America 1978: Starting with Hand-Crushed Chips
Chapter 118 Finished Product
"I don't care about the yield rate. Even if only one chip works, I'll consider it a success." Tang Mu took out 20,000 US dollars from his bag.
"I need a week's worth of your wafers, super expedited, I want to see the wafers produced within seven days."
During the semiconductor winter of 1980, both Intel and AMD were laying off employees, and small foundries like Mike, which depended on large companies for survival, were in dire straits.
He offered a price far exceeding the market rate, partly to ensure confidentiality and partly because he didn't want to owe Professor Lawson too much of a favor.
Mike wanted to refuse, but the money was enough to cover next month's loan interest and staff salaries.
"Fine!" he replied through gritted teeth, "but let's get this straight: your mask is non-standard, and you set the process yourself. If it's a failure, I won't charge you the $20,000, but you'll still have to pay for the silicon wafers and electricity."
He then looked at Lawson beside him and asked, "Teacher, do you think this is alright?"
Lawson replied casually, "Business is your business. As long as you keep it confidential, there won't be any problem."
……
For the next week, Mike practically lived at the factory, keeping Tang Mu company as he went crazy.
The polished silicon wafers are fed into a high-temperature diffusion furnace, where a blue oxide layer slowly covers the silicon substrate at 1000 degrees Celsius.
Tang Mu stood in front of the old Cobilt exposure machine, manually adjusting the micron-level knobs through a binocular microscope, pressing the first mask onto the silicon wafer.
"Exposure!" The ultraviolet light flashed, and the shadow of the circuit was etched onto the photoresist.
The fourth day arrived quickly. Tang Mu was extremely cautious as the most dangerous step, annealing after ion implantation, was about to begin.
In pursuit of the electron mobility required for the RISC architecture, Tang Mu forcibly modified Mike's process formula, raising the temperature of the diffusion furnace by a full 20 degrees and shortening the time by 10%.
"Are you crazy! You'll warp the silicon wafer!" Mike stared at the alarming thermometer, refusing to press the confirm button. "Once the silicon wafer warps by more than 50 micrometers, the underlying metal layers won't align at all!"
After spending the last few days with him, Mike is indeed an expert in this field, but he has a rather bad temper.
In other words, he has a foul mouth, and Tang Mu thinks that his business failures may also be related to this.
"I'll accept it even if it's bent!" Tang Mu pulled the microphone to input the parameters. "What I want is ultimate speed; I won't sacrifice speed for yield!"
The furnace door closed, and the red light illuminated Tang Mu's sweat-covered face.
On the seventh night, when ten gleaming 4-inch wafers were removed from the vacuum sputtering chamber, Mike finally shut up.
Although the edges are slightly warped and the color is not as uniform as Intel's products, it looks exceptionally simple.
"Let's test it directly." Tang Mu turned and walked towards the probe station.
He placed the wafer under a microscope, and several tungsten needles slowly pierced the pads of one of the wafers.
Mike held his breath and stared at the oscilloscope.
"Short circuit".
The ammeter went straight to the bottom, and the chip burned through.
Tang Mu was already prepared. He moved the probe station expressionlessly: "This one is dead too."
"Electrical leakage."
"Logic gates do not flip."
One after another, Mike's heart sank lower and lower. He tested twenty in a row, and they were all defective.
In a commercial wafer fab, this would have been shut down for maintenance long ago.
"Hmm?" Suddenly, Tang Mu's hand stopped.
The current reading of this one is stable at 200 mA.
"Connect Vcc to 5 volts, input clock signal, and pull Reset high," Tang Mu issued a simple instruction.
A straight line lit up on the oscilloscope screen, and then the line began to jump, forming a square wave as strong as a heartbeat.
The frequency meter next to Mike started jumping wildly: 1MHz...5MHz...8.2MHz!
Mike, watching from the sidelines, was dumbfounded: "My God...unbelievable..."
"Intel's 8086 is only 5MHz, yours is 60% faster than Intel's!"
He couldn't believe his eyes. This meant that if the instruction set efficiency was really as Tang Mu said, the actual computing speed of this chip would be 10 times or even higher than that of the 8086!
Mike composed himself and continued helping with the testing.
Ultimately, out of the more than 200 chips in this batch, only 15 could operate stably at frequencies above 8MHz, resulting in a yield rate of only 7.5%.
But Tang Mu didn't care at all. Successfully fabricating the chip on the first try was a great victory!
He used a slicing knife to cut off the precious bare pieces, carefully placed them into the black ceramic shell, and sealed it with gold wire.
Then he took out a marker and wrote on the casing.
"R-Star 01, 1980, Sample."
……
Stanford was silent in the dead of night, but many windows in the Terman Center were lit.
The office was filled with smoke, and Lawson was gazing out the window, lost in thought.
"Come in." Upon hearing the knock, Lawson put down his pipe.
"Professor Lawson, I've made it!" Tang Mu pushed open the door and entered. Although his eyes were bloodshot, his spirit was as strong as ever. "As promised, the first one is for you."
As he spoke, he opened the box and took out the packaged chip.
Lawson picked up the chip and examined it under the light: "Is this the monster you were talking about... that only requires thirty commands?"
Because he lacked the necessary motherboard and software, he had to borrow a Tektronix logic analyzer from the office to perform electrical characteristic tests.
"Professor, look at the rising edge."
Lawson stared at the screen; the rising edge of the waveform was as steep as a cliff, with almost no overshoot or ringing visible.
"8.2MHz...and that's without any process optimization."
Lawson's eyes were filled with disbelief: "If it's Intel's process, it can run at 12MHz."
But after the initial excitement, as a scholar deeply involved in computer architecture, Lawson quickly analyzed the chip with a critical eye.
He put down the chip and asked the most critical question: "Tang, RISC is indeed very fast, but all the software on the market now, VisiCalc, WordStar, are based on CP/M, which is all CISC instruction set."
"Your race car is certainly excellent, but all the roads in the world were built for horses (Happy Year of the Horse!). Without roads, how can you commercialize it?"
Innovation is valuable, but only useful innovation can ensure that a project doesn't fail at the outset. There must be sufficiently suitable application scenarios for large-scale production.
This is a dead end faced by all non-CISC instruction set chips: without an ecosystem, no matter how strong the performance, it is still electronic waste.
"Also, you cut the floating-point unit to simplify the instruction set, right? Then we can only rely on software simulation."
Tang Mu was amazed by Lawson's gaze. Intel in later generations did not make the best chips, but it was precisely because of backward compatibility that it made him laugh the longest.
"Professor, I don't need general-purpose software," he said, already prepared. "I'll use fixed-point arithmetic to shift all the decimals 16 bits to the left to become integers, and then shift them back to the right to restore them."
"Although I sacrifice some precision, I have a speed advantage of a hundred times in graphic coordinate transformations."
You'll Also Like
-
Kamen Rider: High Beam Rider - Starting with Saber.
Chapter 417 8 hours ago -
Fairy Tail: Capturing the Demon Mirajane at the Start.
Chapter 143 8 hours ago -
No job: Check the cards, Loki, are there any problems with the cards?
Chapter 169 8 hours ago -
Crossing Tokyo, I have a nine-tailed fox inside me.
Chapter 138 8 hours ago -
Jujutsu Kaisen: The Flame Mage Who Starts with Gacha Pulls
Chapter 303 8 hours ago -
In Warhammer, throwing a muddle is impossible.
Chapter 743 8 hours ago -
He's courting disaster! A coward forced himself on a tyrannical warlord.
Chapter 164 8 hours ago -
Damn it! The girl he likes insists on dragging him to a blind date.
Chapter 150 8 hours ago -
After we broke up, my ex-girlfriend gave birth to a daughter.
Chapter 150 8 hours ago -
Courtyard Houses: Rewards for Reporting? I, an Outlaw!
Chapter 650 8 hours ago