What a load of baloney. CXMT's DRAM is much higher density than Samsung DRAM at a similar process node. And in the later 16nm nodes Samsung uses EUV. Which CXMT has no access to. So claims it is a straight copy are plain bullshit.
Chinese semiconductor thread II
- Thread starter vincent
- Start date
Frankly I don't care where CXMT get the technology as long as they can keep progressing
Only a fool will care where they get the technology. Even if thennews is true ( which I think it is) who cares? It happens all the time, and almost every major power /country does that to get an edge or catch up to their competition . South Korea did this on an even larger scale towards Japan to catch up and take a lead over Japan with time , the US did it with Great Britain an a larger scale towards catch up and surpass them, in fact as recently as a few months ago Japan did it to taiwans TSMC to steal some of their techniques and technology (which caused a huge news in Taiwan) etc etc. it’s nothing new , countries will keep doing that. You just need to try and keep an edge over your completion by constant R&D since no technology remains static and new processes and techniques are developed each year and others who are lagging behind you will keep trying to poach your technology or even your staffs by every means they can, and there’s nothing you can do stop it short of imprisoning your staffs. lol. So only way to is keep a technology edge over others by constant research .
Huawei releases open source technology SINQ
The core benefits of SINQ are its fast, calibration-free, and easy integration into existing model workflows, reducing the memory requirements of models by 60 to 70 percent through a unique quantization method, depending on the model architecture and bit width.
This means that models that would have required more than 60GB of video memory to run can now run in an environment of about 20GB. Large models that would otherwise need to run on high-end enterprise GPUs such as the A100 or H100 can now run on more economical hardware, such as a single RTX 4090.
SINQ has been evaluated on multiple architectures and models, such as the Qwen3 series, LLaMA, and DeepSeek, and has performed well in benchmarks such as WikiText2 and C4, significantly reducing confusion and flip rates.
The core benefits of SINQ are its fast, calibration-free, and easy integration into existing model workflows, reducing the memory requirements of models by 60 to 70 percent through a unique quantization method, depending on the model architecture and bit width.
This means that models that would have required more than 60GB of video memory to run can now run in an environment of about 20GB. Large models that would otherwise need to run on high-end enterprise GPUs such as the A100 or H100 can now run on more economical hardware, such as a single RTX 4090.
SINQ has been evaluated on multiple architectures and models, such as the Qwen3 series, LLaMA, and DeepSeek, and has performed well in benchmarks such as WikiText2 and C4, significantly reducing confusion and flip rates.
Juchen Co., Ltd.: DDR4 SPD chips have been successfully introduced to several leading memory module manufacturers in the industry
Juchen Co., Ltd. has released a record of its investor relations activities. The company has nearly two decades of mass production experience in the field of memory module supporting chips. Since the DDR2 generation, it has been developing and selling SPD chips for supporting DDR memory modules. Leveraging its long-term technological accumulation, understanding of industry standards, and years of product and industrialization experience, it has established a clear technological advantage and has now become a core supplier of DDR2/3/4/5 SPD series chips worldwide. With its excellent product performance, reliable product quality, and comprehensive customer service, the company promptly seized the market opportunities brought about by the shortage of DDR4 SPD chips in the first half of 2022. DDR4 SPD chips have been successfully introduced to several leading memory module manufacturers in the industry, making Juchen one of only two companies in China qualified to directly supply supporting chips to leading memory module manufacturers.
According to public information, Juchen Semiconductor Co., Ltd. was established in 2009 and is headquartered in Zhangjiang, Shanghai. It was listed on the Science and Technology Innovation Board of the Shanghai Stock Exchange in December 2019. Juchen Co., Ltd. is a global high-tech chip design company with subsidiaries, offices, or sales offices in Silicon Valley, South Korea, Hong Kong, Taiwan, Shenzhen, Nanjing, Suzhou, and other regions. It has long been committed to providing customers with storage, digital, analog, and mixed-signal integrated circuit products and providing application solutions and technical support services. Its main product lines include non-volatile memory chips (EEPROM & NOR Flash), voice coil motor driver chips, and smart card chips. Its products are widely used in many fields such as smartphones, memory modules, automotive electronics, LCD panels, industrial control, communications, Bluetooth modules, white goods, and medical equipment.
Ultra High Precision Wireless Wafer Temperature measurements for EUV systems.
Domestic 12-inch silicon carbide breaks through, increasing production capacity by 2.5 times and reducing costs by 40%!
Recently, Jingsheng Electromechanical confirmed in its investor relations activities that the company's first 12-inch silicon carbide substrate processing pilot line was officially put into operation at its subsidiary Zhejiang Jingrui SuperSiC on September 26, 2025. This technological breakthrough marks a significant progress in China's third-generation semiconductor materials field.
The key breakthrough behind the pilot line's commissioning lies in the complete self-control of the entire supply chain, achieving independent research and development and 100% localization of all equipment, from crystal growth and processing to testing . All stages of the production line utilize domestically produced equipment and proprietary technologies. In particular, core processing equipment such as high-precision thinning machines, chamfering machines, and double-sided precision grinding machines were developed by Jingsheng Electromechanical over many years, achieving industry-leading performance. This marks the official establishment of a complete industrial closed loop for 12-inch SiC substrates, from core equipment to material production.
compared to the current mainstream 8-inch product, 12-inch silicon carbide substrate single-wafer chip output can increase by approximately 2.5 times , significantly reducing unit costs in crystal growth, processing, polishing, and other steps in large-scale production. Industry data indicates that after mass production of 12-inch substrates, the cost per wafer can be reduced by 40% , and the unit price of automotive-grade modules is expected to drop from the current $150 to $90 . This will provide key support for cost optimization in downstream applications such as new energy vehicles, photovoltaics, and 5G communications.
The key breakthrough behind the pilot line's commissioning lies in the complete self-control of the entire supply chain, achieving independent research and development and 100% localization of all equipment, from crystal growth and processing to testing . All stages of the production line utilize domestically produced equipment and proprietary technologies. In particular, core processing equipment such as high-precision thinning machines, chamfering machines, and double-sided precision grinding machines were developed by Jingsheng Electromechanical over many years, achieving industry-leading performance. This marks the official establishment of a complete industrial closed loop for 12-inch SiC substrates, from core equipment to material production.
compared to the current mainstream 8-inch product, 12-inch silicon carbide substrate single-wafer chip output can increase by approximately 2.5 times , significantly reducing unit costs in crystal growth, processing, polishing, and other steps in large-scale production. Industry data indicates that after mass production of 12-inch substrates, the cost per wafer can be reduced by 40% , and the unit price of automotive-grade modules is expected to drop from the current $150 to $90 . This will provide key support for cost optimization in downstream applications such as new energy vehicles, photovoltaics, and 5G communications.
I think it's just samsung using their control over the government to scare employees against jumping ship. All south koreans can only be slaves to samsung