Xueba's computing system

Chapter 67: My mentor is not serious?
Three days later, there are five days left before school starts.

To Lin Yuan's surprise, the two H100s provided by OCR were actually high-end models, not entry-level models.

Because the first draft of Lin Yuan's algorithm was completed on the third day of training.

For AI training, as long as the model is well made and the computing power is sufficient, all you need to do is tie a dog next to it.

When Lin Yuan used the system to solve the model building problem and got computing power equipment for free from OCR, everything was already done. Just like when you hold a gun in your hand, you don't have to find someone to shoot it to prove its power.

Everything just happened naturally.

Lin Yuan originally intended to submit the first draft of the algorithm, together with the simulation test results of the algorithm, to the head of the technical department of the Haotuan Zijin Branch.

Before submitting, Lin Yuan first sent the email of submitting the algorithm to his tutor, thanks to Professor Zhang's special reminder. It was the tutor who forwarded it to the team personally.

There are two things in the email: one is a PDF document that summarizes the AI network model used, records the key nodes in the training process, and the evaluation results. The other is the core Python code of the AI network model.

The data used for specific training and the algorithm entities trained by the AI model are too large to be included in the attachments, as they are hundreds of GB in size (mainly because the training sample data is too large).

However, a qualified practitioner only needs to look at the data on the PDF document to know the value of Lin Yuan's algorithm. As for the hundreds of GB of data, it is used to prove that everything written in the PDF document is true. This can naturally be done later.

As for why he had to submit his work through his tutor, Lin Yuan had not thought of this at first. Experience is always the key, and the tutor must have seen the infighting in the workplace and was afraid that Lin Yuan would be taken advantage of.

When the mentor takes action, people naturally dare not act rashly.
-
Lin Yuan was very confident in the system. He was sure that the algorithm he submitted would attract the attention of the good group. Now he just had to wait.

From now on, you can say bye bye to OCR!
I deleted the WeChat account of that application specialist and pretended that nothing ever happened.

However, I don't know what the instructor was thinking. He called Lin Yuan on the day of algorithm submission. After the two of them talked about the most important algorithm submission, the instructor suddenly mentioned the issue of OCR computing equipment.

"Lin Yuan, what are you going to do with the computing equipment at OCR?"

"Ah? I think I should cut ties with that place from now on."

"Cut?"

Hearing the instructor's doubtful tone, Lin Yuan said tentatively, "I don't have any evidence to accuse OCR of this organization, otherwise I would definitely report it."

In fact, even if Lin Yuan had evidence, he was unwilling to get involved in such muddy water.

"What are you reporting?"

"So what the teacher means is... if reporting it fails, I will just destroy it? Give it a virus to paralyze it?"

"Think more divergently."

Lin Yuan couldn't figure out how to react for a moment.

"Think about it, teacher, what did I do when I saw those caviar and black truffles at the tea break?"

"I wasn't even born at that time, so how could I know what you did?"

"As you young people would say, can't you use your brain to make up your mind?"

"eat?"

Professor Zhang answered himself, "Eat and take at the same time. If it's hard to take away, just lick it."

"Uh..." Lin Yuan was a little... 'Is this something an intellectual would do?'

"So, you understand now." "I, what do I understand? Should I learn from you to eat and take? Should I lick it again?"

"Yes."

"But how can I get it? How can I use it for mining?"

Lin Yuan suddenly woke up from his dream, "Yes, why don't I use it for mining?"

"exactly!"

After a brief moment of excitement, Lin Yuan immediately realized a problem. "Two H100s are powerful, but at most they can only be used for a few days. It's not worth much money."

"Ahem~~" There was a cough from the other end of the phone, and then Professor Zhang continued, "Well, if you need to pay for the electricity yourself, it really isn't much money."

"I understand. Someone is paying for my electricity bill." Lin Yuan loved his unruly mentor so much. Professor Zhang really taught him by example and never hid anything. Lin Yuan was so moved that his eyes were filled with tears. The mentor must have known that he was too poor and cared about his own self-esteem, so he reminded him of this way to make money.

"Just focus on mining. As for the algorithm, as long as your algorithm is correct, you will get everything you deserve. As for the group, I'll take care of it for you."

The most expensive part of mining is not the equipment, because no matter how expensive the equipment is, it is a one-time investment, and the heavy asset itself is valuable. The most expensive part is actually the electricity bill.

Lin Yuan roughly calculated that with the computing power of two high-end H100s, he could probably make about a thousand dollars a day.

He checked the price of the high-end version of H100, which is about 30 yuan. The two together cost 60 yuan. If you don't consider the electricity cost, it should be okay to get a thousand yuan. Otherwise, what fool would be willing to take the risk to mine.
-
Thus, Lin Yuan was finally able to start his mining plan.

After a failed attempt to mine with the system, Lin Yuan finally found another way with the help of his mentor.

"Xia Ye, we have the money for our baby's milk powder."
-
Lin Yuan didn't expect that after submitting the first draft of the algorithm, what he would do next was to study how to use OCR's computing power equipment for mining.

But even though we are happy, problems also arise. Because mining is not something you can do whenever you want.

Mining is all about computing power, but computing is not something that can exist alone. The CUDA computing cores that realize computing power in GPUs are like processing plants that need something to go in before they can produce. And the thing that goes in is data.

The production capacity of a processing plant depends not only on the operating speed of the machines in the processing plant, but also on the speed of transportation - how fast the raw materials are transported out.

Therefore, in addition to the main frequency, another important performance indicator of the GPU is the bandwidth of the internal memory bus.

This bandwidth indicates how fast things can be delivered in the factory inside the GPU.

In order to ensure that the bandwidth is large enough, there are certain requirements for the GPU's own video memory. Video memory represents the total amount of data you can cache, which can be simply understood as the capacity of a warehouse in a factory.

It is for this reason that mining has minimum requirements for GPU video memory.

Because the data used for mining is the "ledger", the ledger must be placed in the video memory.

Lin Yuan was too lazy to care about the data structure of the ledger and the principle behind it. What he cared about was that once the ledger was loaded into the GPU's video memory, it would be easy for the OCR to scan out the data features. Then they would know that he was mining.

The ledger of each digital currency has its own data characteristics, because all data in the world has its characteristics, and data without characteristics is itself a bunch of gibberish.

Various security software can tell what you are mining based on data characteristics alone.

Therefore, the problem that Lin Yuan is facing now is: how to mine under the nose of OCR.

Even if you can make 1,000 yuan a day, that would be 7,000 yuan in seven days. Don't miss out on this money.

(End of this chapter)