Reborn in 2010, the AI ​​in my mind is too powerful.

Chen Yiran's fake news about acquiring AMD on April Fool's Day was purely a joke, and he didn't expect AMD to respond to him.

Moreover, AMD took advantage of the situation and released a lot of related news on April Fool's Day, as if the two companies had a close relationship. Then, they called to contact us the next day, Saturday.

The two sides already have a lot of cooperation. AMD itself is currently using Chen Yiran's EDA software, and engineers from both sides have been exchanging technical information.

But this time was different. Instead of following the normal procedure of contacting engineers or sales staff, they went directly to Gao Dewei and said they wanted to meet with Chen Yiran alone to talk.

Gao Dewei assumed that AMD wanted to gain some exclusive rights in EDA, even if it was just for an update of one or two weeks, it would be a huge advantage for the chip industry. He didn't take it too seriously and informed Chen Yiran to make the decision.

Chen Yiran immediately became interested. He had a feeling that AMD was up to something and was probably planning something big. He immediately agreed, and the two sides agreed to meet on April 5.

After hanging up Gao Dewei's call, Chen Yiran stared at the screen in front of her.

This is the working page for Deepflow and CUDA. The system they have built is already very perfect, but from all aspects, Chen Yiran has a reason to break this relationship.

He needs to make CUDA dependent on deepflow, not the other way around.

In his past life, he didn't understand these technologies, but in this life, whether it's Jasmine or his own experiences, he clearly knows that CUDA will be Nvidia's biggest competitive advantage in the future.

Previously, GPUs only did one thing: rendering graphics. Whether it was game graphics or 3D modeling, GPUs were professionals. However, if you wanted the GPU to perform other tasks, such as scientific computing or physics simulations, you had to disguise the problem as graphics rendering to trick the GPU, which was very troublesome.

But Nvidia created CUDA, finally allowing engineers to write programs directly in C and run them on the GPU. Without needing to disguise themselves as graphics tasks, the GPU became a general-purpose calculator.

Nvidia has put a lot of effort into improving the CUDA ecosystem, but with the emergence of the powerful Deepflow, they had to quickly seek cooperation.

In simple terms, DeepFlow is an abstraction layer that unifies the interface and innovatively adds a computing engine. Therefore, engineers now write code from the DeepFlow entry point, rather than directly supporting CUDA.

Ultimately, however, CUDA can only support NVIDIA graphics cards, so the relationship between the two is rather delicate; it's fair to say that both sides want to be the most important one.

Chen Yiran had a reason for agreeing to prioritize support for NVIDIA cards: AMD's CUDA-compatible solution, OpenCL, was simply not very effective.

This is a specification initiated and first submitted by Apple, which was later improved by AMD, IBM, Intel and NVIDIA into a general open API that can be used to develop general-purpose GPU computing software.

Without CUDA, this thing would still be okay, but with CUDA, nobody would want to use OpenCL. Nvidia is especially cunning; they are very enthusiastic about supporting this format, and then they tell everyone that the same thing will be 3% faster with CUDA.

This isn't really OpenCL's fault, since it's just a standard and each company still makes its own drivers. AMD's drivers are obviously very poor, making them incredibly difficult to use, riddled with bugs and extremely poor optimization.

Chen Yiran doesn't have any great ideas yet from this exchange with AMD, but at least he has a basic goal: to help AMD provide better support. However, AMD also needs to do its job well.

On Tuesday, April 5th, Chen Yiran received representatives from AMD in his company office. The visitor was Harry Waring, Global Vice President in charge of strategy.

In fact, AMD is currently in a very turbulent period, with its top management being quite unstable.

Former CEO Dirk Mayer resigned in January, and the current CEO, Thomas Seifert, is on a temporary basis and also serves as CFO. AMD can now be described as a massive, makeshift operation.

After the two sides shook hands, Harry Waring spoke very politely about the power of FlowEDA: "Mr. Chen, you are truly a genius. In less than a year, your company has already become the dominant force in EDA. You have also made Kaiden's software more and more user-friendly, and the feedback from our engineers has been very positive."

Chen Yiran found this kind of mutual flattery as a starting point in business negotiations annoying, but he couldn't escape it either, so he could only chime in: "Not at all, it's just that we're good at communicating with our clients. AMD's engineers are excellent, and our communication efficiency is very high, which makes the work so easy."

Then came a series of mutual flattery, with virtually no useful information exchanged in the ten minutes.

Don't think of this as a sign of inefficiency. In fact, there are established patterns in business negotiations. Managers who have received formal business education are more likely to adopt this strategy, and there must be a reason for it.

For someone like Chen Yiran who doesn't have a business background, this kind of probing would be very uncomfortable, but who knows, maybe that's Harry's intention.

In a past life, there was a widely circulated video of Shou Zi Chew, CEO of TikTok, being repeatedly questioned eight times by a well-known anti-China senator about whether he had any strange connections with China. Chew repeatedly and helplessly replied that he was definitely Singaporean, and even had served in the military.

Everyone thought that the congressman was out of his mind and had gone mad, but in fact, repeatedly asking questions about the same issue in a roundabout way is an important negotiation technique, and even an important interrogation technique.

Chen Yiran knew this and felt that it was better to just get things done quickly. They weren't really in competition, and their relationship was actually quite good. Going through all the formalities was just too disgusting.

He interrupted Harry's rambling and asked directly, "Since you've come in person, I believe it's not just an EDA issue. Even if it requires special optimization, it shouldn't be you who needs to come forward, nor should you directly ask me to come forward."

Harry smiled and said, "That's right, my main purpose in coming here today isn't related to FlowEDA's business. We came specifically to see you, even though you've delegated the related technical updates to FlowEDA. But actually, DeepFlow is purely your personal business, and we hope to collaborate on that framework."

"Are you trying to make me stop prioritizing CUDA support? To be honest, it doesn't make much difference to me which hardware vendor I prioritize; the main issue is still technical."

Your technology is really lacking; your OpenCL support is a nightmare. I can agree to treat you and Nvidia the same, but you need to optimize your drivers first!

Harry was very happy. He was actually quite unsure of himself before he came, mainly because Chen Yiran was Chinese, not for business reasons.

It's not about racial discrimination, but because Nvidia's Leather Jacket Huang is also Chinese. There are many reports from Nvidia saying that Chen Yiran has a good relationship with Leather Jacket Huang's family and is also a friend of Little Leather Jacket.

This is a nightmare for business negotiations. You talk to him about interests, and he talks to you about feelings. How can you even negotiate?

However, judging from Chen Yiran's attitude now, it doesn't seem to be what Nvidia said.

Harry nodded, then said somewhat seriously, "Indeed, our OpenCL support is very poor, and the GPU driver team's budget is also insufficient, leading to a high turnover rate of OpenCL engineers. So we'd like to ask, if we agree to give you some shares, could you contribute your technical expertise as equity and help us resolve this issue?"