What's new

China Dominates the World TOP500 Supercomputers

20191121-SC19-SCC-Tsinghua-675x380.jpg
Tsinghua University Racks Up Its Ninth Student Cluster Championship Win at SC19
By Oliver Peckham
November 27, 2019

Tsinghua University has done it again. At SC19 last week, the eight-time gold medal-winner team took home the top prize in the 2019 Student Cluster Competition (SCC), bringing their total wins to nine gold medals, three silver, and three bronze.

“The SCC,” SC19 says, “is an opportunity for students to showcase their expertise in a friendly, yet spirited, competition.” In essence, the competition is a trial by fire: teams of students assemble real computing clusters on the conference’s exhibit floor and race to complete actual workloads across a number of applications.


...

https://www.hpcwire.com/2019/11/27/...nth-student-cluster-championship-win-at-sc19/
 
22:53, 14-Jan-2020
ASC20: Largest student supercomputer challenge launched in China
By Gong Zhe


Supercomputer sounds like a serious thing that ordinary people shouldn't have access to. But college students with a passion for high-level computing can still get their hands on these monster machines if they can prove their ability.

The Asian Supercomputer Community (ASC) Student Supercomputer Challenge is the world's largest supercomputer contest. It was founded in 2012, and this year's event has just started in Beijing.

A total of 356 teams from colleges all over the world have registered. And they are busy preparing for the phase-1 contest.

Phase-1 is basically a coding match. The teams submit their code according to the requirements of ASC. The best 20 teams will enter phase-2, which is a week-long, face-to-face match.

This year, the phase-2 will be held in Shenzhen's SUSTech on April 25.

During phase-2, the teams will build their own supercomputers and use them to solve scientific problems related to math, AI, language, and so on.

"ASC is unique among the three largest supercomputer contests. Students are well-hosted, and they can experience the Chinese culture," said Wang Endong, founder of the contest and a member of the Chinese Academy of Science.

The ASC20 is hard. New challenges include a new field in science and technology. In one challenge, students have to simulate a 30-bit quantum computer using a supercomputer – a scaled-down version of Google's "quantum supremacy" test.

Another challenge requires students to invent an AI program to automatically solve the English language test, which is much harder than "just" understanding humans' natural language.

Most of the teams are from Chinese universities, with less than 30 teams from non-Chinese schools.

But foreign participation has opened the eyes of the Chinese students.

"The foreign teams have a great attitude of enjoying the event, and they are better at communication, co-op, and expression," explained Liu Jun, head of AI and HPC of Inspur, one of the largest cloud computing and big data service provider in China.

"Chinese students should learn from them," he added.

(Part of the video footage is taken during ASC19 back in 2019.)
 
Japan Captures TOP500 Crown with Arm-Powered Supercomputer | TOP500
June 22, 2020
top5_june2020.png__250x250_q85_subsampling-2.png

FRANKFURT, Germany; BERKELEY, Calif.; and KNOXVILLE, Tenn.—The 55th edition of the TOP500 saw some significant additions to the list, spearheaded by a new number one system from Japan. The latest rankings also reflect a steady growth in aggregate performance and power efficiency.

The new top system, Fugaku, turned in a High Performance Linpack (HPL) result of 415.5 petaflops, besting the now second-place Summit system by a factor of 2.8x. Fugaku, is powered by Fujitsu’s 48-core A64FX SoC, becoming the first number one system on the list to be powered by ARM processors. In single or further reduced precision, which are often used in machine learning and AI applications, Fugaku’s peak performance is over 1,000 petaflops (1 exaflops). The new system is installed at RIKEN Center for Computational Science (R-CCS) in Kobe, Japan.

Number two on the list is Summit, an IBM-built supercomputer that delivers 148.8 petaflops on HPL. The system has 4,356 nodes, each equipped with two 22-core Power9 CPUs, and six NVIDIA Tesla V100 GPUs. The nodes are connected with a Mellanox dual-rail EDR InfiniBand network. Summit is running at Oak Ridge National Laboratory (ORNL) in Tennessee and remains the fastest supercomputer in the US.

At number three is Sierra, a system at the Lawrence Livermore National Laboratory (LLNL) in California achieving 94.6 petaflops on HPL. Its architecture is very similar to Summit, equipped with two Power9 CPUs and four NVIDIA Tesla V100 GPUs in each of its 4,320 nodes. Sierra employs the same Mellanox EDR InfiniBand as the system interconnect.

Sunway TaihuLight, a system developed by China’s National Research Center of Parallel Computer Engineering & Technology (NRCPC) drops to number four on the list. The system is powered entirely by Sunway 260-core SW26010 processors. Its HPL mark of 93 petaflops has remained unchanged since it was installed at the National Supercomputing Center in Wuxi, China in June 2016.

At number five is Tianhe-2A (Milky Way-2A), a system developed by China’s National University of Defense Technology (NUDT). Its HPL performance of 61.4 petaflops is the result of a hybrid architecture employing Intel Xeon CPUs and custom-built Matrix-2000 coprocessors. It is deployed at the National Supercomputer Center in Guangzhou, China.

A new system on the list, HPC5, captured the number six spot, turning in an HPL performance of 35.5 petaflops. HPC5 is a PowerEdge system built by Dell and installed by the Italian energy firm Eni S.p.A, making it the fastest supercomputer in Europe. It is powered by Intel Xeon Gold processors and NVIDIA Tesla V100 GPUs and uses Mellanox HDR InfiniBand as the system network.

Another new system, Selene, is in the number seven spot with an HPL mark of 27.58 petaflops. It is a DGX SuperPOD, powered by NVIDIA’s new “Ampere” A100 GPUs and AMD’s EPYC “Rome” CPUs. Selene is installed at NVIDIA in the US. It too uses Mellanox HDR InfiniBand as the system network.

Frontera, a Dell C6420 system installed at the Texas Advanced Computing Center (TACC) in the US is ranked eighth on the list. Its 23.5 HPL petaflops is achieved with 448,448 Intel Xeon cores.

The second Italian system in the top 10 is Marconi-100, which is installed at the CINECA research center. It is powered by IBM Power9 processors and NVIDIA V100 GPUs, employing dual-rail Mellanox EDR InfiniBand as the system network. Marconi-100’s 21.6 petaflops earned it the number nine spot on the list.

Rounding out the top 10 is Piz Daint at 19.6 petaflops, a Cray XC50 system installed at the Swiss National Supercomputing Centre (CSCS) in Lugano, Switzerland. It is equipped with Intel Xeon processors and NVIDIA P100 GPUs.

General highlights
Aggregate list performance is now 2.23 exaflops, up from 1.65 exaflops six months ago. The majority of that increase is the result of the new number one Fugaku supercomputer. The new entry point on the list (system number 500) is 1.24 petaflops, only a slight increase from the previous list. Overall the number of new systems in the list is only 51, a record low since the beginning of the TOP500 in 1993.

China continues to dominate the TOP500 with regard to system count, claiming 226 supercomputers on the list. The US is number two with 114 systems; Japan is third with 30; France has 18; and Germany claims 16. Despite coming in second on system count, the US continues to edge out China in aggregate list performance with 644 petaflops to China’s 565 petaflops. Japan, with its significantly smaller system count, delivers 530 petaflops.

Technology trends
A total of 144 systems on the list are using accelerators or coprocessors, which is nearly the same as the 145 reported six months ago. As has been the case in the past, the majority of the systems equipped with accelerator/coprocessors (135) are using NVIDIA GPUs.

The x86 continues to be the dominant processor architecture, being present in 481 of the 500 systems. Intel claims 469 of these, with AMD installed in 11 and Hygon in the remaining one. Arm processors are present in just four TOP500 systems, three of which employ the new Fujitsu A64FX processor, with the remaining one powered by Marvell’s ThunderX2 processor.

The breakdown of system interconnect share is largely unchanged from six months ago. Ethernet is used in 263 systems, InfiniBand is used in 150, and the remainder employ custom or proprietary networks. Despite Ethernet’s dominance in sheer numbers, those systems account for 471 petaflops, while InfiniBand-based systems provide 803 petaflops. Due to their use in some of the list’s most powerful supercomputers, systems with custom and proprietary interconnects together represent 790 petaflops.

Vendor highlights
Chinese manufacturers dominate the list in the number of installations with Lenovo (180), Sugon (68) and Inspur (64) accounting for 312 of the 500 systems. HPE claims 37 systems, while Cray/HPE has 35 systems. Fujitsu is represented by just 13 systems, but thanks to its number one Fugaku supercomputer, the company leads the list in aggregate performance with 478 petaflops. Lenovo, with 180 systems, comes in second in performance with 355 petaflops.

Green500 results
The most energy-efficient system on the Green500 is the MN-3, based on a new server from Preferred Networks. It achieved a record 21.1 gigaflops/watt during its 1.62 petaflops performance run. The system derives its superior power efficiency from the MN-Core chip, an accelerator optimized for matrix arithmetic. It is ranked number 395 in the TOP500 list.

In second position is the new NVIDIA Selene supercomputer, a DGX A100 SuperPOD powered by the new A100 GPUs. It occupies position seven on the TOP500.

In third position is the NA-1 system, a PEZY Computing/Exascaler system installed at NA Simulation in Japan. It achieved 18.4 gigaflops/watt and is at position 470 on the TOP500.

The number nine system on the Green500 is the top-performing Fugaku supercomputer, which delivered 14.67 gigaflops per watt. It is just behind Summit in power efficiency, which achieved 14.72 gigaflops/watt.

HPCG Results
The TOP500 list has incorporated the High-Performance Conjugate Gradient (HPCG) Benchmark results, which provided an alternative metric for assessing supercomputer performance and is meant to complement the HPL measurement.

The number one TOP500 supercomputer, Fugaku, is also now the leader on the HPCG benchmark with a record 13.4 HPCG-petaflops. The two US Department of Energy systems, Summit at ORNL and Sierra at LLNL, are now second and third, respectively, on the HPCG benchmark. Summit achieved 2.93 HPCG-petaflops and Sierra 1.80 HPCG-petaflops. All the remaining systems achieved less than one HPCG-petaflops.
 
China Speeds Up Advanced Chip Development
Efforts underway to develop 7nm, DRAM, 3D NAND, and EUV domestically as trade war escalates.
June 22nd, 2020 - By: Mark LaPedus
pop_rating_lev4.png



China is accelerating its efforts to advance its domestic semiconductor industry, amid ongoing trade tensions with the West, in hopes of becoming more self-sufficient.

The country is still behind in IC technology and is nowhere close to being self-reliant, but it is making noticeable progress. Until recently, China’s domestic chipmakers were stuck with mature foundry processes with no presence in memory. Recently, though, a China-based foundry entered the 14nm finFET market, with 7nm in R&D. China also is expanding into memory. And in the fab equipment sector, China is developing its own extreme ultraviolet (EUV) lithography system, which is a technology that patterns the most advanced features in chips.

It’s unlikely that China will develop its own EUV system in the near term. And for that matter, the nation’s foundry and memory efforts are modest, at least for now. And China won’t overtake multinational chipmakers anytime soon.

Nonetheless, it is developing its domestic IC industry for several reasons. For one thing, China imports most of its chips from foreign suppliers, creating an enormous trade gap. China has a sizeable IC industry, but it isn’t large enough to close the gap. In response, the nation is pouring billions of dollars into its IC sector with plans to manufacture more of its own chips. Simply put, it wants to become less dependent on foreign suppliers.

China recently accelerated those efforts, especially when the U.S. launched a multi-prong trade war with the nation. In just one example, the U.S. has made it more difficult for Huawei to obtain U.S. chips and software. And recently, the U.S. blocked ASML from shipping an EUV scanner to SMIC, China’s largest foundry vendor. China sees these and other actions as a way to hamper its growth, prompting it to speed up the development of its own technologies.

Meanwhile, the U.S. says its trade-related actions are justified, claiming that China is engaged in unfair trade practices and has failed to protect U.S. intellectual-property. China dismisses those claims. Nonetheless, the industry needs to keep an eye on the trade issues as well as China’s progress in semiconductors. They include:

  • SMIC is shipping 14nm finFETs, with a 7nm-like process in R&D.
  • Yangtze Memory Technologies (YMTC) recently entered the 3D NAND market with a 64-layer device. A 128-layer technology is in R&D.
  • ChangXin Memory Technology (CXMT) is shipping its first product, a 19nm DRAM line.
  • China is expanding into compound semis, including gallium nitride (GaN) and silicon carbide (SiC).
  • China’s OSATs are developing more advanced packages.
This all sounds impressive, but China is still trailing. “China is spending like crazy. China’s strategy is to be a player in semiconductor manufacturing. It comes from wanting to have a bigger share of its domestic manufacturing capabilities, as well as for security considerations,” said Risto Puhakka, president of VLSI Research. “But China’s share in memory is small. On the logic side, they are behind TSMC. China is far from being self-sufficient from any reasonable aspect.”

Those aren’t the only issues. “There are still many challenges for China, including the need for more talent and IP in semiconductor manufacturing, and the need to further narrow the gap in the leading process technologies,” said Leo Pang, chief product officer at D2S. “The top challenge is the tension between the U.S. and Chinese governments, which is causing uncertainty in the supply of manufacturing equipment and EDA software.”

China’s strategy
China has been involved in the IC industry for decades. In the 1980s, it had several state-run chipmakers with outdated technology. So at the time, China introduced several initiatives to modernize its IC industry. With help from foreign concerns, the country launched several chip ventures in the 1980s and 1990s.

Still, China found itself behind the West in semiconductor technology for several reasons. At the time, the West implemented strict export controls on China. Equipment vendors were prohibited from shipping the most advanced tools to China.

Then in 2000, China launched two new and modern domestic foundry vendors — Grace and SMIC. By then the export controls were relaxed in China. Equipment vendors simply required a license to ship tools to China.

Around that time, China became a large manufacturing base with low labor rates. Demand for chips skyrocketed. Over time, the nation became the world’s largest market for chips.

Starting in the late 2000s, multinational chipmakers began building fabs in China to gain access to the market. Intel, Samsung and SK Hynix built memory fabs in China. TSMC and UMC built foundry fabs there.

By 2014, China consumed $77 billion worth of chips, according to IC Insights, but it imported most of them. Plus, China only manufactured 15.1% of those chips, according to IC Insights. The rest were manufactured outside of China.

In response, and armed with billions of dollars in funding, the Chinese government unveiled a new plan in 2014. The goal was to accelerate China’s efforts in 14nm finFETs, memory and packaging.

Then, in 2015, China launched another initiative, dubbed “Made in China 2025.” The goal is to increase the domestic content of components in 10 areas — IT, robotics, aerospace, shipping, railways, electric vehicles, power equipment, materials, medicine and machinery. In addition, China hopes to become more self-sufficient in ICs and wants to increase its domestic production to 70% by 2025, according to IC Insights.

In 2019, China consumed $125 billion worth of chips, according to IC Insights, but it still imports most of them. China only manufactured 15.7% of those chips, so it’s unlikely the country will reach its production targets by 2025.


Fig. 1: China’s IC market vs. production trends Source: IC Insights

China faces other challenges, as well, particularly a shortage of technical talent. “China is still seeking more talent in semiconductor manufacturing,” D2S’ Pang observed. “That is mainly because China is building a dozen new fabs. It has already recruited thousands, if not tens of thousands, of experienced semiconductor engineers from fabs in Taiwan, Korea, Japan and even the U.S. by paying them with very attractive compensation packages.”

On the bright side, China made a quick recovery from the Covid-19 pandemic earlier this year. In the first half of 2020, chip and equipment demand were strong in China and elsewhere. “200mm capacity has continued to be running full with a wide range of end applications. In the 300mm area, this has been a similar situation over this past year,” said Walter Ng, vice president of business development at UMC.

Others see similar trends. “China semiconductor test and packaging markets have been resilient throughout the Covid-19 period,” said Amy Leong, senior vice president at FormFactor. “The demand remains solid, fueled by the combination of the momentum built over the last few years from the ‘Made in China 2025’ initiative, and the recent ‘panic build/buy’ amid China-U.S. tensions. With this said, we are seeing an increasing level of demand uncertainties in China as the fear of a global economic recession mounts.”

The mood is also tense. Starting in 2018, the U.S. launched a trade war with China, slapping tariffs on Chinese-made goods. China has retaliated.

The trade war is escalating. Last year, the U.S. added Huawei and its internal chip unit, HiSilicon, to the “entity list,” saying the companies pose as a security risk. To do business with Huawei, a U.S. company must obtain a license from the U.S. government. Many U.S. vendors have been denied, which impacts their bottom lines.

Then, earlier this year, the U.S. expanded the definition of a “military end user” in China. This is designed to prevent China’s military from obtaining any U.S. technology.

In May, the U.S. moved to stem the flow of chips to Huawei from overseas fabs. “Going forward, an overseas fab must halt sales to Huawei if it meets the following three conditions: A) fab uses U.S. equipment or software to make chips; B) the chip is designed by Huawei; and C) the chipmaker has knowledge the item produced is destined for Huawei,” said Paul Gallant, an analyst with Cowen. “(This requires) foreign chipmakers using U.S. equipment to get a license before selling chips to Huawei. But the language of the new rule may not actually ban such sales. On the upside, the new rule only covers chips actually designed by HiSilicon, not all chips made by overseas fabs being sold to Huawei.”

At some point, TSMC may halt new orders to Huawei. It’s unclear how this will all play out. The rules are fuzzy and could change overnight.

Foundry, EUV efforts
Even before the trade war, China was in the midst of a major fab expansion program. In 2017 and 2018, China had 18 fabs under construction, according to SEMI’s “World Fab Forecast Report.” Eventually, these fabs were built.

China currently has 3 fabs under construction, according to SEMI. “Two of those fabs are for foundry. One is 8-inch and another is 12-inch. There is another one for memory (12-inch). Still on the drawing board are 7 more,” said Christian Dieseldorff, an analyst at SEMI.

The foundry industry makes up a large percentage of China’s fab capacity. China’s foundry industry is split into two categories—domestic and multinational vendors.

TSMC and UMC are among the multinationals. TSMC operates a 200mm fab in Shanghai. In 2018, TSMC began shipping 16nm finFETs in another fab in Nanjing.

UMC is manufacturing chips in a 200mm fab in Suzhou. UMC also has a new 300mm foundry venture in Xiamen, which is shipping 40nm and 28nm.

Meanwhile, China’s domestic foundry vendors, such as ASMC, CS Micro and the Huahong Group, all focus on mature processes. On the leading edge, startup HSMC is developing 14nm and 7nm in R&D.

SMIC, China’s most advanced foundry company, is the world’s fifth largest foundry vendor, behind TSMC, Samsung, GlobalFoundries and UMC, according to TrendForce.

Up until last year, SMIC’s most advanced process was a 28nm planar technology. In comparison, TSMC introduced 28nm a decade ago. Today, TSMC is ramping up 5nm with 3nm in R&D.

This is a sore spot for the Chinese government. Because China is behind, Chinese OEMs must obtain their most advanced chips from foreign suppliers.

On the other hand, there isn’t a gap for mature processes in China. “The technology node gap is not an issue for most fabs, since the majority of chips used in IoT and automotive applications do not require leading-edge nodes,” D2S’ Pang said.

Nonetheless, SMIC is trying to develop advanced processes. In 2015, SMIC, Huawei, Imec and Qualcomm formed a joint R&D chip technology venture in China with plans to develop a 14nm finFET process.

This is a big step. “Moving to finFETs at 14nm is not easy. Everybody struggled with it,” VLSI Research’s Puhakka said. “So did SMIC. It’s difficult what they are trying to do.”

Still, that move is essential to continue scaling. At 20nm, traditional planar transistors run out of steam. This is why in 2011 Intel moved to finFET transistors at 22nm. FinFETs are faster with lower power than planar transistors, but they are also harder and more expensive to manufacture.

Later, GlobalFoundries, Samsung, TSMC and UMC moved to finFETs at 16nm/14nm. (Intel’s 22nm process is roughly equivalent to 16nm/14nm from the foundries.)

Finally, after years of R&D, SMIC in 2019 reached a milestone by shipping China’s first 14nm finFETs. Today, 14nm represents a tiny percentage of SMIC’s sales. “Our customers’ feedback on 14nm is positive. Our 14nm is covering both communications and automotive sectors with applications including low-end application processors, baseband and consumer-related products,” said Zhao Haijun and Liang Mong Song, SMIC’s co-CEOs, in a conference call.

Still, SMIC is late to the party. For example, the application processor is the most advanced chip in a smartphone. Today’s smartphones incorporate application processors based on 7nm. Most other chips in smartphones, such as image sensors and RF, are based on mature nodes.

And 14nm isn’t cost-competitive for the most advanced application processors. “SMIC is starting to do 14nm. But if you look at smartphones, the designs are at 7nm,” said Handel Jones, chief executive of IBS. “If you look at the transistor costs at 7nm, a billion transistors cost from $2.67 to $2.68. A billion transistors at 14nm cost about $3.88. So you have a big cost difference.”

14nm is viable in other markets, though. “14nm technology can be used for low-end 4G and 5G smartphones, but not for mainstream or high-end smartphones. 14nm can be used for 5G infrastructure applications with the appropriate processor and system architectures,” Jones said.

Now, with funding from the government, SMIC is developing 12nm finFETs and what it calls “N+1.” 12nm is a scaled down version of 14nm. Slated by year’s end, N+1 is billed as a 7nm technology.

N+1 isn’t quite what it seems. “SMIC’s N+1 is equivalent to Samsung’s 8nm, which is slightly better than TSMC’s 10nm,” said Samuel Wang, an analyst at Gartner. “SMIC’s N+1 is unlikely for this year. 12nm may become production ready by the end of 2020.”

Once again, SMIC may miss the market window. By the time it ships 8nm in 2021, smartphone OEMs will move to 5nm for the application processor.

That’s not the only issue. SMIC could manufacture 8nm or 7nm using existing fab equipment. Beyond that, the current lithography equipment runs out of steam. So beyond 7nm, chipmakers require EUV, a next-generation lithography technology.

However, the U.S. recently blocked ASML from shipping its EUV scanners to SMIC. If SMIC can’t obtain EUV, the company is stuck at 8nm/7nm. “The U.S. blocked the EUV sale to SMIC (last year) under the Wassenaar agreement. I can’t envision a EUV shipment to China in the foreseeable future. But with 14nm just over 1% of SMIC’s sales, they don’t need EUV technology for a few years,” said Krish Sankar, an analyst at Cowen and Co.

At some point, though, China wants to go beyond 7nm. This is why China is working on its own EUV technology. China hasn’t developed a full-blown EUV scanner—it may never develop one. But work is underway in the arena. The EUV subsystems are being developed at several research institutes. For example, the Shanghai Institute of Optics and Fine Mechanics of the Chinese Academy of Sciences (CAS) last year described the development of EUV driven by a kilowatt laser. In 2020, researchers from the Institute of Microelectronics of the CAS published a paper on “EUV multilayer defect characterization via cycle-consistent learning.”

“There is a lot of research being done around different components of EUV,” VLSI Research’s Puhakka said. “I don’t think they have advanced to have a manufacturable EUV tool. Developing its own EUV will be a long process. I won’t say never, but it’s a long and hard road.”

Others agreed. “I assume that we see only part of what China is doing. It’s like an iceberg, most is hidden from view. Their academicians publish papers on EUV technology, but the work that I have seen has been mostly theoretical. I assume that there is some underlying hardware,” said Harry Levinson, principal at HJL Lithography.

Memory, non-memory efforts
China, meanwhile, has a huge trade gap in memory, namely DRAM and NAND flash. DRAM is used for main memory in systems, while NAND is used for storage.

China imports most of its memory. Intel, Samsung and SK Hynix operate memory fabs in China, which produce chips for both the domestic and international markets.

To reduce its dependence here, China is developing its domestic memory industry. In 2016, YMTC emerged with plans to enter the 3D NAND business. And CXMT is currently ramping up China’s first home-grown DRAMs.

Both are competitive markets, especially NAND. 3D NAND is the successor to planar NAND flash memory. Unlike planar NAND, which is a 2D structure, 3D NAND resembles a vertical skyscraper in which horizontal layers of memory cells are stacked and then connected using tiny vertical channels.

3D NAND is quantified by the number of layers stacked in a device. As more layers are added, the bit density increases in systems. But the manufacturing challenges escalate as you add more layers.

“There are two big challenges in scaling 3D NAND,” said Rick Gottscho, executive vice president and CTO at Lam Research. “One is the stress in the films that builds up as you deposit more and more layers, which can warp the wafer and distort the patterns. Then, when you go double deck or triple deck, alignment becomes a bigger challenge.”

Meanwhile, YMTC appears to have overcome some of those challenges. Last year, YMTC shipped its first product–a 64-layer 3D NAND device. Now, YMTC is sampling a 128-layer 3D technology.

The company is behind. In comparison, multinational vendors are shipping 92-/96-layer 3D NAND devices. They are also ramping up 112-/128-layer products.

Still, YMTC could become a factor, at least in China. YMTC’s chips are being incorporated in USB cards and SSDs from China-based companies. If Chinese OEMs adopt YMTC’s technology, “it could become a disruptive situation in NAND market share,” said Jeongdong Choe, an analyst with TechInsights.

To be sure, though, China has a long way to go in memory before it becomes a major competitor. “IC Insights remains extremely skeptical whether the country can develop a large competitive indigenous memory industry even over the next 10 years that comes anywhere close to meeting its memory IC needs,” said Bill McClean, president of IC Insights.

The same is true for analog, logic, mixed-signal and RF. “It will take decades for Chinese companies to become competitive in the non-memory IC product segments,” McClean said.

Meanwhile, several China-based GaN and SiC vendors have emerged in China. They appear to be foundry vendors and materials suppliers, but clearly, China is behind in the arena. GaN is used for power semis and RF, while SiC is targeted for power devices.

“The Chinese market represents a significant opportunity in the global power electronics industry, mainly in the automotive and consumer segments,” said Ahmed Ben Slimane, technology and market analyst at Yole Développement. “Driven by the electric-vehicle/hybrid-electric vehicle applications, SiC devices started to be adopted by leading Chinese car makers, such as BYD in its Han EV model. In the power GaN industry, the Chinese smartphone OEMs, such as Xiaomi, Huawei, Oppo and Vivo have opted for GaN in fast charger technology. Driven by strong system makers in China, Chinese wafer and device players are certainly well-positioned in terms of cost-competitiveness and increasing quality given the current context of the U.S.-China conflict.”

This in turn is fueling the development of the ecosystem. “Following the emergence of wideband-gap semiconductors in the power electronics market, China is indeed pushing for innovative technologies and it has started building up its domestic value chain,” said Ezgi Dogmus, technology and market analyst at Yole Développement. “In the Chinese power SiC ecosystem, we see various players getting involved at wafer, epiwafer and device level. This includes players such as Tankeblue and SICC in wafers, Epiworld and TYSiC in epiwafer and Sanan IC in the foundry businesses. Regarding the power GaN market, starting from 2019, we have witnessed entry of competitive GaN device manufacturers such as Innoscience and various system integrators in the domain of fast chargers.”

Packaging plans
China also has big plans in packaging. JCET is China’s largest packaging house. It has several other OSATs as well.

“China’s OSAT technology is quite current to the mainstream industry capability, perceived as a much narrower technology gap compared to front-end wafer fabrication technology. They are capable of supporting nearly all popular package types,” FormFactor’s Leong said. “The emerging 2.5D/3D heterogeneous integration technology is still under development in China, noticeably behind the industry leaders like TSMC, Intel and Samsung.”

Potentially, though, advanced packaging is where China could close the gap. This is not just in packaging, but in semiconductor technology.

Today, for advanced designs, the industry typically develops an ASIC using chip scaling. This is where you shrink different functions at each node and pack them onto a monolithic die. But this approach is becoming more expensive at each node.

The industry is looking for new approaches. Another way to develop a system-level design is to assemble complex dies in an advanced package. “As Moore’s Law slows down, heterogeneous integration with advanced packaging technology represents a once-in-a-lifetime opportunity for China to catch up in semiconductors,” Leong said.

https://semiengineering.com/china-speeds-up-advanced-chip-development/
 
the difference between "have" and "not have" is much bigger than the difference between "trailing edge" and "leading edge".

going from 0% market share to 15% market share within 20 years is a much, much bigger deal than going from 15% to 40% in the next 20.
 
November 2019

The number of TOP500 installations in China continues to rise and now sits at 227, up from 219 six months ago. Meanwhile, the share of US-based system remains near its all-time low at 118.
https://www.top500.org/news/china-e...mputers-us-holds-on-to-performance-advantage/

June 2020

China continues to dominate the TOP500 with regard to system count, claiming 226 supercomputers on the list. The US is number two with 114 systems;
https://www.top500.org/news/japan-captures-top500-crown-arm-powered-supercomputer/
 
New Top500 list,

TOP500 Expands Exaflops Capacity Amidst Low Turnover

FRANKFUR, Germany; BERKELEY, Calif.; and KNOXVILLE, Tenn.—The 56th edition of the TOP500 saw the Japanese Fugaku supercomputer solidify its number one status in a list that reflects a flattening performance growth curve. Although two new systems managed to make it into the top 10, the full list recorded the smallest number of new entries since the project began in 1993.

....

HPCwire also has an article on the new Top500 list,

The following excerpt from the above regarding China supercomputer may be of interest ,
As was the case six months ago, only two machines have crossed the 100 Linpack petaflops horizon (Fugaku and Summit). Four if you count the two (Sugon) Chinese systems that were nearly benchmarked over the last couple of years ago but not officially placed on the list (sources reported one system measured ~200 petaflops and a second reached over 300 petaflops). China has curtailed its supercomputing PR push in response to tech war tensions with the U.S. that came to a head 18-months ago.
 
Last edited:
2020 ACM Gordon Bell Prize Awarded to Team for Machine Learning Method that Achieves Record Molecular Dynamics Simulation
New Tool Simulates Interactions of 100 Million Atoms

New York, NY, November 19, 2020 – ACM, the Association for Computing Machinery, named a nine-member team, drawn from Chinese and American institutions, recipients of the 2020 ACM Gordon Bell Prize for their project, “Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning.”

Winning team members include Weile Jia, University of California, Berkeley; Han Wang, Institute of Applied Physics and Computational Mathematics (Beijing, China); Mohan Chen, Peking University; Denghui Lu, Peking University; Jiduan Liu, Peking University; Lin Lin, University of California, Berkeley and Lawrence Berkeley National Laboratory; Roberto Car, Princeton University; Weinan E, Princeton University; and Linfeng Zhang, Princeton University.

....
 
2020 ACM Gordon Bell Prize Awarded to Team for Machine Learning Method that Achieves Record Molecular Dynamics Simulation
New Tool Simulates Interactions of 100 Million Atoms


New York, NY, November 19, 2020 – ACM, the Association for Computing Machinery, named a nine-member team, drawn from Chinese and American institutions, recipients of the 2020 ACM Gordon Bell Prize for their project, “Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning.”

Winning team members include Weile Jia, University of California, Berkeley; Han Wang, Institute of Applied Physics and Computational Mathematics (Beijing, China); Mohan Chen, Peking University; Denghui Lu, Peking University; Jiduan Liu, Peking University; Lin Lin, University of California, Berkeley and Lawrence Berkeley National Laboratory; Roberto Car, Princeton University; Weinan E, Princeton University; and Linfeng Zhang, Princeton University.

....
Surprisingly NO member from IT SUPA POWA INDIA.
Only 1 non ethnic Chinese in this award winning team.

No wonder Trump so FRIGHTENED of the Chinese.
.
 

1613009885578.png


1. Introduction
Currently, the computation capability of No.1 supercomputer in TOP500 list 1) reaches hundreds of Petaflops. It is expected that the first exascale supercomputer will be debuted in around 2021, which will open the exascale era for high performance computing (HPC). However, because of the slowdown of both Moore's law [1] and Dennard Scaling law [2], the improvements of system performance and efficiency are becoming increasingly difficult, bringing unprecedented challenges on architecture design for exascale supercomputer [3-12]. As one of the leading teams of supercomputer research and development in China, Sunway has solid foundations on both theoretical research and engineering practice on the supercomputer design, which has been proven by the successful implementations of Sunway BlueLight [13,14], Sunway TaihuLight [15-19] and Sunway exascale prototype. The successful developments of Sunway supercomputers demonstrate that the comprehensive co-design for system architecture, including the processor, interconnect network, assembly structure, power supply, cooling system, system software, parallel algorithm and application support, is crucial to achieve optimal system performance and efficiency.​
....​
 
We were supposed to nominate the fastest 300 petaflop com last year but to prevent antagonising US and tone down the Chinese tech challenge threat, we had to hide our capability. How long can we keep doing this?
 
We were supposed to nominate the fastest 300 petaflop com last year but to prevent antagonising US and tone down the Chinese tech challenge threat, we had to hide our capability. How long can we keep doing this?

It is absurd to appease these genocidal white supremacists. Appeasement never works with these racists. It’s a weak move and one that must be reversed.
 

Pakistan Defence Latest Posts

Back
Top Bottom