After inspecting the carbon-based chip laboratory and gaining an in-depth understanding of the current carbon-based chip research at Xinghai Research Institute, the group left Qixia Industrial Park.
Returning to the villa at the foot of Purple Mountain, Xu Chuan shouted after entering the living room.
"Xiao Ling, help me collect relevant reports from this year's 'ISSCC International Solid State Circuit Conference', focusing on content related to processors and communication system-level chips."
As soon as he finished speaking, the response from the AI academic assistant sounded in the living room.
"Okay! Master."
As he walked forward, Xu Chuan walked towards the bathroom without stopping, intending to wash his face and wake up first.
ISSCC International Solid-State Circuits Conference, the full English name is International Solid-State Circuits Conference, is the highest-level conference in the field of integrated circuit design recognized by the global academic community and industry.
This conference, held in San Francisco, USA, in the first half of each year, is known as the "Olympic Conference in the field of integrated circuit design."
The organizer of the conference is the International Institute of Electrical and Electronics Engineers.
This summit, which began as the ISSCC International Solid State Circuit Annual Conference in 1953, has a history of more than 70 years. It has received great attention in international academic and industrial circles, attracting more than 3,000 industrial participants from all over the world every year.
and academic participants.
The most important thing is that in the more than seventy-year history of ISSCC, many landmark inventions in the history of integrated circuits were disclosed for the first time.
For example, the world's first TTL circuit was exhibited in 1962; the world's first integrated analog amplifier circuit was exhibited in 1968.
There are also GHz microprocessors, multi-core processors and other chips and designs that are more familiar to the public today, all of which were also exhibited at this world-famous summit.
And this kind of top-level summit is often not only open to the industry.
Many times, in addition to Intel, Nvidia, Qualcomm, Samsung and other top companies in the chip field, they will display this year's latest products at the venue. Top scholars in the semiconductor industry will also participate, giving public lectures or displaying paper technologies, etc.
It is no exaggeration to say that the products and technologies disclosed at this summit are the latest products of major manufacturers at that time, as well as the development path of silicon-based chips in the next few years.
...
After wiping off the water stains on his face, Xu Chuan sat in front of the computer and clicked on the report that the AI assistant had collected for him.
The 2025 ISSCC International Solid-State Circuit Conference will still be held in San Francisco, USA. More than 3,500 experts and scholars participated in this conference.
Xiaoling has already sorted out the important contents of this summit according to his instructions.
The first thing that caught his eye was the speech by Nvidia CEO Jensen Huang.
Standing on the spacious stage, Huang Renxun, wearing a leather jacket and with white hair, looked at the empty lecture hall on the stage with a smile on his face.
"Dear guests, I am very honored to stand here again. First of all, I would like to thank IEEE Solid State Circuits for providing us with this venue to hold the event..."
"Before I get into the discussion, I want to emphasize this: NVIDIA sits at the intersection of computer graphics, simulation, and artificial intelligence, which forms the soul of our company..."
"For the past two decades, we have been committed to the research of accelerated computing. For example, CUDA technology enhances the functions of the CPU and offloads and accelerates tasks that special processors can complete more efficiently."
"...The new Blackwell Ultra AI chip we launched this time is manufactured using HBM4 memory chips and TSMC's 3nm cutting-edge process. It is the core of the new generation of AI chips and supercomputing platforms."
"As you can tell from the name, it is a chip specially built for AI and supercomputing platforms."
“In terms of performance, Blackwell Ultra is fully capable of living up to this reputation!”
"It can provide at least 30 times performance improvement for large language model (LLM) inference workloads compared with chips of the same level of technology, and reduce costs and energy consumption by 25 times."
"To put it simply, if you buy it, for every 1 meter you invest in processing data in the cloud, you can get up to 60 times the performance improvement."
"The acceleration is 100 times, while the power is only increased by 3 times, and the cost is only increased by 1.5 times. The cost savings are real!"
"It is no exaggeration to say that the new Blackwell Ultra AI chip will be the most powerful chip in the world!"
When the speech came to this point, the audience was already in a commotion.
The new Blackwell Ultra AI chip, specially built for AI and supercomputing platforms, can provide up to 30 times performance improvement, and reduce costs and energy consumption by a full 25 times on the original basis!
If the performance of such a chip is really as good as what the NVIDIA CEO said, then it is indeed a super chip!
As we all know, with the further improvement of technology, the 7-nanometer process, which was originally limited by the quantum tunneling effect, is no longer the threshold limiting silicon-based chips.
However, the 5-nanometer chip preparation process is still the process limit of most foundries, and only a few manufacturers, such as TSMC, Samsung and other top wafer foundries, can produce 3-nanometer chips.
Especially after reaching the 3nm and 2nm processes, silicon-based chips face many problems.
Whether it is the limitations of the material itself or the increasing impact of the quantum tunneling effect on electrons, it is difficult to work hard on the chip manufacturing process to improve the performance of silicon-based chips.
For example, although TSMC adopts GAAFET all-around gate transistor technology in 3 nanometers, it does reduce power leakage and improve the performance and efficiency of the chip.
However, compared with the 5nm process, the performance improvement of the 2nm process is only about 10%-15% while maintaining the same energy consumption and complexity.
The new Blackwell Ultra AI chip created by Nvidia can provide up to 30 times performance improvement compared to the 5-nanometer chip they previously produced.
Although the performance of the chip is far from simply looking at multiples, after all, this is a chip specially built for AI and big data, and it will definitely be better at large language model (LLM) inference loads.
But based on the HBM4 memory chip, this new Blackwell Ultra AI chip can be called a black technology.
Seeing the shocked expressions on the faces of the audience and feeling the noisy discussion, a proud smile appeared on Huang Renxun's face.
NVIDIA...will shock this International Solid State Circuit Summit and the whole world!
He has already seen that tomorrow's stock will show that beautiful upward curve.
For him personally, this will be another beautiful increase in assets!
...
Sitting in the audience, Qualcomm's chief engineer Cristiano Anmon crossed his legs and watched his colleagues on the stage crazily showing off their products, with a mysterious smile on his face.
It has to be said that the new AI products launched by Nvidia this time are indeed very strong and can be called black technology.
But this year, he came with a mission!
Nvidia's Blackwell Ultra AI chip is very powerful, but it is only limited to AI and supercomputing platforms.
But Qualcomm is different!
Mobile platforms are the biggest core of competition in the global semiconductor industry today!
And this time he brought Qualcomm Snapdragon 8Gen4, which is manufactured using TSMC’s second-generation N3E process and equipped with Qualcomm’s self-developed Oryon core!
The most important thing is that the CPU frequency of Snapdragon 8Gen4 reaches 4.4GHz!
Compared with the previous 3.39GHz of Snapdragon 8Gen3, the main frequency of 8Gen4 has been increased by nearly 1GHz, far exceeding the 0.1-0.2GHz improvement of 8Gen3 to 8Gen2.
Nvidia's Blackwell Ultra AI chip is indeed very powerful, but sorry, Snapdragon 8Gen4 is the protagonist today!
Qualcomm's chief engineer Cristiano Amon has the same idea as Apple's director of hardware engineering Steve Saccoman.
They also bring their own killer weapons!
.......
In the study room, after watching the key report of the ISSCC International Solid State Circuit Summit, Xu Chuan forked the video on his computer.
No wonder Yuan Zhouli, the head of science and technology, told him this morning that he hoped Xinghai Research Institute could help share some of the pressure in the chip field.
If it weren't for the major breakthroughs in the carbon-based chips they were researching, this carnival in the semiconductor field would be simply disappointing.
Whether it is Nvidia or Qualcomm, or Apple and Samsung, they have come up with what can be described as a "unique skill" in their respective chip fields.
Compared with the toothpaste-like performance advancement in 2024, the progress of major manufacturers in the chip field in 2025 is simply explosive advancement.
Just one year is enough to be worth the past five or six years.
Whether it is the Blackwell Ultra AI chip launched by Nvidia, or the Snapdragon 8Gen4 with a new self-developed Oryon core launched by Qualcomm, or the new 5G chip launched by Apple, in the field of chips and semiconductors, each of them has a different impact on the finished products of previous years.
It can be called a black technology existence.
Especially for the domestic semiconductor industry, it is a technology that is like a dimensionality reduction blow.
Huawei's Kirin chip is indeed very powerful, and HiSilicon's design is also very powerful, enough to enter the high-end field.
But for Huawei, how to manufacture 5nm, 3nm and 2nm chips is the biggest problem.
After all, under the control of those unfriendly Western interest groups, wafer foundries capable of producing top-level 3-nanometer and 2-nanometer chips refused to accept orders, and even refused to provide them with low-nanometer, high-performance chips.
I have to say that the strength displayed by major semiconductor manufacturers at this International Solid State Circuit Summit is indeed a bit exaggerated.
Fortunately, Lady Luck is on their side.
A breakthrough in carbon-based chip technology is enough to change everything!
The electrical conductivity of carbon is better than that of silicon. In comparison, carbon-based chips can provide higher-speed computing and higher storage capacity.
Carbon-based chips made of carbon nanotubes are more suitable than silicon-based chips for use in rapidly developing technological fields, both in terms of superior performance and lower power consumption.
More importantly, this is a blank field with very broad prospects for future development.
Compared with silicon-based chips, the upper limit of carbon-based chips will be higher.
In addition, with the continuous development and maturity of technology, the technical difficulty of manufacturing carbon-based chips will be reduced, and the cost will also be reduced.
There is no suspense to say that now that finished carbon-based chips have appeared, they truly control the future information field!
The thoughts in his mind were spinning, and a smile appeared on Xu Chuan's lips.
After closing the reports related to the ISSCC International Solid State Circuit Summit, he created a new paper report.
After thinking for a while, he hit the keyboard.
"High-performance carbon-based chips based on high-density integrated carbon nanotube arrays!"
After tapping on the keyboard a few times, an easy-to-understand title quietly came into Xu Chuan's eyes.
This chapter is not finished yet, please click on the next page to continue reading the exciting content! Looking at the title of the paper on the screen, the corner of his mouth curled up.
Since he is needed to help share some of the pressure in the semiconductor field, and he happens to have this ability now, why not take advantage of this opportunity to further promote the influence of the "Discovery" journal!
I believe that no one will miss the next latest issue of "Exploration·General Issue" and "Exploration·Materials" of "Exploration" journal!
.......
With the help of an AI academic assistant, I spent most of a day writing the paper "High-performance carbon-based chips based on high-density integrated carbon nanotube arrays!"
Compared with papers on basic subjects such as mathematical physics, industrial-oriented papers with finished products and detailed experimental data are much easier to write.
Whether it is the paper itself or related experimental data, there are complete documents for reference.
After all, the carbon-based chip laboratory has actually prepared a finished chip integrating 10 million transistors per square millimeter with a frequency as high as 5.8GHz.
It couldn't be easier for him to just sort it out and then simply use some of his own understanding and language to describe the future of carbon-based chips.
If it weren't for the fact that some experimental data or places may involve some confidentiality, writing this kind of paper could be faster.
After asking Xiao Ling to check it, Xu Chuan completely checked the newly released paper. After confirming that there were no problems, he compared it with another paper, "High-density integrated carbon nanotube array based on carbon-based semiconductor materials."
Methods" was sent to Ouyang Ji, deputy editor-in-chief of Discovery, through a dedicated intranet.
The former was published in "Exploration·General Journal", and the latter was used as the opening paper of "Exploration·Materials".
I believe that in the next period of time, the academic and industrial circles will be lively and even crazy!