Chinese semiconductor industry

Status
Not open for further replies.

ansy1968

Brigadier
Registered Member
Probably based on observation of major fabs in Taiwan , Korea and China, smic use high degree of US tools.

Chinese tools are not battle tested. No idea how its yield would go. Very possible there would be hiccups.

Like I said before, another option is go rogue and take over taiwan and Intel fab equipment in china and that would buy some time.

If 45nm completely domestic and give good yield on production level, hell, it's already a major breakthrough.
Hi TD739,

Like I said before, another option is go rogue and take over taiwan and Intel fab equipment in china and that would buy some time.

No need, its an act of desperation, right now Huawei and SMIC is in a lot pain but they are well prepared. It might take 2 years the most for the eventual production of 7nm, if you had the time you may read WTAN post and our discussion on NEWS On CHINA SCIENTIFIC AND TECHNOLOGICAL DEVELOPMENT thread. Latest news is that SMIC had a major breakthrough in 7nm tech, its a surprise ,skeptic and doubters are abound, only 3 FABS had that capability and they are all Asian with SMIC being the third. Defeating INTEL, with their advance EUVL from ASML, they are still stuck at 14nm and will only introduce its 7nm process possible 2023 -2024 timeframe with no guarantee. So you see China had the talent and resource to fulfill its need, what it needed is time. The US sanction will only accelerated the development with urgency to help Huawei and SMIC.

As I said before stick around and enjoy the show, it will be fun, 2021 will be an explosive year, you will learn more about China IC development and see all our discussion come to fruition.
 
Last edited:

Hendrik_2000

Lieutenant General
Who is this Innosilicon ? a little known company via broadsword

Innosilicon completes chip tape-out based on SMIC FinFET N+1
Alex Chen, Taipei; Willis Ke, DIGITIMES Tuesday 13 October 2020

China-based Innosilicon, a provider of high-speed mixed signal IPs and customized chip design services, has recently announced it has completed a chip tape-out and testing based on SMIC's FinFET N+1 process.

Innosilicon said on its website that all IPs are homegrown, having passed the one-off functional testing following months of collaborative efforts with SMIC, paving the way for volume production of the foundry's latest manufacturing node.

SMIC co-CEO Liang Mengsong revealed earlier that the firm's FinFET N+1 process is similar to 7nm node in terms of power consumption and stability, but some 35% lower in performance. Liang added that SMIC will introduce N+2 process later to support higher-performance applications.

SMIC stressed that its FinFET N+1 process, now pending validations at potential customers, is expected to enter small-volume production by the end of 2020.

Innosilicon now has three major product lines: IP solutions, ASICs and digitally encrypted server chips. It has been building an IP ecosystem in China for years, and has helped volume produce customized SoCs, GDDR6 chips and chiplets for heavyweight clients in China and abroad with manufacturing nodes ranging from 55nm to 7nm, according a China-based IT Home report.
 

ansy1968

Brigadier
Registered Member
Hi Hendrick_2000

I check google and lo and behold, they design GPU chips, maybe a major competitor for NVIDIA?

Innosilicon Innovates the Industry's First Next-Genneration GPU Targeting Cloud Computing
  • Sep 03, 2020
new-43.jpg

The fast-paced development of new infrastructure including 5G, cloud computing, and data centers and more, results in increasing demand for high-performance chips for desktop computers and servers. To address this expanding use cases, INNOSILICON is about to release two intelligent rendering GPU of the “Fantasy” series delivering high-performance datacenter-class graphics cards. Backed by years of technical expertise, the Innosilicon’s expert team has finished the design and will deploy solutions into production at scale within the year. The “Fantasy” series will be widely used in mainstream infrastructure deployment such as desktop rendering, 5G datacenter, cloud gaming, cloud office, cloud education and much more.
GPU stands for Graphics Processing Unit or graphics card chip, which enables all the fantastic dynamic 3D effects in games and engineering software. GPU not only performs various graphics processing tasks for personal computers, servers and mobile devices, but also, due to the emerging artificial intelligence, is widely used in the mainstream cloud computing fields like scientific computing, autonomous vehicles, intelligent analysis, password cracking, image recognition, big data, financial transactions and more.
The Innosilicon’s “Fantasy” series, soon globally available, powers the 5G infrastructure deployment with high performance, high security, and high reliability. Its built-in physically unclonable iUnique Security PUF information security encryption technology that improves data security and anti-attack ability, along with a series of global leading technologies like 16Gbps GDDR6 high-speed video memory, HDMI2.1 8K display and Cache-consistent Innolink Chiplet, will make their debut in the Fantasy GPU.
“Fantasy” GPU series, featuring floating point and intelligent 3D graphics processing, fully customized multi-level pipeline computing core, together with high-performance rendering and intelligent AI computing, can be interconnected with multiple chips for processing and flexibility enhancement. As a result, it is ideally suited for the 1080P/4K/8K high quality display for desktop applications, supporting VR/AR/AI, multi-server cloud desktop, cloud games, cloud office and other big data graphics application scenarios in the 5G era.
Leveraging the Innosilicon’s world-class long-term close collaboration with various advanced semiconductors and years of mass-production experience, the “Fantasy” series of high-performance intelligent rendering GPU chips will be rapidly iterated on multiple advanced processes, continuously and substantially optimized in the performance and power consumption according to the needs of customers, to adapt to the fast-growing software and hardware applications and speed up development and deployment of GPU-accelerated data centers and desktop computers.
About Innosilicon (INNOSILICON.com.cn)
“Ten years to sharpen a sword”, as a world-class one-stop provider of mixed signal IPs and ASIC custom solutions, Innosilicon team has been fully devoted to innovation for 14 years with a determined mind, focusing on global advanced process technologies. With its proven track record in the IP industry, Innosilicon has supported the mass production of billions of high-end chips.
In 2018, Innosilicon and NVIDIA was the first in the world to break the memory wall, overcome the GDDR6 high-bandwidth video memory technology bottleneck, develop fully customized computing cores, and successfully produce high-performance computing GPU products at scale. In 2019, Innosilicon announced availability of the HDMI2.1 IP supporting 4K/8K display and high-speed 32Gbps SerDes Memory IP. In 2020, Innosilicon was the first to launch INNOLINK Chiplet and HBM2E and other high-performance computing platform (supporting CPU/GPU/ NPU). With dozens of famous customers and more than billions of chips in mass-production, the remarkable innovation capabilities of Innosilicon have been proven in the fields such as high-performance computing, high-bandwidth memory, encrypted computing, AI cloud computing and low-power IoT.
 

ansy1968

Brigadier
Registered Member
Another one and aside from Huawei, they are a major customer of SMIC 14nm node, wow if SMIC can able to mass produce 14nm chips with ingenious content they are set for life. Now I know why investor are so upbeat about SMIC shares.

Industry Debut | Innosilicon’s Multiple High-Speed Interface IPs, Based on the SMIC 14nm Process, Proven in Mass Production
  • Jul 02, 2020 |
  • One-stop, multi-standard, first launch of domestic advanced IP mass production
news_32_top.jpg

Zhuhai Hengqin, China—Innosilicon has announced today that a number of high-performance domestic autonomous and controllable high-speed interface IPs, based on SMIC 14nm process, have been successfully one-time verified in the SOC products from domestic mainstream customers, and entered the commercial mass production.
Innosilicon IPs, focused on the domestic development, with 14 years of shipping experience of billions of high-end IP authorization, have gained competitive advantages, especially in the mainstream 28/22nm FINFET process, helping multiple top customers at home and abroad to achieve one-time mass SOC production.
With full adherence to the international general standards, Innosilicon’s high- security and high- reliability IPs can be customized and optimized in the size, power consumption and packaging of PPA to perfectly suit the customer application scenarios. Assisted by our fast one-stop turnkey integration, the success of customer products will be accomplished with differentiated competitive advantages. Innosilicon IPs have been widely used in the mainstream chips in the fields of the high-performance computing, AI, multimedia terminals, automotive electronics, IoT and etc.
The high-speed interface IP, which is first adopted by the customer in the SMIC 14nm process and put into mass production, includes MIPI DSI&CSI/LVDS combo IP, DDR4/3/LPDDR4/3 combo and other general IPs. Along with the multi-standard compatible IO, these combo IPs can also be customized according to customer requirements, optimized in the power consumption and area, which have passed the relevant functional tests in product application scenarios at one time, and quickly realized commercial mass production in new process——this is an industry first.
Up to now, Innosilicon has years of cooperation experience in IP ecological co-construction with the six major foundries in the world, achieving constant breakthroughs from 0.13μm, 65nm, 55nm, 40nm, 28nm to the advanced 14/12/8/7/5nm FinFET process, to provide one-stop high-speed interface IP with leading coverage and maturity, and the number of mass-produced chips reaches billions. In the field of mixed signal IP, Innosilicon will continue to enhance long-term cooperation with major foundries, to help customers provide more differentiated products with great social value to the high growth markets, and continue to empower the high-end domestic ecological chain.
In addition to the above-mentioned IPs, Innosilicon’s other high-security and high-reliability domestic IPs including USB3.2/USB2.0, HDMI2.0/2.1, eDP /VOB, PCIe3/4/SATA3, ADC, Audio Codec, Video-DAC, assist global customers with the successful mass production of SOC products in various mainstream advanced processes.
Innosilicon high-end combo IP includes:
  • Free combination of standards under 32Gpbs high-speed combo Serdes: PCIe4, USB3.1, SATA3, 10G ETHERNET, XAUI, CHIPLET, etc.;
  • Various multimedia IPs: MIPI CPHY/DPHY, Audio codec, MIPI/LVDS/TTL Combo, HDMI2.1 /eDP1.4 TX/RX solution;
  • World's top high-performance DDR combo IPs: DDR4/3/LPDDR4/3 PHY, Industry leading GDDR6, DDR5/LPDDR5, etc.
About Innosilicon:
Innosilicon, as a world class and domestic leading one-stop provider of high-speed mixed signal IP and chip customization, has supported billions of mainstream high-end SOC mass production at home and abroad, covering from 0.11μm to 5nm provided by the world’s 6 top semiconductor factories, with the market share leading for 10 consecutive years.
In 2018, Innosilicon was the first in the world to overcome the top difficulty GDDR6 high bandwidth data bottleneck, and successfully developed the applicable performance-leading computing GPU as well as the near-threshold voltage design for mass production, and then launched the INNOLINK Chiplet -- high-performance computing platform (CPU / GPU / NPU) of Chinese standard in 2020. We have effectively customized a number of chips in various FDX / FinFET advanced technologies, and have excellent innovation capabilities in the fields such as high-performance computing, high bandwidth storage, encrypted computing, AI cloud computing, and low-power IOT technology.
Innosilicon is fully committed to the autonomous and controllable localization of chips, serving global customers with high-security, high-reliability IP and customized products through a flexible and win-win business model.
 

Quickie

Colonel
Who is this Innosilicon ? a little known company via broadsword

Innosilicon completes chip tape-out based on SMIC FinFET N+1
Alex Chen, Taipei; Willis Ke, DIGITIMES Tuesday 13 October 2020

China-based Innosilicon, a provider of high-speed mixed signal IPs and customized chip design services, has recently announced it has completed a chip tape-out and testing based on SMIC's FinFET N+1 process.

Innosilicon said on its website that all IPs are homegrown, having passed the one-off functional testing following months of collaborative efforts with SMIC, paving the way for volume production of the foundry's latest manufacturing node.

SMIC co-CEO Liang Mengsong revealed earlier that the firm's FinFET N+1 process is similar to 7nm node in terms of power consumption and stability, but some 35% lower in performance. Liang added that SMIC will introduce N+2 process later to support higher-performance applications.

SMIC stressed that its FinFET N+1 process, now pending validations at potential customers, is expected to enter small-volume production by the end of 2020.

Innosilicon now has three major product lines: IP solutions, ASICs and digitally encrypted server chips. It has been building an IP ecosystem in China for years, and has helped volume produce customized SoCs, GDDR6 chips and chiplets for heavyweight clients in China and abroad with manufacturing nodes ranging from 55nm to 7nm, according a China-based IT Home report.

The report made the mistake of saying the process have 35% lower performance.

What Liang actually said was this:

When it comes to specific data, he revealed that SMIC's N+1 process has a 20% performance increase and 57% power consumption reduction compared to 14nm.

The 7nm process market benchmark performance improvement should be 35%, so the company's N+1 process is for low-power applications.

So, the difference in performance is 35% - 20% = 15% which is way more passable.

SMIC's N+2 process (using DUV) should be able to overcome this discrepancy soon and even surpass it with EUV.
 
Last edited:
Status
Not open for further replies.
Top