High Performance Computing by HPC technical experts at the NSA, the DOE, the National Science Foundation and other agencies.

"To maintain U.S. leadership in HPC," the report says, "a surge" of U.S. "investment and action is needed to address HPC priorities."

Concern about China's technical advances have been raised before by U.S. scientists and industry groups, but never in such striking terms -- or by representatives of a spy agency.

Indeed, the report says that "assuming status quo conditions, the meeting participants believe that a change in HPC leadership was extremely likely, with only minor disagreement on the timescale; many suggested that China would be leading the U.S. as early as 2020."

China supercomputing systems have been leading the Top 500 list, the global ranking of supercomputers, for several years. But that's not a measure of supercomputing leadership alone.

One workshop attendee, Paul Messina, a computer scientist and distinguished fellow at Argonne National Labs and the head of its Exascale Computing Project, sketched out the HPC leadership criteria: It means leadership in producing and using systems, as well as "first mover advantage." It also means staying in the lead at all times. The U.S. needs to control its HPC destiny and "can't depend on other countries to sell us what we need," he said in an email.

Something to keep in mind is that this report was written at a time when many assumed that supercomputing funding was not under threat. The report calls for more spending while the Trump administration, along with the Republican-controlled Congress, is planning major cuts in the federal budget.

"National security requires the best computing available, and loss of leadership in HPC will severely compromise our national security," the report says. "Loss of leadership in HPC could significantly reduce the U.S. nuclear deterrence and the sophistication of our future weapons systems."

Among those at the meeting was Barry Bolding, a senior vice president and chief strategy officer at supercomputer company Cray. "I will say from Cray's view, [the report] accurately reflects the discussion of the workshop and mostly accurately reflects some of our primary concerns regarding HPC competitiveness."

Steve Conway, an HPC analyst and research vice president at Hyperion Research, said the meeting "and report are important for alerting the U.S. HPC community, especially government officials, to the dangers of taking U.S. HPC leadership for granted when other nations, particularly China, are intent on seizing global leadership of the market for supercomputers."

The report makes three overarching observations about China's Sunway TaihuLight system, which at 93 petaflops, is ranked first on the Top500 list of supercomputers.

The TaihuLight supercomputer is "homegrown," and includes processors that were designed and fabricated in China. The Chinese chip design "includes architectural innovations," and was designed using "a true co-design approach" where the applications are tuned to take advantage of the chip design, the report said.

The machine "is not a stunt," the report notes, meaning China didn't develop this system for bragging rights. The machine "is being used for cutting edge research," and three of the six finalists for the Gordon Bell Prize, the top research award in HPC, were the result of Chinese efforts.

The report offers something particularly insightful about China's motivations.

"Meeting participants, especially those from industry, noted that it can be easy for Americans to draw the wrong conclusions about what HPC investments by China mean – without considering China's motivations," the report states.

"These participants stressed that their personal interactions with Chinese researchers and at supercomputing centers showed a mindset where computing is first and foremost a strategic capability for improving the country; for pulling a billion people out of poverty; for supporting companies that are looking to build better products, or bridges, or rail networks; for transitioning away from a role as a low-cost manufacturer for the world; for enabling the economy to move from 'Made in China' to 'Made by China,' " the report states.

But it also pointed out that the computer codes developed for industry, "are good proxies for the tools needed to design many different weapons systems."

To express your thoughts on Computerworld content, visit Computerworld's Facebook page, LinkedIn page and Twitter stream.

http://computerworld.com/article/31...-china-nearing-supercomputing-leadership.html

 

[Pyr0test]

it shows that sanctions against china will only make us stronger

 

[kankan326]

“Something to keep in mind is that this report was written at a time when many assumed that supercomputing fundingwas not under threat. The report calls for more spending while the Trump administration, along with the Republican-controlled Congress, is planning major cuts in the federal budget.”----Real purpose of this article?

 

[samsara]

Thanks for the article, but I think this kind of info is better aggregated into one single dedicated thread with topic about THE CHINESE SUPERCOMPUTER and related news/info.

@MOD, please look into this matter and rectify it.

Note: a quick search for "Sunway TaihuLight" returns several postings under separate entries just within the 2017... these scattering posts related to China's Supercomputer make info retrieval gets tougher. And to put this Supercomputer info under the way much broader topic of "China Science & Technology Forum" will also cause it difficult to locate.

Therefore a dedicated thread for China's Supercomputer under 'China & Far East' section is what I suggest. Hopefully all concerned will support this idea

 

[bobsm]

TOP500 List Refreshed, US Edged Out of Third Place

TOP500 Team | June 19, 2017 00:22 CEST

FRANKFURT, Germany; BERKELEY, Calif.; and KNOXVILLE, Tenn.— The 49th edition of the TOP500 list was released today in conjunction with the opening session of the ISC High Performance conference, which is taking place this week in Frankfurt, Germany. The list ranks the world’s most powerful supercomputers based on the Linpack benchmark and is released twice per year.

In the latest rankings, the Sunway TaihuLight, a system developed by China’s National Research Center of Parallel Computer Engineering & Technology (NRCPC) and installed at the National Supercomputing Center in Wuxi, maintains its top position. With a Linpack performance of 93 petaflops, TaihuLight is far and away the most powerful number-cruncher on the planet.

Tianhe-2, (Milky Way-2), a system developed by China’s National University of Defense Technology (NUDT) and deployed at the National Supercomputer Center in Guangzho, China, occupies the number two position with a Linpack mark of 33.9 petaflops. Tianhe-2 was the number one system in the TOP500 list for three consecutive years, until TaihuLight eclipsed it in June 2016.

The new number three supercomputer is the upgraded Piz Daint, a Cray XC50 system installed at the Swiss National Supercomputing Centre (CSCS). The upgrade was accomplished with additional NVIDIA Tesla P100 GPUs, doubling the Linpack performance of the system’s previous mark of 9.8 petaflops in November 2016, which itself was the result of a significant upgrade. Piz Daint’s current Linpack result of 19.6 petaflops enabled the system to climb five positions in the rankings.

As a result of the Piz Daint upgrade, Titan, a Cray XK7 system installed at the Department of Energy’s (DOE) Oak Ridge National Laboratory, drops to number four in the rankings. Its Linpack mark of 17.6 petaflops has remained constant since it was installed in 2012.

Rounding out the top 10 are:

Sequoia (17.2 petaflops), an IBM BlueGene/Q system installed at the DOE’s Lawrence Livermore National Laboratory, at number five;
Cori (14.0 petaflops), a Cray XC40 system housed at the National Energy Research Scientific Computing Center (NERSC), at number six;
Oakforest-PACS (13.6 petaflops), a Fujitsu PRIMERGY system running at Japan’s Joint Center for Advanced High Performance Computing, at number seven;
Fujitsu’s K computer (10.5 petaflops), installed at the RIKEN Advanced Institute for Computational Science (AICS), at number eight;
Mira (8,6 petaflops), an IBM BlueGene/Q system installed at DOE’s Argonne National Laboratory, at number nine; and
Trinity (8.1 petaflops), a Cray XC40 system running at Los Alamos National Laboratory, at number ten.
With the two Chinese supercomputers and one Swiss system occupying the top of the rankings, this is the second time in the 24-year history of the TOP500 list that the United States has failed to secure any of the top three positions. The only other time this occurred was in November 1996, when three Japanese systems captured the top three spots.

Nevertheless, the US still claims five of the top ten supercomputers, which is more than any other nation. It also leads the pack in the total number of TOP500 systems, with 169. China is a close second with 160. However, both countries lost share compared to six months ago, when they each claimed 171 systems. Besides the US and China, the most well-represented countries on the list are Japan, with 33 supercomputers, Germany, with 28, France, with 17, and the UK, with 17.

Aggregate performance on the TOP500 rose to 749 petaflops, a 32 percent jump from a year ago. Such an increase, though, is well below the list’s historical growth rate of approximately 185 percent per year. The slower growth in list performance is a trend that began in 2013, and has shown no signs of reversal.

Intel continues to be the dominant supplier of TOP500 chips. Either Xeon or Xeon Phi processors power 464 of the 500 systems. IBM Power processors are in 21 systems, while AMD Opteron CPUs are present in six systems. Processor share, with regard to supplier, has not changed appreciably over the last year.

A total of 91 systems on the list are now using accelerator/coprocessor technology, up slightly from 86 in November 2016. The most popular choices are NVIDIA GPUs, which are present in 74 systems, and Xeon Phi coprocessors, which are employed in 17 systems. (Three of those use a combination of NVIDIA GPUs and Xeon Phi coprocessors). Outliers include a single ATI Radeon-accelerated system and two that use PEZY technology. An additional 13 systems now use the Xeon Phi (Knights Landing) as the main processing unit.

For system interconnects, Ethernet and InfiniBand continue to be the most prevalent technologies. Ethernet is present in 207 systems; InfiniBand is present in 178. However, for the top 100 systems, their relative share changes dramatically, with Ethernet installed in just a single system, while InfiniBand is used in 42 of these elite machines. Intel Omni-Path interconnect technology, which made its first appearance a year ago with 8 supercomputers, is now installed in 38 systems.

In the system vendor arena, Hewlett Packard Enterprise (HPE) claims the most TOP500 systems, with 144. These include 25 systems originally installed by SGI, which HPE purchased in 2016. Lenovo is the second most popular vendor, with 88 systems, and Cray is in third place, with 57.

Cray systems, however, continue to lead in overall performance, claiming 21.4 percent of the list’s total performance. HPE is well back in second place, with an overall performance share of 16.7 percent. Thanks to its number one Sunway TaihuLight system, NRCPC retains the third spot with 12.5 percent of the total performance.

Energy efficiency on the list continues to rise, as reflected in the latest Green500 results. The top four positions are all occupied by newly installed systems in Japan, with the upgraded Piz Daint supercomputer capturing the number five spot. All of these use NVIDIA’s latest P100 GPUs. In fact, the top 13 systems on the latest Green500 are all equipped with the P100 hardware.

The most energy-efficient system on the Green500 list is the new TSUBAME 3.0, a modified HPE ICE XA system installed at the Tokyo Institute of Technology. It achieved 14.110 gigaflops/watt during its 1.998-petaflop Linpack performance run. It is ranked number 61 on the TOP500 list.

The number two Green500 entry is the kukai system at the Yahoo Japan Corporation. Built by Exascaler, this system achieves 14.045 gigaflops/watt, just 0.3 percent behind TSUBAME 3.0. It occupies position 466 on the current TOP500 list.

The number three system on the Green500 list is the AIST AI Cloud system at the National Institute of Advanced Industrial Science and Technology, Japan. This system from NEC achieves 12.68 gigaflops/watt and is ranked number 148 on the TOP500.

The number four Green500 system is the RAIDEN GPU system, manufactured by Fujitsu. It is installed at RIKEN’s Center for Advanced Intelligence Project. It attained 10.6 gigaflops/watt and is in position 306 on the TOP500 list.

Piz Daint, the fifth-ranked supercomputer on the Green500, conducted a power-optimized run of the Linpack benchmark, achieving 10.4 gigaflops/watt. At number three on the TOP500, it represents the most energy-efficient supercomputer in the top 50 of that list.

At 10:30 am CET on Monday, June 19, the TOP500 and Green500 awards will be handed out to the winning recipients. The award ceremony will be followed by a presentation from TOP500 author Erich Strohmaier, who will provide additional highlights of the list.

https://www.top500.org/news/top500-list-refreshed-us-edged-out-of-third-place/

 

[TaiShang]

Planned supercomputer would be 10 times faster than today's No 1

(China Daily) 08:24, May 10, 2018

Technology experts check the condition of the Tianhe-1 supercomputer at the National SuperComputer Center in Tianjin. [Photo/Xinhua]

The National Supercomputing Center in Shenzhen plans to build a next-generation supercomputer that will be 10 times faster than the world's current speed champion, a senior executive said.

"The investment is likely to hit 3 billion yuan ($470.6 million), and key technologies for the supercomputer are expected to be developed independently," Wang Zhenglu, director of the project management department of the center told China Daily.

The supercomputer will be "exascale", meaning that it will be capable of making a quintillion (1 followed by 18 zeros) calculations per second, much faster than the current No 1 supercomputer, Sunway TaihuLight, Wang said. The new supercomputer is expected to be developed by 2020.

The exascale supercomputer will play an important role in promoting the development of the Guangdong-Hong Kong-Macao Greater Bay Area, especially in cloud computing, big data and artificial intelligence, he said.

Supercomputers are a new driving force in China's computer innovation and are important for the nation's information security. China has established six supercomputing centers, in Tianjin, Guangzhou, Wuxi, Changsha, Jinan and Shenzhen.

China is now at the vanguard of supercomputer development, having raced ahead of the United States and Japan. The latest biannual ranking of the world's 500 fastest supercomputers showed China's Sunway TaihuLight and Tianhe-2 are the world's fastest and second-fastest machines, with Titan from the US in fifth place.

The Shenzhen center, founded in 2009, has focused on high performance computing and cloud computing. It maintains leading positions in a variety of areas like weather forecasting and disaster prevention.

The center's work in detailed exploration and large-scale data processing has helped a domestic petroleum company reduce petroleum exploration costs.

The growing shortage of cloud computing resources led the center to recently purchase 420 more rack-mounted servers, 10 graphics processing servers and a group of storage systems to expand its computing resources.

China has an edge in hardware, but more efforts are needed to develop tailor-made software for supercomputers to widen their application, Wang Gang, an associate professor at the College of Computer Science and Technology at Jilin University, said in an earlier interview.

Zhang Zhihao contributed to this story.

@qwerrty , @xunzi , @cirr

 

[AndrewJin]

 

[qwerrty]

undisputed champs for 5 years in a row now
you tried man. you tried.. with that dirty sanction trick. it didn't work. better luck next time, amelica

.

 

[xunzi]

China like to surprise the world. When we said 2020, it could potentially be this year or next year 2019.

 

[JSCh]

Tsinghua University defends ASC18 title, First-time finalist ShanghaiTech University wins AI award
May 10, 2018
SPONSORED CONTENT BY INSPUR

On May 9, the final round of the 2018 ASC Student Supercomputer Challenge (ASC18) ended at Nanchang University. Tsinghua University was once again crowned Champion, while first-time finalist ShanghaiTech University won second place and received the e Prize Award.

2018 ASC Student Supercomputer Challenge (ASC18)​

Jointly hosted by the Asia Supercomputer Community, Inspur Group and Nanchang University, ASC18 enrolled over 300 participating teams from around the world, with the top 20 entering the final round. In addition to independently designing and building supercomputers up to 3000 Watts, the teams were required to run HPL/HPCG benchmark, tackle problems in AI machine reading and comprehension, along with optimizing cutting-edge scientific and engineering applications including RELION – a core application of the cryo-EM that won the 2017 Nobel in Chemistry, and NASA’s famous computational fluid-dynamics code CFL3D.

Among the finalists, the defending champion Tsinghua University completed the highest number of tasks, claiming a strong lead in the final round and retaining its title won the previous year. In the HPGG benchmark test, SIESTA, RELION and CFL3D challenges, Tsinghua University gave an impressive performance, demonstrating a profound comprehension of supercomputing systems and applications as well as excellent capabilities in optimizing performance.

Tsinghua University, ASC18 Champion​

ShanghaiTech University entered the ASC finals for the first time, finishing second and winning the e Prize Award in a surprise upset. In the highlight task in AI reading and comprehension, the team made innovations and improvements in model algorithm and performance training and completed the parallel model training of large-scale dataset within 8 hours. Moreover, the team’s prediction accuracy reached 46.46, close to the world cutting-edge level, winning them the e Prize Award.

The NTHU team designed a heterogeneous supercomputing system with Inspur servers and Tesla GPU accelerators with optimized power dispassion control. Their HPL performance reached 42.99 TFLOPS, improving the ASC17 record to a large extent and winning the Highest Linpack.

“It’s exciting to participate in a competition like ASC, and It’s heartening to see these students who have worked so hard and traveled so far to perform such challenging tasks and solve problems with hardware and software,” said Jack Dongarra, ASC Advisory Committee Chair, professor at the Oak Ridge National Laboratory of the United States and the University of Tennessee. “I hope the contest can be a truly satisfying experience for all participants, whether now or in the near future.”

Wang Endong, the initiator of the ASC challenge, academician of the Chinese Academy of Engineering, noted that currently, supercomputing and big data, cloud computing, and artificial intelligence are becoming increasingly integrated and interconnected. Supercomputers will become an important infrastructure for a smart society. The development speed and capability of supercomputers will be closely related to social development, improvement of living standards and the progress of civilization. As an international supercomputer competition, ASC will play an even more important role in cultivating integrated supercomputer talents and international academic exchanges, laying a solid foundation for the development of smart computing.


Tsinghua University defends ASC18 title, First-time finalist ShanghaiTech University wins AI award | HPCwire

 

[cirr]

Planned supercomputer would be 10 times faster than today's No 1

(China Daily) 08:24, May 10, 2018

Technology experts check the condition of the Tianhe-1 supercomputer at the National SuperComputer Center in Tianjin. [Photo/Xinhua]

The National Supercomputing Center in Shenzhen plans to build a next-generation supercomputer that will be 10 times faster than the world's current speed champion, a senior executive said.

"The investment is likely to hit 3 billion yuan ($470.6 million), and key technologies for the supercomputer are expected to be developed independently," Wang Zhenglu, director of the project management department of the center told China Daily.

The supercomputer will be "exascale", meaning that it will be capable of making a quintillion (1 followed by 18 zeros) calculations per second, much faster than the current No 1 supercomputer, Sunway TaihuLight, Wang said. The new supercomputer is expected to be developed by 2020.

The exascale supercomputer will play an important role in promoting the development of the Guangdong-Hong Kong-Macao Greater Bay Area, especially in cloud computing, big data and artificial intelligence, he said.

Supercomputers are a new driving force in China's computer innovation and are important for the nation's information security. China has established six supercomputing centers, in Tianjin, Guangzhou, Wuxi, Changsha, Jinan and Shenzhen.

China is now at the vanguard of supercomputer development, having raced ahead of the United States and Japan. The latest biannual ranking of the world's 500 fastest supercomputers showed China's Sunway TaihuLight and Tianhe-2 are the world's fastest and second-fastest machines, with Titan from the US in fifth place.

The Shenzhen center, founded in 2009, has focused on high performance computing and cloud computing. It maintains leading positions in a variety of areas like weather forecasting and disaster prevention.

The center's work in detailed exploration and large-scale data processing has helped a domestic petroleum company reduce petroleum exploration costs.

The growing shortage of cloud computing resources led the center to recently purchase 420 more rack-mounted servers, 10 graphics processing servers and a group of storage systems to expand its computing resources.

China has an edge in hardware, but more efforts are needed to develop tailor-made software for supercomputers to widen their application, Wang Gang, an associate professor at the College of Computer Science and Technology at Jilin University, said in an earlier interview.

Zhang Zhihao contributed to this story.

@qwerrty , @xunzi , @cirr

The 2nd known "exascale" machine in development.

 

[cirr]

从中科院到思朗科技：中国超高性能芯片的诞生之路原创

2018-05-07 17:52 投资界　 李梅

"摩尔定律不可能永远适用，芯片性能升级遇到了世界性瓶颈，这刚好是我们追赶的最佳机遇。”王东琳说。
　　
5G到来的进程正在加速。

目前，5G正处于标准确定的关键阶段，今年6月，国际标准组织3GPP即将完成5G第一版本国际标准。同时政策利好也不间断，4月24日，发改委、财政部发布通知，将降低5G公众移动通信系统频率占用费标准……

5G技术不仅能支持包括汽车在内的各类机器人顺畅地互联互通，也将是智能手机、智能家居、人工智能、大数据及云计算等多个领域实现“质”的升级的基础技术。

面对这股迎面而来的5G浪潮，中国的芯片行业准备好了吗？虽然道阻且长，但以华为海思为代表的中国智造依然让人期待。目前在基带领域，华为海思是唯一可与高通相比的中国公司，这是华为30多年来各种要素积累的结果，也非一蹴而就。

而在创业公司层面，也有一家公司显得尤为特别：脱胎于中科院自动化研究所、原国家专用集成电路设计工程技术研究中心（1992年组建）的思朗科技，由该中心原主任、原中科院自动化所所长王东琳博士带队，已经研制出高性能领域微处理器MaPU, MaPU首次实现了代数算法级全局优化且高度可编程，已于2015年流片成功。

MaPU不仅完全可以实现国际巨头的可编程处理器的性能，而且功耗比可媲美ASIC。基于MaPU，思朗科技进一步研制出了三大领域处理器：面向5G通信领域的UCP、面向多媒体领域的UMP，以及面向超算领域的HPP。同时，还配备了AI领域处理器：深度神经网络引擎NNE。

近日，在中科院自动化所，投资界采访到了思朗科技创始人兼首席科学家王东琳。王东琳在国际上最早提出代数算法级“全局优化计算”架构，基于此架构设计的MaPU在计算能力和性能功耗比方面具有国际领先水平。

思朗科技创始人兼首席科学家王东琳

高性能领域微处理器MaPU

王东琳介绍，MaPU最大的特点是计算能力强，功耗低。

当前，市面常用的处理器有几类，一类是可编程的处理器，比如英特尔、TI的处理器，因为可编程，适应性非常广。但是当执行一个数学算法的时候，它的运算器的利用率一般在15%左右，高的能到20%。TI的处理器，运算器利用率最高也仅可以到40-50%。就是说这些处理器空有那么高的频率，空有这么多资源，但它的执行效率不高。

还有一种是不用编程的ASIC方案，它用硬件把算法流程和对算法的控制都已经写好了。这种实际上就是算法的加速器，因此执行效率会非常高，几乎可以接近100%。

很明显，可编程处理器和ASIC之间在功耗上有着巨大的差别。ASIC的问题是，效率高，但是算法不可变，算法只要变化一点，这个芯片就不能用了。

而MaPU既可以做到接近ASIC的效率（计算资源利用率可以达到90%以上），同时也高度可编程，兼具两者的优势。

以超算芯片为例，MaPU的性能功耗比全球第一

在王东琳看来，当前主流可编程处理器的核心问题在于它是传统体系结构，指令层次低并试图在运行时刻通过乱序多发射等技术尽可能实现局部并行执行。这造成了芯片中计算资源利用率不高，数据IO量大，动态功耗大，整体性能功耗比不高，已经不适应当今社会对微处理器巨大计算能力以及极低功耗的双重渴求。如果能从应用算法整体来考量时间及空间等不同维度并行特性，并利用这些并行特性来进行整体性优化整理，其内核中的运算器的使用率将获得大幅提高。于是王东琳和他的团队经过精密的测算和试验，提出代数算法级全局优化的解决方案。

“一条指令，就可以实现一个代数算法，所以叫做代数指令。传统架构的指令集都是算术运算级的指令。”王东琳介绍说，MaPU将其升级为代数级算法指令，“MaPU通过代数指令软流水线来零延时动态重构（与算法相适应的）硬件架构，达到与ASIC基本相同的算法架构，实现整个算法的全局优化执行过程。”

简而言之，MaPU既能支持应用算法级全局优化，又能通过高度可重构的计算架构与存储体系在软件层面实现的这点，可灵活适应领域（5G通信、多媒体、超算或人工智能）内各种算法，可以说MaPU集合了ASIC、FPGA、CPU的优势，是几乎可以与ASIC的性能功耗比相媲美的“软ASIC。”

“MaPU-代数运算微处理器，在并行代数运算、并行存储体系指令系统和硬件架构方面产生重大原始性创新，将微处理器硬件支撑从标量/超标量运算提升至代数运算层次，数量级地提升计算密集型领域微处理器能效比。”王东琳如此总结。

那么，具体的性能以及功耗比指标如何，王东琳给出了一组直观的对比数据：

以极光H1.0超算芯片为例，芯片内部集成32个HPP处理核，双精度浮点处理能力将达到4,659GFLOPS@64，经评估功耗仅为40W左右，性能功耗比达到 116GFLOPs@64/W，为全球第一。

因此，MaPU应用量产之后，它独创的架构优势将有望使我国在微处理器架构上实现重大突破，在同等能耗比下释放巨大计算能力，引领中国电子行业的自主创新发展。

对此，中科院院长白春礼在今年3月曾在媒体表示：“在高科技产品研发方面，中科院即将发布一款具有完全自主知识产权的微处理器——MaPU代数处理器，达到国际领先水平。相信MaPU系列处理器问世以后，将在世界计算机、通信等领域以及广大消费电子市场产品中大放异彩。”

“MaPU的三个孩子”

在MaPU的基础上，思朗科技进一步研制出了三款性能强大的领域处理器：5G通信领域处理器UCP、多媒体领域处理器UMP，以及超算领域处理器HPP。

UCP：全球首次彻底实现软件定义无线电。

UCP是MaPU针对移动通信增强的通用通信处理器是5G宏基站基站处理器的芯片内核，UCP内核每秒可以完成5.8G定点复数FFT，每秒可以完成55GBPS的LDPC编码以及2.5GBPS译码。根据初步测算，一颗含有二十个UCP内核的基带处理器就可以满足64天线5G宏基站全部基带处理需求。

“如果使用FPGS搭建5G系统，则需要多片互联才能实现系统解决方案，而基于FPGA的电路系统主频一般<400-600MHz，片间互联总线带宽受限，运算能力受限，这将成为实现5G系统的瓶颈。”王东琳说。

目前UCP内核是国际领先的在可接受代价下实现全软件定义5G无线传输基带处理系统的处理器内核，王东琳将UCP定义为“移动通信领域彻底实现基带处理的软件定义无线电。”

除了基站设备制造商以外，UCP内核还可以提供给5G终端制造商。由于5G标准的原因，所有终端都要重新嵌入适应5G算法的基带内核（原有的基带内核或者DSP内核无法应对5G下行接收和上行发送时的大运算量），这也是UCP内核的机会。多合一无线通信设备、各领域宽带自组网终端也都是UCP通过全软件定义无线电技术施展能力的空间。

UMP：依靠它智能手机、智能电视都可在线升级视听体验，并可为高清摄影、摄像等应用提供超级引擎。

UMP是MaPU的第二个“孩子”，面向智能手机、智能电视的多媒体微处理器内核。

UMP在MaPU基本架构基础上发明了更高效的并行处理架构，使得各类视频处理运算的性能功耗比与ASIC相媲美，某些方面甚至胜出，同时还保持了高度可编程特性。

“这种特点可以使家用电视随着电视厂家视音频处理算法演进在线改善视音频体验效果，这是ASIC电视芯片无法比拟的。”王东琳表示，“同时也可以通过算法和软件改进快速推出新产品。”

4个UMP内核+1个ARM内核构成的超高清电视引擎芯片（功耗不超过8 瓦）可以满足4K超高清电视的所有视音频处理以及基于安卓系统电视管理的全部需求，视听效果可以媲美索尼与三星最高端电视。14个UMP内核+1个ARM内核构成的超级电视引擎芯片可以满足8K超高清电视所有处理与计算需求。

加上思朗研发的AI领域深度神经网络引擎NNE内核后，不论是智能电视还是智能手机，都可以在高清视频、机器视觉、人机交互方面实现性能大跃升，让国产消费电子产品在较低成本下大幅提高用户使用体验。第一款超级电视引擎芯片将于2018年底完成流片，并可以开始面向智能电视制造商进行应用推广。

HPP内核超算处理器相较于Intel最新型处理器性能功耗比提高近一个数量级。

目前，MaPU的第三个“孩子”HPP内核已经是一项比较成熟的产品。

“MaPU通过架构创新获得的核心能力还是高密集度计算。”因此，王东琳一直希望通过MaPU针对通用计算领域增强型内核HPP（高性能处理）构成超算型微处理器，满足高端服务器领域的需求。

从而开发了极光H1.0——超级计算类MaPU微处理器（预计年底完成流片及封测），经设计评估极光H1.0的性能功耗比远超国际上其它超算类微处理器，可以作为超算系统及超级服务器的核心处理器。

王东琳给出了一组数据：

基于HPP内核的超算处理器极光H1.0与Intel最新型Xeon Phi性能相当，但可16/32/64/128位灵活重构，性能功耗比提高近一个数量级：在提供同等64位浮点运算能力时极光H1.0（2x16个HPP内核）功耗是40W，Intel Xeon Phi处理器是300W。

NNE：顶级深度神经网处理

思朗的另一款产品：加速神经网络引擎NNE 继承并发展了MaPU的“同心圆存储体系优化模型”，并针对深度神经网络进行了优化，优点是针对主流神经网络，整网吞吐率高，访存需求及功耗低，与英伟达深度神经网内核性能相当但效率更高。思朗的NNE在存储体系和配置深度神经网络时吸收MaPU全局优化思想，性能功耗比上优势明显。

“NNE可以支持深度学习训练，特别支持智能推理，在视频图像识别方面做了专门优化，在目标检测、识别、视频图像结构化方面具有较突出的优势。”。王东琳介绍。

一个可期待的应用场景就是智能驾驶，在这个场景里，UMP可以高速处理多路摄像头图像并提取待识别目标；NNE负责理解判断处理路况、车况并提供决策及驾驶控制所需要的关键信息；而UCP负责提供极短时延的车联网通信能力。

甘坐冷板凳，坚持研发了近10年的科研团队

思朗及其前身团队，从2009年就开始部署新指令集体系结构的研究，到研发出完全自主创新的微处理器架构MaPU已经经历了9年艰苦岁月。研发团队系原国家专用集成电路设计工程技术研究中心的70多位核心科研人员。

2017年二季度开始，团队开始公司化运作。

王东琳是个有技术信仰的人，团队伙伴也是这样。集成电路领域向来人才不足，人员短缺。因为做芯片等硬件太苦，收益又不高，不少优秀学生毕业后更愿意选择金融和互联网业。做芯片就需要有工匠精神，需要顶尖的技术人才愿意沉下心做研发，并能长期承受住研发成功与否的压力。思朗科技的研发团队做到了。

基于MaPU的几款增强领域处理器在各自领域各有优势，实际上也可以组合使用，从而落地到各种不同的实际场景中：5G通信、智能手机、智能家居、超算、智能驾驶、智慧城市、机器人和无人机等等。

“摩尔定律不可能永远适用，芯片性能升级遇到了世界性瓶颈，这刚好是我们追赶的最佳机遇。”王东琳说，“芯片是一个需要耐心的行业，我们已经做了9年，基础已经打好，接下来的重点是到实际应用中去调整、去优化。“

http://news.pedaily.cn/201805/430955.shtml?from=singlemessage

http://www.smartlogictech.com/

@Bussard Ramjet