bloom america

Chapter 299 GPU developed successfully!

After discussing it with the UK for a while, Catherine handed over all the remaining work to Ralph, which ultimately required the intervention of the American telecommunications company.

"Why do British people need the Internet?" Elsa seemed a little confused.

"Aren't you from England?" Catherine was a little funny.

"So I don't even know what they need the Internet for. The UK is not as big as the United States, and communication is also very convenient. When my wife and I went to Wales, it only took more than an hour."

Compared to Americans, Britain is indeed very small.

"Perhaps they want to strengthen the ties between Europe as a whole. They probably don't want to leave any opportunities for the Soviet Union to take advantage of."

Although she said this, Catherine had a strange feeling in her heart. Although Europe's move seemed aimed at the Soviet Union, on the other hand... it seemed that some of it was aimed at the United States.

In fact, the EU and the euro appeared in a sense in Europe to fight against the United States and the US dollar, and it was not because of the Soviet Union.

"I guess so."

The matter ended here, and Catherine's work returned to normal. At this time, Catherine was discussing the cloud computer issue with Olson.

——California has done very well in popularizing the Internet, especially in Los Angeles. Similarly, New York has also built a huge intra-city interconnection network. It is completely possible to set up a cloud computing system.

At this time, Olson was calculating the cost for Catherine: "Assuming that each city has 10,000 users, the research and development cost of a cloud machine is about US$10 million, and the cost is about US$2 to 5 million.

Excluding R&D costs, it takes about 2 years to recover the cost of a cloud system, and then it can continue to operate for another two years. If R&D costs are also included, we need to invest in at least five or six big cities.

Only in this way can we make money by using cloud systems. After the four-year lifespan is up, the speed of development of our computers will force us to choose to eliminate the original cloud computing machines."

Compared with the current computer development and the future, they are simply two worlds.

In the 21st century, the highest floating-point computing speed of personal computer graphics cards has reached 8TL, which is already comparable to that developed by IBM in 2000, and it only took ten years.

This situation would have been unimaginable before the 1980s.

——This is also because the functions of current computers are very poor. Without a qualitative leap, it is difficult to see any advantages.

If we place it in the 21st century, maybe within three years, or even two years, we will have to choose to purchase new machines or upgrade them.

Olson continued to add: "However, we don't need to invest too much in upgrading. If the number of users of the previous machine is reduced to expand the machine's performance, it can still be sustained for a period of time, and it can probably be reduced.

It’s about a quarter of our cost.”

The supercomputer requirements of cloud systems require a certain degree of flexibility, which is very effective in reducing costs.

"But there is another problem... Although our cloud system is expected to load up to 10,000 people, in that case, the speed of the computer will inevitably be affected."

"Well, system optimization must be done. However, the possibility of being online at the same time is not high, and we can also let other computers bear the burden of calculations. If it is not possible, we can reduce the number of people. Is it okay to have about 8,000 people?"

Olson nodded: "Okay, we will re-calculate and evaluate this."

When Olson went down, the phone suddenly rang.

Catherine picked up the phone, and the voice from the reception desk came from the other side: "Ms. Edson, Mr. Roberts from Xunlei Laboratory is calling for you. Do you need to put the call in?"

"Um."

Catherine responded.

"Miss Edson! Good news! Oh, my God, this is so wonderful!"

Roberts on the opposite side seemed very excited.

"Calm down, Mr. Roberts."

Catherine is a little strange.

"Oh...oh, okay, I understand." The other person seemed a little embarrassed.

"What's wrong?" Catherine asked.

"The GPU you mentioned has been developed. Its floating-point computing performance is twice as high as the current SFC!"

The voice on the other side had a high pitch and seemed really happy. Catherine was also impatient: "Okay, I'll be there right away, Ma Shan."

Xunlei Laboratory is not inside Noah, but is directly adjacent to Rama Laboratory. However, both locations are not too far from Noah, and it only takes the first ten minutes.

"Roberts! Is it done?"

As soon as she saw Roberts, Catherine couldn't wait to ask.

At this time, Roberts seemed to have lost the nervousness on the phone. He smiled, which seemed a bit mysterious: "Please follow me and I will show you something."

Catherine followed Roberts and entered the interior of the laboratory.

Roberts took Catherine to a 14-inch monitor, and then showed Catherine what was on it.

"This...this is..."

Catherine was shocked when she saw what was on it - it was actually a full 3D scene!

"It took us a lot of time to make this thing." He smiled: "This is a fully 3D scene. We can watch it from any angle."

The scene is very simple, it looks like a palace with gray and white walls, but it's pretty good to be able to do this.

"It's really bad."

Catherine is also very happy.

Americans have a strange love for 3D. Even in the black and white computer era, they used lines to build 3D scenes. For Americans, 3D technology is definitely a big killer.

"However, this GPU is only the most basic experimental version. After linking with the CPU, the maximum computing speed is only 1mips."

"One million times per second..." Upon hearing this sentence, Catherine seemed to have thought of something: "You mean our CUDA platform is also completed?"

The other party nodded: "We have always designed the GPU based on the CUDA architecture."

Catherine nodded excitedly.

Research on GPUs is much simpler than CPUs. Not only that, the performance of GPUs is also much better than CPUs. Because of this, Catherine seems so happy.

CPU and GPU are two directions, and there is a big difference in development. GPU only needs to consider floating point operations, so it can enhance xing performance to the limit, while CPU not only needs to consider calculation problems, but also consider adding (super)

Pipeline design, multi-shot (luàn sequential shot) system, instruction simplification and optimization, and separate design of floating-point arithmetic and integer arithmetic, etc. When it comes to CPUs on smartphones, sometimes 3G needs to be considered

support etc.

This is like running between a heavily armed soldier and an athlete wearing a thin shirt. The former is definitely not as good as the latter.

That's why Catherine is so excited after having the GPU - this means that today's 16-bit machines can already do some of the things that 32-bit machines will be able to do in the future.

Hey, is there anything more outrageous than this?

"We have tried applying it to nimbus game consoles. Our game console's resolution can reach 16 bits (65535sè). If it is on the same screen, it can display up to 2000sè and the maximum resolution can reach 640x480. But we only made one

test program."

Catherine nodded: "Take your time, don't rush."

Being able to reach 16-bit color is enough, and Catherine can be said to be satisfied with it.

If you can have this GPU, the future of 16-bit computers will definitely be more perfect.

——But Catherine also knows that GPU is a product that only thrives on 32-bit computers.

"Roberts... have you ever tried to build a supercomputer using GPUs?"

Upon hearing this, Roberts' eyes lit up.

"Oh...yes, if such a GPU is used in a supercomputer...oh, my God, this is simply a revolution!"

Yes, Roberts also realizes the value in this.

With the current SFC, it is possible to build a supercomputer with 100mfl - that is, 100 million floating point operations per second. If you use GPU acceleration... it is possible to increase it ten times!

——If placed in the original history, a 1gfl billion supercomputer would not have been developed until 1983 by Seymour Cray.

"But this is just an experimental product. It may take a month or two for the real product to be released."

Roberts seems to have thought of this again.

"Don't worry, the company will fully support you."

Since this was just an experimental product, Catherine didn't want to disturb the other party's development of a formal GPU, so she left early.

After returning from Xunlei Laboratory, Catherine found Seymour Clay.

"Clay, do you think it would be feasible if we used GPUs to develop supercomputers?"

Clay was stunned for a moment by Catherine's unclear question.

"Theoretically it's possible. If your GPU theory is correct...wait a minute, has the GPU been developed?"

Catherine smiled and nodded.

"Oh...can I have this product?"

Catherine shook her head: "Not yet. The product development in Xunlei Laboratory is still in the completion stage, and it will take another month or two to complete."

"I seem to be a little impatient."

"No need to rush."

Catherine suddenly felt that she seemed to be a little too excited. After Seymour Clay left, Catherine washed her face and calmed down.

"Oh, I seem to have trouble sleeping and eating...I really want to see the finished product as soon as possible!"

Catherine complained to Elsa.

…

There is another update, which will be a little later.