Chapter 86: Tape-out Completed (Ch 828214888)

“How long will this take to build?”
After a few days, the lithography machine and production line were still under manufacturing, which made Chen Tong a little anxious.
After all, Chen Tong only needed one day to complete the construction of the screen panel factory and production line.
However, the lithography chip foundry in front of us has already taken almost seven days.
And judging from the current manufacturing progress, it will probably take some time.
“This is a very delicate construction task and it will probably take about one or two months. Please wait patiently, Mr. Chen!”
The head of ASML explained to Chen Tong with a smile at this time, after all, the assembly of lithography machines and the production of production lines require a lot of time to complete.
It would be difficult to build it completely without a month or two.
In this regard, Chen Tong can only accept the current extremely long construction time!
At this time, employees within Huateng Semiconductor also received a special confidentiality contract from the Legal Department.
Although the factory expansion has become a key company secret, the company’s internal employees have also curiously found out some clues.
Photolithography machine! Chip foundry production line! Delivery room!
These are undoubtedly shocking and heavy news!
On the second day after Huateng Semiconductor received the components delivered by car, the company’s employees were collectively required to sign a confidentiality agreement.
After the news was leaked, the amount of compensation that the leaker had to pay reached a astronomical level.
When Huateng Semiconductor’s chip foundry technology cannot reach the industry-leading level, Chen Tong still hopes to temporarily hide some of the company’s current top strengths as much as possible.
Of course, Chen Tong also understood that these things might be kept secret for a while, but the higher-ups would probably know about it soon.
However, Huateng Semiconductor is currently lagging behind in technology, and its chip foundry level is not even as good as that of Huaxin International.
As long as the company is careful and low-key and does not do anything illegal, it will naturally be able to continue to hide its strength and develop.
The factory is still under rapid construction, and new news has also come from Taiwan.
After nearly two weeks, TSMC finally completed the tape-out production of the two chips.
This time, the 7-nanometer wafer can be divided into 1,680 HT2018A chips, and through testing, the yield rate can reach about 90%.
That is to say, the number of chips with good yield in a wafer is only between 1500 and 1520.
Nearly 150 of the chips were defective, but fortunately these chips were just defective products rather than scrapped products.
After some tests, Li Zhongyu’s team who stayed in Taiwan found that these defective chips could still be used to generate some residual heat through some means.
The biggest problem with the defective HT2018A chip is that the current frequency of the core in the chip is too high when connected to the super-large core, which eventually leads to gate leakage.
The solution to this defective product problem is also very direct. Shield a core and directly change the core of the chip to 1+2+4!
Although the performance will be weakened, the original defective product will no longer be a defective product.
As for the low-end HT2018B wafer, after segmentation, nearly 2,180 chips can be obtained.
And the chip yield rate is almost close to 100%.
During the detailed testing process, only two chips had problems.
“At present, TSMC’s wafer foundry and packaging fees are about 50,000 per piece…”
Li Zhongyu also gave Chen Tong a report on all the subsequent fees that TSMC needed to pay.
Not counting the 30 million paid for tape-out, each subsequent foundry wafer production will cost 350,000 yuan.
Including the defective HT2018A chip, the cost of each chip is basically 208 yuan.
As for the mid-to-low-end HT2018B, the cost of each chip is 161 yuan.
“Send back the chip’s test data first…”
Compared with the chip’s foundry production cost, Chen Tong is more concerned about the chip’s specific data and power consumption.
Soon, Li Zhongyu sent all the specific data of the two chips and the defective products to Chen Tong.
As the company’s most proud top processor chip, HT2018A, in the latest BG5 test, the chip’s single-core performance score can reach 750, and its multi-core performance can reach 2980 points.
In the latest GPU Manhattan 3.1 performance benchmark test, its overall performance frame rate can reach 65.0.
The current mainstream Fire Dragon 845 has a single-core performance of 520 and a multi-core performance of 2375; the frame rate of GPU Manhattan 3.1 is 61.0.
That is to say, the overall performance of this chip is stronger than that of Fire Dragon 845, but in terms of GPU performance, it cannot create a big gap with Fire Dragon 845.
Compared with the most powerful A11 processor chip currently on the planet, the A11 processor chip has a single-core performance of 940 points and a multi-core performance of 2550 points.
The frame rate of the GPU 3.1 of Apple A11 is at 67.0.
Chen Tong looked at the data comparison in front of him and probably had a certain understanding of his own chip.
The competitor of the flagship processor chip HT2018A is not A11 and Snapdragon 845, but the next generation products of the two chips.
After all, it takes at least half a year for chips to go from foundry to sales and then to be used in mobile phone products.
This means that the company’s chip will not officially appear in front of consumers until the end of September next year at the earliest.
Judging from the current comparison, the next-generation A12 is definitely able to surpass itself, and the Fire Dragon 855 can also surpass itself in terms of GPU.
The multi-core performance of the processor chip that blocked the comprehensive defective version dropped to 2780 points.
At present, the company’s main chip is the HT2018B chip, which tends to be mid-to-low-end.
The single-core performance of this chip reached 630 points and the multi-core performance reached 2480 points. The overall CPU performance can catch up with the Fire Dragon 845.
The only shortcoming is in the GPU performance, which only scored 46.0 in the Manhattan 3.1 test.
In other words, the performance of HT2018B is very close to that of the flagship mid-range processor chip, Fire Dragon 845.
And because it uses a 7-nanometer manufacturing process, the chip’s power consumption performance is surprisingly good.
Chen Tong recalled the mainstream mid-range processor chips on the market next year.
Most processor chips basically use a process technology of about 10 nanometers, and even the core architecture of the chip adopts a dual large core plus six small core design.
7 nanometers + four large core design!
This is basically a combination design that is completely comparable to the flagship.
Chen Tong believes that with this combination of strengths, HT2018B will be able to emerge as a dark horse in the mid-range market next year.
The company can even rely on this chip to gradually come into the public eye.