Hand rubbing nuclear fusion live in the wilderness

Chapter 427 120 trillion times
In the computer room, Han Yuan explained the matters needing attention in the detection of carbon-based chips, while controlling the equipment in his hand.

He did not explain the testing process of nanoscale silicon-based chips because it is unnecessary.

After all, carbon-based chips are about to appear, and silicon-based chips are about to completely withdraw from the stage of history. It is meaningless to talk about the manufacturing and testing methods of silicon-based chips.

At least for South Korean won, it is meaningless for Huaguo.

Anyway, no need.

He doesn't know about other countries, but Huaguo and South Korean won can be sure, needless to say, as long as there is a chip whose performance is close to silicon-based chips, domestic chips will definitely be replaced slowly.

Not to mention carbon-based chips that outperform silicon-based chips.

We must know that the patents paid to Qualcomm, Intel and other companies by companies large and small across the country, as well as the cost of purchasing chips, materials, technologies and other things, are astronomical sums every year.

If Huaguo can develop its own chip, don't think about how to do it.

This point won is sure!
Of course, he also has enough confidence in the performance of the carbon-based chip he made.

Although the performance of the carbon-based chip he manufactured is definitely not as good as the standard in the "carbon-based integrated circuit board preparation information".

But even so, the performance of these carbon-based chips far exceeds that of ordinary silicon-based chips.

Ten times the performance cannot be achieved, but five or six times should be available.

With this performance, as netizens said, this is the grandpa of silicon-based chips.

In the computer room, Han Yuan started testing the carbon-based chip in his hand. After the basic failure test was completed, the first thing he did was to run the speed test.

This is what he cares most about, and it is also what the audience in the live broadcast room cares about most.

68 billion transistors, this number is nothing in a silicon-based chip, put it in 22 years, even the number of transistors in a medium-level chip is more than this.

But if it is put into a carbon-based chip, how powerful will it be?
Everyone in the live broadcast room, even those outside the live broadcast room, including Han Yuan himself, are concerned about this issue.

How strong is a carbon-based chip with 68 billion transistors?

Looking forward to it, the chip that has passed the basic fault test was placed in the corresponding test equipment, and the control program was started through the intermediate computer.

To calculate the basic performance of a chip, whether it is a silicon-based chip or a carbon-based chip, the most basic thing is the floating-point computing speed.

As for the testing methods of floating-point computing speed, he still used the two methods used to test silicon-based chips.

The first is a common metric algorithm.

The second is the numerical calculation of π.

These two methods are used from the magnetic core board computer at the beginning to the integrated chip computer later.

Even the nano-scale silicon-based chips manufactured some time ago are tested in this way.

In particular, π numerical calculations can reflect the performance of a chip or a computer more accurately than measurement algorithms.

After arranging the carbon-based chips, South Korea controlled the central computer to first open the program loaded with the 'measurement algorithm', and transfer the calculation function to the carbon-based chips arranged in the test equipment.

The rules for measuring algorithms are actually quite simple.

It is the multiple measurement calculation of several common orders such as constant order, logarithmic order, linear order, linear logarithmic order, square order, and cubic order.

比如让一块芯片计算1、2、4、8、16千平方次，需要的秒数分别进行测试一次，而后再将数据记录下来。

After multiple calculations and averaging, you can get a relatively accurate floating-point calculation speed.

This test method is relatively simple, the program is also simple, easy to use.

However, mechanized numerical calculations cannot fully reflect the performance of a chip.

Han Won still remembers the first transistor computer he made. At the beginning, he also used two methods of measurement algorithm and π value calculation.

The calculation speed given by the metric algorithm is about 600 million per second, while the calculation of the π value is lower, at more than 500 million, close to 600 million.

In addition, regarding the first transistor computer, this system in his mind also gives a calculation speed.

Han Yuan clearly remembered that the running speed given by the system was 720 million, which was several times higher than the floating-point computing speed measured by himself.

Such a large deviation between the two should be caused by the different measurement methods used. The measurement method used by this system is more efficient for the use of hardware.

In fact, Won really wanted to see how this system evaluates carbon-based chips, but that kind of test was only triggered once under a specific task.

The various chips he manufactured later have not been evaluated by this system.

Shaking his head, Han Yuan recovered from his thoughts, and the first round of measurement and calculation for carbon-based chips has been completed.

The measurement and calculation program has fed back the floating-point calculation data back to the central computer, the Korean Won called it, and a number appeared on the display.

"129562325125614 times per second."

Staring at the numbers that appeared on the screen, Han Yuan's lips parted slightly, and he counted over and over again, swallowing in disbelief.

At the same time, the audience in the live broadcast room were also blown away.

[One, two, three, four, five, six, seven, shit, I can't count, how many are these? 】

[Full fifteen digits! 】

[How many billions are fifteen digits? 】

[Fuck, 12 trillion calculations per second! 】

[I pinched my fingers and counted for a long time, isn't fifteen billions? 】

[Upstairs you missed a single digit, which is 120 trillion. 】

[Counting, it really is, 100 trillion! 】

[It's against the sky, this carbon-based chip. 】

[The speed of 120 trillion floating-point calculations per second, is it high or low? 】

[It’s very high. For example, the number of floating-point operations (single-precision floating-point computing power) of a common GTX1080 graphics card is 8.8-9TFlops (1080 trillion floating-point operations-11.5 trillion floating-point operations), GTX[-]TI has reached [-]TFlops ([-] trillion floating-point operations), do you think the anchor is awesome? 】

【英特尔最新的I9十八核处理器的运算速度应该已经达到了1万亿次每秒的水平，也就是1000GFLOPS多的样子，主播这一块碳基芯片能顶120九块的I9十八核处理器。】

[I'll be good, isn't this a bit too exaggerated? 】

[I remember that the supercomputer wind tunnel built in our country in 2012 seems to have a floating-point computing capability of 200TFlops (200 trillion floating-point operations), and this chip can support half a supercomputer. 】

[Strong and invincible, carbon-based chips! 】

[Silicon-based chips are finished, and the entire silicon semiconductor industry is finished. 】

Seeing the numbers displayed on the live broadcast screen, the entire live broadcast room became lively in an instant.

120 nine trillion floating-point computing power per second, which shocked everyone's jaws.

With a floating-point computing power of 120 nine trillion times per second, it can be said to be comparable to supercomputers in the 10s.

You must know that a supercomputer is assembled by tens of thousands of core processors and graphics acceleration processors.

But now such a small carbon-based chip can achieve such a performance, I can only say: "My lord, the times have changed!"

Not to mention the surprise of the audience in the live broadcast room, even experts from various countries gasped after seeing this number.

The computing power of a single chip can reach more than 100 trillion operations per second, which really crushes all silicon-based chips.

Even if this is single-precision floating-point computing power, it is really scary.

Seeing this data, the hearts of chip manufacturers such as Intel, Qualcomm, and AMD immediately became cold, and even practitioners in the entire silicon-based semiconductor industry felt cold.

As the netizens in the live broadcast room said, this is the end of silicon-based chips.

We must know that the best commercial chips currently available in the world, after unlocking the frequency limit and reaching overclocking, their single-precision floating-point computing power is about ten trillion times.

Even if there is a technical reserve, it will not be too high.

Moreover, overclocking computing also has great disadvantages. In addition to increasing energy consumption, it is also easy to accelerate the aging of computer hardware, and it is also prone to "electromigration" phenomenon.

Although the "electromigration" phenomenon does not immediately damage the chip, it damages the chip in a slow process.

But every electronic migration will more or less reduce the life of the CPU.

At the same time, in the case of overclocking, the possibility of system freeze or error will increase.

If a CPU can run for ten years under normal use, under the condition of overclocking calculation, the service life of the CPU can be one year, which is very good.

Staring at the numbers on the display, countries all over the world couldn't help but wonder whether this 'carbon-based chip' could run overclocked.

Obviously, the tests conducted by this anchor are all basic tests.

Metric calculation is one of the single-precision floating-point operations and is the simplest mode.

If the overclocking operation is to be opened, the method used will definitely be more complicated and detailed, so as to maximize the performance of the chip.

But thinking of this, these experts shuddered again.

Ordinary tests alone can reach a computing speed of more than 100 trillion times per second. If the limit can be unlocked for overclocking, how terrible will its performance be?
150 trillion times? 180 trillion times?Or 200 trillion times?or higher?
From the perspective of the performance of silicon-based chips, any generation of chips that have been launched and used commercially are relatively conservative in setting the chip frequency.

When most CPUs with new architectures first come out, as long as the performance is slightly better than their previous generation/competitors, the frequency can be set relatively conservatively.

However, under the premise that the process technology allows and the power consumption allows, the overclocking range may be very large.

For example, Intel's Core 2 Duo based on the Conroe architecture is much better than its own Pentium 4 and AMD's K10 because of its same-frequency performance.

Therefore, when it was first launched, the low-end E6300 was only 1.86 GHz, and even the high-end E6700 was only 2.66 GHz.

As a comparison, Intel's previous generation Pentium 4 with the same process technology has a 3.73GHz model. As long as the motherboard supports it, it is not too difficult for the Core 2 Duo series to exceed 3.6G.

For the E6700, the frequency increase has reached 35%, and if it is the E6300, it is as high as 93%.

However, at that time, the CPU was limited by frequency multiplier and motherboard and other accessories. For E1066 with FSB frequency of 6300, exceeding 93% means that it is still a bit difficult for FSB to climb to the index of 2000.

But it is generally possible to exceed 3GHz. Even so, the frequency increase is more than 60%.

If this data is substituted into the carbon-based chip in front of you, the overclocking power of up to 90.00% can make the floating-point computing capability of this chip easily break through 200 trillion times per second.

If it was placed in the 10s, a proper chip would be a supercomputer.

Even many supercomputers do not necessarily have the powerful performance of this small chip.

This speculation made experts from various countries swallow their saliva involuntarily, just like Won, and was frightened.

If calculated according to this data.

Silicon-based chips are really going to be swept in the trash.

Although the current anchor is only manufacturing chips used in computers, the manufacturing principles of chips used in computers, mobile phones, and other devices are generally interoperable.

As long as one is developed, the others can be solved quickly with the influx of funds.

It is difficult to create something, but it is much easier to imitate and modify it.

In the computer room, Han Yuan finally recovered after swallowing the air.

Not to mention the audience and experts from various countries in the live broadcast room, even he was a little frightened.

The floating-point computing power of 120 nine trillion times per second is calculated by the most basic measurement algorithm, but it also affirms the performance of the carbon-based chip.

Won knew that carbon-based chips would be better than silicon-based chips, but he didn't expect to be so good.

After all, there is still a big gap between his craftsmanship and the top craftsmanship in the 'Carbon-based Integrated Circuit Board Knowledge Information'.

In his original prediction, the carbon-based chip can achieve a computing power of about 80 trillion times per second, which is already very good.

But the reality gave him a huge surprise.

The first test data actually reached 120 trillion times per second.

After taking a deep breath, Han Yuan gradually calmed down and started the second measurement.

"Hopefully it's not false data or a blip."

Han Yuan stared at the screen and prayed silently in his heart.

Time passed bit by bit, and the data on the reality screen kept beating. In just a few minutes, the Korean won never felt so long.

Finally, the data jumping stopped, and the second measurement data was displayed on the screen.

"130010025122309 times per second."

130 trillion operations per second!

Looking at the data on the display screen, Han Yuan couldn't help but clenched his fists.

It's not an illusion, this carbon-based chip really has such a high number of floating-point operations!
The data from the re-test made the live broadcast room extremely lively, but Han Yuan was in no mood to answer the various questions of these sand sculpture netizens.

Now he just wants to complete the 'chip testing' and 'package testing' processes of the carbon-based chips as soon as possible. After confirming the stability, he will speed up the production of chips with various functions and assemble them into a new central computer.

(End of this chapter)