After completing the technological accumulation and validation, Tom immediately began constructing a new chip factory.
Compared to the previous chip factories, this one had a larger area; the entire industrial park, including various supporting facilities, covered a total area of 10 square kilometers.
This was not only because the factory’s designed production capacity was higher, reaching approximately 20 million 45-nanometer process chips of various specifications and models annually, but also because, compared to the previous 800-nanometer chips, the 45-nanometer chips now had a more complex process, more steps, and higher environmental requirements.
From the initial purification of silicon dioxide to the final chip packaging and ex-factory, all processes combined totaled over 1,000 steps!
Several hundred types of factories specifically supplying, powering, or providing assistance to this factory were built.
Such a massive production base would have required Tom to go to great lengths and spend several years to build in the past.
But now, with the maximum consciousness connection count soaring to over 5 million, this chip base was fully constructed and all equipment installation and debugging completed in just half a year.
All other auxiliary factories were also completed, so it immediately went into production.
At the same time, the other previously built chip factories, with their relatively outdated technology, did not cease production but continued to operate at full capacity.
The reason was simple: while 45-nanometer process chips were the most advanced and certainly good, they were not needed everywhere.
For example, a hydraulic control chip for an excavator could meet requirements even with a millimeter-level process chip; why use a 45-nanometer chip for that?
Such waste was simply not justifiable.
In fact, within Tom’s entire base cluster, the highest usage was for chips with lower and more outdated processes.
Thus, with various different process chips being produced simultaneously and factories operating concurrently, Tom’s overall chip production capacity, with the commissioning of this newest factory, successfully surpassed the 500 million mark, reaching an annual output of 520 million chips.
Even before the first batch of 45-nanometer chips was produced, Tom began the construction of another key project.
The supercomputer!
Supercomputers use a large number of chips for parallel computing, achieving performance far exceeding ordinary computers, and are indispensable key facilities in almost all fields such as industrial development and scientific research.
The AI technology Tom was about to develop also could not do without it.
Tom had a supercomputer before, which used about 20,000 800-nanometer process chips, achieving a floating-point operation capability of about 100 billion times per second.
This computing power was sufficient before, but it was no longer enough now. The source of this content ɪs novel⁂fire.net
This time, Tom planned to build a supercomputer using approximately 130,000 chips, with a floating-point operation capability reaching 400 trillion times per second!
It wasn’t that Tom didn’t want to build a larger one. Rather, for facilities like supercomputers, increasing computing power was not as simple as crudely stacking chips.
Compared to increasing the number of chips, what was more important was its parallel computing capability, that is, how to set up an algorithm and use corresponding hardware to effectively distribute computing tasks to each chip.
Making them work together was the most crucial aspect.
In addition, aspects such as memory bandwidth and capacity, storage system read/write speed, data transmission, and latency all constrained the supercomputer’s computing power.
Without breakthroughs in other technologies, sometimes adding more chips could actually reduce the supercomputer’s computing power.
Tom deployed thousands of Clones and performed extensive optimization before finally designing this supercomputing system.
Currently, this supercomputing center has been completed, just waiting for the chips to arrive.
After only a few days, the specially designed supercomputer chips were finally manufactured.
Transporting them to the supercomputing center, Tom controlled the Clones to quickly begin the installation work.
These approximately 130,000 chips were divided by Tom into 65,536 nodes, with two chips placed in each node, connected to each other using specially developed optical fibers, and the memory reached 64TB.
Powering it was a power station with an installed capacity of 15,000 kilowatt-hours specifically built by Tom.
The power consumption of this thing was terrifying. In just one hour, it consumed 15,000 kilowatt-hours of electricity, and in a year, it would consume 130 million kilowatt-hours.
With terrifying power consumption, the heat generation was, of course, also extremely terrifying.
For this, Tom specifically designed a water cooling system, laying a large number of pipes around the supercomputing center, with a total length reaching over a thousand kilometers and a coverage area of over ten square kilometers.
After installation was complete, with an order from Tom, it finally roared to life.
In an instant, countless calculations began to occur within each chip. With the change in the most basic circuit units, heat began to generate.
The heat generated by a single chip was very small. But when tens of thousands of chips were combined, their heat generation was extremely terrifying.
As a result, in the cooling pipes, the icy cold cooling water instantly began to heat up and flow outwards.
The surface of Loshen Star was perpetually at a low temperature of over -250 degrees Celsius. At this moment, through the pipes laid within it, the heat from the supercomputing center was dispersed into the ground of Loshen Star.
After being cooled, this water would then flow back into the supercomputing center, once again carrying heat and flowing outwards.
Through such repeated cycles, the supercomputing center could always operate at a suitable temperature.
At this moment, as the supercomputing center began to produce heat, the sensors deployed by Tom nearby keenly detected that this part of Loshen Star’s ground began to slowly warm up, and even some gases with lower melting points began to melt and diffuse into interstellar space.
If observed with an infrared telescope, one could even clearly see that the color of this area of the ground was darker, clearly distinguishing it from the surrounding areas.
If there were an ecosystem on Loshen Star, Tom could even say that this single supercomputer alone could change the ecology of this area.
After testing was complete and everything was confirmed to meet expectations, Tom immediately uploaded the learning algorithm written by thousands of programmer Clones to the supercomputer.
Subsequently, data generated by chips from every factory, every power plant, every scientific research base, planting base, even animal husbandry base, as well as every truck, train, tractor, excavator, spaceship, satellite, and so on—almost every single chip-generated data point—began to converge on this supercomputer.
This massive data was the "food" for this learning algorithm.
Fed with this high-quality data, this learning algorithm continuously iterated and optimized itself, rapidly maturing.