Chinese semiconductor thread II

[tokenanalyst]

important progress in the research of ultra-steep vertical transistor devices​

Xidian research team used an ultra-thin two-dimensional heterostructure to create a VTFET semiconductor channel and vertically integrated it with a resistive threshold switch (TS) to realize an ultra-steep vertical transistor (TS-VTFET). This device technology relies on the excellent electrostatic regulation of ultra-thin two-dimensional semiconductors to greatly improve the device's gate control capability; at the same time, with the help of the voltage-controlled "insulating-conducting" phase change characteristics of the resistive threshold switch, the device's room temperature sub-threshold swing reaches 1.52 mV/dec, which is well below the theoretical limit of conventional MOSFET room temperature subthreshold swings higher than 60mV/dec. In addition, in the published proof-of-concept work, the ultra-steep vertical transistor prepared by the research team showed strong performance, including a current switching ratio higher than 8 orders of magnitude, a sub-60mV/dec current range exceeding 6 orders of magnitude, and a leakage current less than 10fA. It provides a new solution for high-performance and low-power transistor technology in the post-Moore era.

Please, Log in or Register to view URLs content!

 

[interestedseal]

I hope so. China needs at least another DRAM manufacturer. Just CXMT isn't enough.

Swaysure and Jinhua are both DRAM makers and HW-related. Not sure why HW has to operate two separate companies on Dram.

 

[tokenanalyst]

UNISOC and ZTE successfully completed the industry’s first joint commissioning of the 5G N102 core and network integrated solution​

​

Recently, UNISOC and ZTE successfully completed the industry's first joint debugging of the integrated core and network solution for the 5G N102 frequency band, including use cases such as 5G NR data calls, latency and peak rate tests. This is a major innovation achieved by both parties in 5G products and research and development, laying a solid foundation for promoting the application of the N102 frequency band in the 5G industry.

​

The N102 frequency band (5925-6425MHz) defined by 3GPP provides important spectrum resources for the popularization of 5G in the industry and the evolution of 5G to 5G-Advanced to meet the rapidly growing demand for mobile data.

The joint debugging of the above solutions is based on the industry's first chip platform V620 launched by Unisoc to fully support 5G R16 broadband IoT features. The chip platform has strong radio frequency capabilities and full network features, supporting up to NR 2CC and LTE 5CC; supports HPUE PC1.5; supports industry-leading 5G R16 features such as 5G TSN and 5G indoor high-precision positioning; has super uplink technology and supports Rich NUL and SUL combinations; equipped with a 4-core Arm® Cortex®-A55 CPU, the computing power is nearly 200% higher than the previous generation. This chip platform can be widely used in various forms of equipment such as 5G FWA, 5G handheld terminals, 5G modules, laptops, gateways, etc., bringing excellent solutions to vertical fields such as broadband applications, power, energy, and high-end manufacturing globally. 5G connection.
As the world's leading platform chip design company, UNISOC adheres to technological innovation as its core, adheres to the values of "professionalism, win-win, and struggle", and is committed to continuously creating high-quality products and innovative solutions for global industries and customers. Value, win-win development with global ecological partners, and use cores to create a better world.

Please, Log in or Register to view URLs content!

 

[henrik]

Quite credible rumors that HW took over Jinhua and started mass production a while ago

Where do they get their machines for production?

 

[tokenanalyst]

Where do they get their machines for production?

Mostly locally.

 

[gelgoog]

Recently, UNISOC and ZTE successfully completed the industry's first joint debugging of the integrated core and network solution for the 5G N102 frequency band, including use cases such as 5G NR data calls, latency and peak rate tests.

...
The joint debugging of the above solutions is based on the industry's first chip platform V620 launched by Unisoc to fully support 5G R16 broadband IoT features.
...
equipped with a 4-core Arm® Cortex®-A55 CPU, the computing power is nearly 200% higher than the previous generation.​

Can't UNISOC make something better than this? This is a CPU core from 2017.

Please, Log in or Register to view URLs content!

Heck HiSilicon is under sanctions and they use the ARM Cortex-A510.

 

[Wahid145]

Can't UNISOC make something better than this? This is a CPU core from 2017.
Heck HiSilicon is under sanctions and they use the ARM Cortex-A510.

I'm always so much disappointed by UniSoC when I see what HiSilicon is doing under extreme sanction by the "most powerful nation on the earth"

 

[tokenanalyst]

When you thought that these overpaid H.G. think tankers cannot get more idiotic you are rewarded with something like this, they are so lazy that they cannot even do their own research.

Let me answer for free. The evidence tells me that they have been running production lines with these domestic immersion lithography machines for years, the evidence points that most development now is ancillary and debugging. The immersion lithography project dates back to 2011, at first was supposed to be for 65nm node then upgraded to a 45nm node then to 28nm node and now with the release of SMEE new holistic lithography tools the evidence points out to something similar to the NXT2000i.

 

[tokenanalyst]

When you thought that these overpaid H.G. think tankers cannot get more idiotic you are rewarded with something like this, they are so lazy that they cannot even do their own research.

Let me answer for free. The evidence tells me that they have been running production lines with these domestic immersion lithography machines for years, the evidence points that most development now is ancillary and debugging. The immersion lithography project dates back to 2011, at first was supposed to be for 65nm node then upgraded to a 45nm node then to 28nm node and now with the release of SMEE new holistic lithography tools the evidence points out to something similar to the NXT2000i.

View attachment 126010

When the immersion lithography project passed "the acceptance" for a 65nm node way back in the early to mid 2010s that means that the technology back then, the wafer stage, the optics, the illumination system, the sensors, the light source and so on, was good enough for a 65nm process node (SE), then in 2017 when was updated for a 45nm process node (basically LELE) the same technology have developed enough that the project was upgraded to that process and the same goes for the 2020 the 28nm immersion (DP) the technology was good enough for that node AND NOW I think that same technology has advanced enough that they can use it for a 14-7nm process node (SAQP).
Is all about all the technology developed by the other players in the lithography project apart from SMEE, U-Precision, Qier Electromechanical, Gouwang optics and many others.​

 

[tokenanalyst]

AI chips using VeriSilicon’s NPU IP have shipped more than 100 million units worldwide.​

VeriSilicon today announced that artificial intelligence (AI) chips integrating VeriSilicon neural network processor (NPU) IP have been deployed globally. More than 100 million units have been shipped within the scope, and they are mainly used in 10 market areas including the Internet of Things, wearable devices, smart TVs, smart homes, security monitoring, servers, automotive electronics, smartphones, tablets, and smart medical care. In the past seven years, VeriSilicon has led the world in the field of embedded AI/NPU, and its NPU IP has been used by 72 customers in 128 AI chips in the above market areas.

VeriSilicon's NPU IP is a high-performance AI processor IP that adopts a low-power, programmable and scalable architecture design. It can be flexibly configured to meet customers' different requirements for chip size and power consumption, making it a cost-effective neural network acceleration engine. The IP is also equipped with an extensive and mature software development kit (SDK) that supports all mainstream deep learning frameworks to ensure that customer products can be quickly brought to market.

VeriSilicon’s latest VIP9000 series of NPU IP provides scalable and high-performance processing capabilities for Transformers and Convolutional Neural Networks (CNN). Combined with VeriSilicon's Acuity toolkit, this powerful IP supports all mainstream frameworks including PyTorch, ONNX and TensorFlow. In addition, it features 4-bit quantization and compression technology to address bandwidth constraints and facilitate the deployment of generative artificial intelligence (AIGC) and large language model (LLM) algorithms such as Stable Diffusion and Llama 2 on embedded devices.

Please, Log in or Register to view URLs content!