Two-dimensional is a plane, and three-dimensional is a solid.
Traditionally, Transistors are "etched" onto silicon wafers using photolithography. Making it three-dimensional is similar to "building" Transistors onto a chip.
The technical difficulty is obviously many times higher.
"The Multiple Exposure Technology I developed earlier seems to be suitable for photolithography of Three-Dimensional Transistor. Let's give it a try first."
Without going through complicated application procedures, without communication, without resource collection, Li Qingsong made a decision the moment Li Qingsong learned of this possibility.
As a result, many clones that were trying other solutions immediately put down their work and joined in trying this possibility.
The necessary resources for this attempt were immediately available without any delay, making the entire process extremely efficient.
After the initial design was completed, the prototype multi-exposure equipment immediately began etching a silicon wafer in the laboratory. This was followed by a series of testing procedures.
When the final result was presented to Li Qingsong eyes widened with delight.
"It actually works!"
After changing the Transistors from two-dimensional to three-dimensional, the Short-Channel Effect indeed disappeared! The performance of the Transistors and the stability of the current were improved again!
"This is it!"
Less than half a month after this idea came to him, Li Qingsong made a major decision based on the experimental data, which was enough to change the direction of the entire industry.
Three-Dimensional Transistor undoubtedly offer higher performance, but this also presents a problem.
That is, the technical difficulty has increased significantly compared to before, and correspondingly, the manufacturing process will be greatly extended to use this technology.
From silicon wafer purification to final chip packaging, the entire process involves more than 1,700 steps!
Assuming that each of these 1700 processes has a yield rate of 99%—which seems quite high—in reality, after 1700 processes, the yield rate of the final product is only 3.8 per hundred million, making it completely unusable.
Only by increasing the yield rate of each process to about 99.99% can the final yield rate be increased to about 80%, which is when it can be considered practical.
But achieving a 99.99% yield rate for each process... is incredibly difficult.
But there was no other way. Li Qingsong could only follow the most stringent design and construction requirements to build this brand-new Chip Factory.
Li Qingsong ultimately chose a basin site far away from any bases and Railway Line to ensure that even the slightest vibration would not be transmitted to this area.
The factory buildings of this huge facility have adopted the most stringent Radiation Protection measures to prevent interstellar radiation from interfering with the production line;
Every clone entering the production workshop must wear the most stringent dustproof suit to prevent bringing in any dust, and so on.
The construction of this huge factory, along with the subsequent installation and commissioning of equipment, took Li Qingsong a whole year.
A year may not seem like a long time, but for the highly efficient Li Qingsong, it is an unusually long period of time.
The Chip Factory has finally been built, and it can now mass-produce advanced chips with a 20-nanometer process.
I've only just mastered this technology, and at this stage, the yield rate of the Chip Factory is not high, currently only about 3%.
Its total production capacity is 10 million chips per year. However, only 300,000 of these chips meet the demand, while the remaining 9.7 million chips have various problems, leading to a decline in overall performance.
Should we throw these chips away? Or rework them?
Are you kidding me!
If a chip has a defect, you can mask that part using algorithms and programs, and the rest will still work.
Although the overall performance will decrease, even if it does, the overall performance is still better than even the most advanced 45 Nanometer process chips. It's still a good product.
Therefore, Li Qingsong used all 300,000 of the highest quality chips produced that year to build a new supercomputer, while the remaining 9.7 million chips were all distributed to factories, bases, and various spaceships and vehicles.
Through continuous production and optimization, the overall yield rate is slowly improving, and it is believed that it will soon reach over 90%.
With access to more advanced chips, Li Qingsong finally built the new supercomputer he had been dreaming of.
Compared to the previous generation of supercomputers, this supercomputer uses a much larger number of chips, a total of 160,000!
Its computing power reached 130 quadrillion floating-point operations per second, which is more than 30 times that of the previous generation of supercomputers!
After construction and debugging were completed, Li Qingsong couldn't wait to put the previously severely limited Combat AI into this supercomputer and start another battle experiment.
This time, Li Qingsong adopted a different fighting style than before.
His main body also arrived at the training ground, manipulating the A-side fleet with almost no delay. At the same time, he consciously delayed controlling the B-side fleet, giving it a 100-second delay.
However, the second party's fleet possesses a Supercomputer Spaceship that can receive real-time assistance from Combat AI.
Thus, one side had no Combat AI but also no delay, while the other side had Combat AI but also a delay. The two sides used blank rounds and simulated weapons to engage in fierce firefights in space.
With the application of new supercomputers, the fleet of Party B has seen a tremendous improvement in terms of flexibility, reaction speed, coordination, and cooperation.
Advance, formation, salvo, cover, suppression—various tactics were employed in turn, while a barrage of bullets, artillery, Interstellar Missile, Electromagnetic Bomb, and other weapons were used in turn.
Ultimately, the fleet directly controlled by Li Qingsong still won the battle.
But the victory was not easy.
Based on comprehensive assessment, with a delay of up to 100 seconds, the combat effectiveness of the second-party fleet, combined with delayed manipulation and Combat AI assistance, can reach approximately 90% of that of the first-party fleet.
Li Qingsong did not stop and continued the experiment.
This time, he increased the delay of the second-party fleet to 500 seconds.
In the next battle, the B fleet showed a significant decline in combat strength. In the final assessment, its overall combat strength was only about 70% of that of the A fleet.
However, without the assistance of Combat AI, if a fleet is controlled entirely by itself with a 500-second delay, its overall combat effectiveness plummets to only about 30%.
The Combat AI contributed 40% to the overall combat power boost!
"A 500-second operational delay corresponds to an operational radius of approximately 75 million kilometers."
From this moment on, with the assistance of Combat AI, even against a spaceship 75 million kilometers away, I can unleash 70% of the fleet's peak combat power!
That's enough!