Rewrite the Technological Landscape

In terms of the baseband chip, Intel has just done a favor to the Dafeng Group, and in another matter, Intel has actually done a favor to the Dafeng Group, although Intel itself probably didn't realize it at the moment.

That is, in September 2007, Intel would acquire Havok, one of the world's three major physics engines, for 110 million yuan, but in September of this life, Havok was acquired by Dafeng Group for 120 million yuan. The reason why it is said that Intel has helped Dafeng Group is because Intel still expressed its willingness to acquire Havok in this life. It is just that Intel did not compete with Dafeng Group when it heard that Dafeng Group wanted to acquire it.

Of course, the reason why there is no dispute is that Intel did not see the value of this company too clearly. After Intel acquired Havok, because it did not make graphics cards, Intel hoped that the CPU would be responsible for the physical calculations, which led to the development of graphics card acceleration Havok FX being directly hidden. up.

A few months later, Nvidia acquired AGEIA, another company of the world's three major physics engines. Nvidia made great efforts to develop and opened the era of GPU physics acceleration.

Therefore, Meng Qian wants to snatch Havok from Intel. Fortunately, Intel has not seen how valuable Havok is. Part of the reason for his acquisition is even for patent fees, because the current ATI and Nvidia both use Havok solutions. , Microsoft's X BOX and Sony's PS also use the Havok solution.

But since the Dafeng Group wanted it, it didn't matter whether the little money was earned or not, so let's treat it as a favor, and just like that, the company was in the hands of the Dafeng Group.

After the acquisition, Meng Qian brought together the company's original graphics card department and ATI's integrated GPU subsidiary Galaxy Technology and the management of Havok. Everyone got to know each other first, and they will work together in the future.

Then the first thing announced was that Havok will become an independent subsidiary of Dafeng Group and will not merge with Galaxy Technology. The reason is also very realistic. IBM owns 6% of Galaxy Technology.

Then, everyone started to have a meeting. David, the current CEO of Galaxy Technology, first announced a good news. At the end of 2006, AMD acquired ATI. The listing brings a new series of product lines.

Among them, X1950XTX is the main high-end, and the application of GDDR4 memory greatly improves the on-chip transmission bandwidth and performance. RV570 is the main product in the mid-to-high-end market. RV570 has 36 PS units, 12TMU, and 12ROPs. It also has a 3:1 architecture, but its performance is very good, and it has become the focus of the mid-to-high-end market. In the low-end market, RV560 outperformed G73 and became one of the most cost-effective display chips in the low-end graphics card market in 2007.

Many people think that God does not give ATI a chance. If ATI resists for one more year, maybe it will survive. This helpless God’s will tricks people. In the past, AMD was cheaper, and Dafeng Group is naturally cheaper in this life.

With the help of this new series of product lines, Dafeng Group has been well received in the graphics card market after acquiring AIT, and its market share has directly increased by 10 percentage points. That's why ATI made the product well, and kept exaggerating the wind group.

After reporting the product sales and market conditions, David began to talk about the company's development, and at the request of Meng Qian, David first tested the people on Havok, "Everyone has a certain understanding of Nvidia's CUDA. understand?"

Everyone nodded. CUDA has been in development for a year. If you haven't learned about CUDA until now, you should really be fired.

"I don't know what Havok thinks about CUDA?"

All eyes at Havok are focused on one person, and that is their chief technology officer, Claude. Claude said very confidently, "In the past two years, everyone has been discussing the potential of the GPU. Many people When talking about this issue, you will think of Nvidia's CUDA, but at least from the current point of view, because the purpose of designing CUDA is at least to allow more developers to use their products better, but it is still aimed at games. and video production.

In fact, everyone began to pay attention to the potential of the GPU. In fact, it started this year. Therefore, I doubt whether the original intention of Nvidia’s CUDA design is related to the potential of the GPU. I have created a little inherent advantage. It is still hard to say whether Nvidia can make good use of this inherent advantage.

At least the development of CUDA this year is a bit slow. "

Meng Qian nodded silently, Claude saw the essence of the problem, and was willing to say these things during the first company meeting.

David glanced at Meng Qian's expression and asked, "If you were in charge of the CUDA project at Nvidia, what do you think you should do?"

"There is no doubt that one thing is to enhance the versatility of the GPU. Nvidia's CUDA does not help the GPU to truly release its potential. In other software applications other than games and video production, the GPU is still idle. Everyone should know that. Judging from the speed of using floating-point operations, the CPU is far lower than the GPU, and now it should be time to explore the general possibilities of the GPU."

Meng Qian looked at Claude more appreciatively, while David continued to ask, "Can you tell me about the specific measures?"

"To put it simply, what the GPU does in the game is to calculate the pixel color, so in theory, in any application direction, we can let the program prepare the corresponding array and determine the way the GPU calculates, so that To achieve the corresponding functions, there needs to be a function library to complete the docking work. This is a point that I think they overlooked after seeing NVIDIA's CUDA. CUDA should not be just a programming environment, but an algorithm platform. Including the environment, including libraries, including hardware.

If such a computing platform can be built, the value of the GPU will be greatly enhanced. "

"Then in what areas do you think GPUs will shine in the future?"

"For fields with higher requirements for parallel computing, GPUs should have their own uses." Claude thought for a while, but did not mention too specific directions.

However, Claude can have such an idea. Meng Qian already thinks that this person is very good. At this time, Meng Qian gave David a look, and David stopped asking, "I am very happy to hear Mr. Claude's point of view. , because Mr. Claude's point of view is in sync with ours."

When these words came out, Havok's high-level executives realized something. Sure enough, what David will introduce next is the GPU computing platform that Gale Group has been quietly developing for several years, the Galaxy computing platform, and the Galaxy computing platform. Huge function library, simple and convenient development environment, rich application testing and compatible design for various hardware.

For now, the Galaxy computing platform is more like a GPU instruction manual for developers. It teaches developers how to use GPUs to implement more applications than game and video production, and provides developers with a special Convenient operating platform.

During the internal presentation, David focused on showing everyone the excellent performance of GPU in the field of deep learning.

Meng Qian, as the pioneer of deep learning theory, combined with the results of the Galaxy platform for GPU deep learning, Claude, who is highly technically sensitive, has already smelled a hint of the future.

After introducing the Xinghe platform, David concluded, "The Xinghe platform, the accelerated physical development of the GPU, and the technology are the three major development directions of our company. Regarding the technology, our goal is to realize the 55nm process in 2008, and about The physical development of the Galaxy platform and the GPU is accelerating, and I hope that we can work together as soon as possible to create a complete ecosystem."

There was warm applause in the meeting room.

Looking at the current team configuration and everyone's mental state, Meng Qian felt quite at ease.

Thanks to Zvman for the reward!