bloom america

At the end of the month, the Xilei Laboratory owned by Intee Company was in a state of joy.

"The animal energy is pretty good."

Catherine was also here at this time, and her eyes were shining.

The maximum frequency of the trial-produced Capricorn has reached 100mhz, but that is only a very small number of ultra-high-end CPUs. It can be said to be the limit that current CPUs can achieve.

As for today's most high-end CPUs, the highest frequency is only 100%, while the average product is around 10%.

What lies before Catherine's eyes is the flagship 42mhz version that is about to be launched on the market.

There is a big gap between laboratory products and official products.

And even intee does not dare to launch 100mhz products immediately.

After all, this is almost ten times the performance of today's mainstream CPUs, and most importantly, this is a 32-bit system CPU.

This also means that there will be an insurmountable cliff between the two.

And now, the launch of new CPUs with a minimum of 10mhz is also to turn the gully into a small one.

Fine seams.

This will help promote the product.

Overly advanced products will lead to a series of terrible problems, which may be fatal.

Catherine was able to make CPUs appear earlier, but it was impossible to develop these products faster.

This is just like the reason why intee has long developed a 100ghz optical processor in history, but regarded it as a technical reserve. It would rather develop an ordinary processor with an iuy architecture than launch such a killer optical processor.

Equipment that is too sturdy will push the company itself to the edge of the cliff.

100mhz products were launched as the next generation products.

According to Catherine's idea, it takes about a year and a half to launch a generation of products, Zodiac, which means a total of 36 years.

Well, wait until all the zodiac signs are finally released, then be lazy and launch a new zodiac sign, and then turn the old unsold CPUs into groups of twelve into "annual collections"

Then take it to auction...

Catherine's wishful thinking is crackling.

"As for the 100mhz experimental product, can the xing be stable?" Catherine asked.

"There are still some problems in the process." Robert replied: "Although the frequency is high, there are big problems in heat dissipation. And similarly, this CPU often makes errors that cause it to crash, and it runs at such a high frequency.

There will be a lot of defective products during production... Therefore, if a 100mhz CPU wants to be launched, it will have to wait at least a year until the process matures or a new process is adopted - then high-frequency CPUs can be promoted to

market."

Catherine nodded.

Originally, she wanted to get a 100mhz CPU to put in her office, but when she heard about the problem of frequent errors and crashes, Catherine felt that it would be better for her to choose a 42mhz product...

It was rare that Catherine wanted to enjoy the special treatment as a boss, but unfortunately, the reality was not as good as she imagined.

"If possible, for our next generation products, we should not worry about the main frequency first. We should consider the proportion of GPU in our processor. The next generation will not only be this generation, we will call the product that integrates CPU and GPU

It’s called APU, and both CPU and GPU will exist as processing units.”

Robert was convinced by Catherine's thoughts.

The frequency of current machines is sufficient. When the technology matures, 100mhz CPUs can be promoted. In fact, even if the technology matures, it may be necessary to wait until the market can accept it before launching it to achieve the best results.

At this time, it is obviously unwise to continue to increase the frequency.

To do a good job, you must first sharpen your tools. It may be better if the new apu architecture can be matured.

Therefore, according to the plan, the next generation CPU will not increase the main frequency, which means that the highest main frequency of the next generation computer will still be 100mhz.

But the next generation of CPU will also undergo a "revolutionary" creation and change.

Generally speaking, CPU is good at task allocation, data pre-reading, decoding, branch prediction, etc.

And the goals Catherine gave these researchers are also in these aspects.

The GPU has strong floating-point computing capabilities. Although Capricorn currently integrates both the CPU and the GPU, they still work independently. Although it has a CUDA architecture, in many cases, this architecture cannot be used to strengthen the CPU.

, in fact, CUDA is generally only effective for external graphics cards.

And because of architectural limitations, even if it is used in a CPU, the efficiency improvement will be very limited.

This is somewhat different from Catherine's idea. According to Catherine's idea, a processor that integrates GPU and CPU should be more awesome.

In order to achieve the goal of performance improvement, Catherine decided to streamline the tasks of the CPU to pre-reading and decoding, while the GPU will be responsible for the main calculation work. In this way, the CPU only pre-reads data and decodes for the GPU, while the GPU

Run at full speed and become the center of computing. If it can be realized, the computer will gain more than 20% performance improvement.

Let the gpu core perform computational functions while letting the cpu prefetch the gpu from off-chip main memory

The required data. This is more efficient because it allows the CPU and GPU to do what they are good at: the GPU is good at calculations, and the CPU is suitable for making decisions and elastically acquiring data.

In fact, the speed at which CPU and GPU obtain data from main memory is almost the same, but if the CPU can be allowed to judge the data required by the GPU in advance and use it to make preparations, the GPU can reduce its burden and devote itself to execution and calculation.

If you go, efficiency will naturally be improved.

In a GPU processor assisted by a CPU, after the CPU starts a GPU program, it will create a pre-execution program. This program is automatically generated from the GPU core using a compiler algorithm and contains memory access instructions for multiple thread blocks of the GPU core.

.cpu preexecution

The execution program will run before the GPU core. This is because: First, the CPU pre-execution thread only contains memory fetch instructions from the GPU core, without floating point calculations. Second, compared to the GPU scalar core, the CPU runs at a higher frequency.

High, the degree of instruction-level parallelism is also higher.

It doesn’t matter if you don’t understand it, as long as you know that it sounds powerful.

Arts students will not understand the resentment of science students.

All in all, in the new CPU, the proportion of CPU and GPU will become equally important, just like the weights on both sides of the scale are equally heavy, balance is the king.

In addition, another technology will also be applied to new CPUs, which is frequency multiplication technology.

Frequency doubling technology is still in the research and development process, but this technology is not difficult to research, so it can be developed quickly.

With frequency multiplication, there is no need to worry about the frequency problem. As long as the frequency multiplication can be achieved by 2 times, the maximum frequency of the CPU will reach 200mhz.

Of course, this frequency doubling technology cannot be used immediately.

Even with the new generation of CPUs, frequency multiplication technology will only be possible on high-end products and flagship products. As for ordinary versions, it is completely impossible.

According to Catherine's analysis of the current situation, the 100mhz CPU has been developed for half a year, but researchers cannot continue to increase the frequency. This means that the increase in CPU frequency has fallen into a stagnant period. If there is no way to find a higher frequency

Before advanced technology and architecture, it was impossible to increase the main frequency.

Therefore, after Capricorn, the main work of Intel will be to improve the efficiency of the processor, rather than simply increasing the main frequency. Efficiency is also very important. It is like using a 3.0ghz i3 processor than using two Pentium processors.

A four-5.0ghz dual-CPU system should be more efficient and smoother.

More importantly, Catherine has decided that frequency multiplication technology will be the main function of the next generation of CPUs. It was naturally impossible to popularize this technology before. In fact,

If we count from the time when Catherine tinkered with the CPU, she should have developed a relatively perfect CPU by now. You must know that it took so much time for the intees in history to already have a 386 CPU.

The 386 CPU is not inferior to Catherine's Capricorn. In fact, if it weren't for the fact that Capricorn is integrated with the GPU, in terms of performance, Capricorn would still be more outstanding.

But after Capricorn, the situation will completely change. The new generation of CPU will take a completely different path from the historical products. This path was completely developed by Catherine. This will be a path that has never been taken before.

road.

Now, as long as the research and development work of the operating system phoenix-www.piotian.comtee has turned to the core of a new generation of processors with more outstanding gpuxing and superior performance.

The new generation of APU will be named aquaniuo - Aquarius.

The plot in [China] will not develop further. If it develops further, it will be harmonious after being reported.

So the last chapter can only come to an end.

Two updates completed~

Next week I am going to build an e3-1230v2 computer. I have removed all the SSD from my laptop!.