The world's richest man started by developing gutter oil

With just a few promotional videos, The Legend of Zu has received positive reviews from 90% of netizens.

It can be seen how good the Zu Mountain Legend filmed by Xu Ke is this time. Of course, if you want to film it well, it must be based on the premise of having money. If you don’t have money, it is impossible to make every detail so perfect.

As for how much box office revenue "Legend of Zu" can earn, we will have to wait until May 5th to find out.

Because the movie tickets for "Legend of Zu" will be pre-sold on May 5th, the more tickets pre-sold, the higher the screening rate will be, and the higher the screening rate, the higher the box office will be.

However, one thing is certain, the box office of "The Legend of Zu" is definitely over 50 billion, and the global total is likely to be billion.

Half past two in the afternoon.

Han Chen once again received an email from Guo Kezun.

The content of the email includes the 14nm Butian CPU, the workstation version of Butianshi CPU, the third-generation Whale Swallow CPU, the fourth-generation Deep Blue CPU, the independent AI accelerator card, the independent motherboard, and the third-generation Kongxuan graphics card.

To be precise, these chips have been mass-produced and have been produced.

Since this generation of Butian CPU technology has made great progress and can accommodate twice as many transistors in the same area, the design of Butian Stone 2014 is much more complicated than the previous generation.

顶配版补天石2013采用1+3的核心的设计，顶配版补天石2014采用4+2+2的核心的设计，4代表着四颗超级效能大核心，这4颗超级效能大核心不支持超线程功能，第一个2代表着2颗中等核心，这2颗中等核心支持超线程功能，最后2颗核心是低效能的小核心。

The 4 super-performance cores reach 4.5Ghz, and the performance in turbo mode reaches 5.2hz. The frequency of the four medium cores is 3.2Ghz, and the frequency of the 2 low-performance small cores is 2.4Ghz. The medium cores and small cores do not support turbo mode.

6颗核心的二级缓存总和是3072KB，三级缓存加起来是76MB，4级换成升级到夸张的512MB。

The difference between the new generation of Butian Stone and the old generation is that the GPU is no longer integrated. The new generation of Butian Stone no longer uses the SOC design method, but uses the Chiplet core packaging process independently developed by Deep Blue.

The new generation of top-end version of Bu Tian Stone has a 237% performance improvement compared to the previous generation of top-end version of Bu Tian Stone.

The mid-range version of the Bu Tian Stone has one less super performance core and one more medium core than the high-range version of the Bu Tian Stone.

The low-end version of the Bu Tian Stone has 2 fewer super performance cores and 2 more medium cores than the high-end version of the Bu Tian Stone.

In addition, the new version of the Bu Tian Stone will no longer integrate running memory and storage memory. Running memory and storage memory will become optional accessories. The motherboard is equipped with 12 hardware-dedicated surfboard interfaces.

Users can use these interfaces to add accessories to their computers to enhance their computer performance.

Under the leadership of Su Ma, the new generation of Deep Blue computers is very different from the previous generation. The new generation of Deep Blue computers has cancelled the all-in-one design because the research and development cost and manufacturing cost of the all-in-one computer are too high. In addition, the all-in-one design has greatly reduced customer choices.

In order to cater to market demand and gain more market share, Su Ma directly cancelled the all-in-one machine design. In fact, there is another reason why Su Ma did this.

The reason why Su Ma did this is because the power consumption of the new generation of CPUs and GPUs has increased significantly due to frequency reasons.

All-in-one computers cannot dissipate the heat of these chips very well, and only an independent host can better solve the heat dissipation problem. Therefore, the new generation of Deep Blue computers will enter the user DIY era.

The top-end version of Butianshi 2014 is priced at 2099, the mid-end version is priced at 1599, and the low-end version is priced at 1099.

As for the motherboard, Su Ma has also designed five models. Because the Butianshi does not have a south bridge chip, the prices of the three motherboards are not very high. The most expensive motherboard is priced at 1599, and the cheapest one is only 699.

The biggest difference between high-end motherboards and low-end motherboards is that high-end motherboards support overclocking function. Of course, the materials used in high-end motherboards and low-end motherboards are also different.

After reading Su Ma's report, Han Chen didn't know whether it was right or wrong to let Su Ma support Deep Blue Semiconductor.

"This Su Ma really knows how to make money." Han Chen couldn't help but complain.

No wonder AMD not only turned losses into profits but even surpassed Intel under Su's leadership.

However, since Deep Blue Semiconductor has been handed over to Su Ma, Han Chen will not participate too much in her management, as long as her decisions can make money for the company.

The third generation of Kongxuan graphics cards is divided into five models.

The performance of the lowest-end model is the same as the previous generation of Bu Tianshi integrated graphics card. The only difference is that the video memory is much larger than the previous generation, upgraded from 4GB to 8GB. Su Ma set the price of this graphics card at 599 RMB.

The performance of the highest-end Kong Xuan graphics card is 2 times that of the Kong Xuan F2.8 graphics card. The highest-end Kong Xuan independent graphics card is named Kong Xuan F3ultra. The core frequency of Kong Xuan F3ultra is 1.9Ghz, it has 4763 stream processors, and is equipped with 24GB of independent video memory. Su Ma set its price at 4999 RMB.

The remaining three graphics cards are priced at 1299, 2999, and 3999 respectively.

最低端的显卡明明为孔宣F3se、另外三款显卡命名为孔宣F3se、孔宣F3s、孔宣F3、孔宣F3pro。

The difference between these three graphics cards and Kongxuan F3ultra is that the stream processors and main frequency show a step-by-step reduction.

Due to the reduction in stream processing, the performance also showed a step-by-step decline.

孔宣F3pro性能相当于孔宣F孔宣F3ultra的80%，孔宣F3pro相当于孔宣F3的80%，孔宣F3s相当于孔宣F3的80%。

The workstation version of the Bu Tian Shi CPU is also divided into three versions.

The top-end workstation version of the Butianshi has eliminated all medium cores and small cores, because this chip was originally designed for workstations, and workstations do not need to consider power consumption issues at all. The top-end workstation version of the Butianshi CPU uses 8 ultra-high-performance cores, and the power consumption reaches a terrifying 185 watts.

You should know that the CPU power consumption of the top-end version of the Bu Tian Stone is only 95 watts, and the workstation version of the Bu Tian Stone is 90 watts higher than the ordinary Bu Tian Stone.

In terms of motherboards, Su Ma has designed three motherboards for workstations. The first motherboard supports 2 CPUs running at the same time, the second motherboard supports 3 CPUs running at the same time, and the third motherboard supports 4 CPUs running at the same time.

A single CPU can support up to 256GB of memory, which means that a 4-way CPU can support 1TB of memory. Of course, 99% of people will not need so much memory.

In addition, the three workstation motherboards also support multi-graphics card functions. The top-end motherboard supports up to 8 graphics cards operating at the same time, and the low-end motherboard supports up to 2 graphics cards operating at the same time.

The difference between the three workstation Butianshi CPUs is the number of their super performance cores. The top-end workstation Butianshi CPU has 8 super performance cores, the mid-end workstation Butianshi CPU has 6 super performance cores, and the low-end workstation Butianshi CPU has 4 super performance cores. The super cores are replaced by medium cores that support hyper-threading.

三颗CPU的价格分别是2999、3999和4999，三块主板的价格分别是999、1599和2599。

In order to improve the cost-effectiveness, the workstation Bu Tianshi CPU can be installed on an ordinary motherboard.

That is to say, the 7999 Bu Tianshi CPU can be installed on a 699 ordinary motherboard. There is absolutely no problem with a small horse pulling a big cart.

The third-generation Whale CPU is also divided into three versions.

The low-end version of the Whale CPU has 64 cores, and the frequency has been increased from 3.0Ghz to 3.5Ghz.

The mid-range version of the Whale CPU has 86 cores, and the frequency has been increased from 3.0Ghz to 4.0Ghz.

The high-end version of the Whale CPU has 108 cores, and the frequency has been increased from 3.0Ghz to 4.5Ghz.

Because the Whale CPU is a customized server CPU, the specific price needs to be determined based on the specific configuration, so I will not introduce it in detail here.

As for the AI ​​training card, it has been mentioned before.

It uses NVIDIA, which is actually a special graphics card. Currently, the R&D department has developed a 14-nanometer AI training card.

The AI ​​training card is composed of a third-generation 32-core CPU and a module card with 6698 stream processors, 16GB of RAM and 40GB of video memory.

The computing power of a single card is only 35.9TFLOPS, which is much less than Han Chen's estimate. Han Chen estimated the computing power to be 60TFLOPS.

There is no way that this is the first time that the R&D team has developed an AI training card, so it is normal that the performance of the developed card is a little worse. I believe that the next generation of training cards will be greatly improved.

However, it is already very impressive that the first AI training card can achieve this computing power. You should know that Nvidia's 3090 is only 35.6 TFLOPS, which means that the performance of Deep Blue's AI training card can be comparable to that of 3090.

You should know that the 3090 uses a 7-nanometer process, and the Deep Blue training card uses a 14-nanometer process, and the process difference between the two is 2 generations.

Being able to beat 2 by 3090 levels is already very awesome.

After reading Su Ma's email, Han Chen immediately sent a message to Andrew Ng asking him to contact Deep Blue Semiconductor immediately.

Replace the existing AI training servers with the latest servers as soon as possible. After replacing them with new servers, the training speed will be increased by at least 20 times. He will also be able to complete the training of saro and other AIs as soon as possible. As for the replaced servers, they will all be sent to DeepBlue Cloud as cloud servers for rent.

As for Shenlan's new products, Han Chen plans to wait until Singles' Day this year to launch them on the market.

After all, the old generation of products was only launched on Singles’ Day last year. It would be strange if consumers didn’t criticize the company for launching new products so soon.

September 2014, 5, China time.

Huawei held a press conference for a new mobile phone in Shanghai. The press conference was held very suddenly. Only the invited media knew about it. Netizens were unaware of it in order to increase the influence of the press conference.

Warwick also cooperated with Douyu and DeepBlue Film and Television to live broadcast today's press conference.

"Hello everyone, I'm very happy that you can come to the Huawei Mate 7 mobile phone launch conference. The Mate 7 will be a breakthrough for Huawei mobile phones." Yu Dazui started talking as soon as he came on stage.

When Yu Dazui introduced the CPU, the audience exclaimed in amazement.

"The CPU used in Huawei Mate 7 is the Kirin 910 independently developed by Huawei. This CPU is also the world's first 14nm mass-produced CPU. Here I would like to thank Deep Blue Wafer Factory for its support. Deep Blue Wafer Factory put our production tasks ahead of theirs and produced Kirin 910 within one and a half months. It is for this reason that Kirin 910 can become the world's first 14nm CPU." Yu Dazui bowed slightly to express his gratitude.

The barrage in the live broadcast room also popped up crazily at this time.

"Fuck, I didn't expect Deep Blue to prioritize producing chips for others this time."

"Those sheepdogs who said on the Internet that Deep Blue couldn't possibly produce 14 nanometers, come out now. They kept talking about how difficult it is to manufacture lithography machines and how Deep Blue couldn't possibly have a 14 nanometer production process. Now you've been slapped in the face, right?"

"Fuck! It takes other wafer factories at least three months to mass produce a chip, but Deep Blue has achieved mass production of a chip in just one and a half months. Do you know what this means? It means that Deep Blue's production technology is more than an era ahead of the world.

Although I don’t know how Deep Blue was able to mass-produce a CPU in such a short time, Deep Blue’s technology is truly amazing.”

"I know why Deep Blue was able to mass-produce a CPU in such a short time."

"why?"

"Because Deep Blue Fab does not need to make photomasks for lithography machines, making photomasks is very time-consuming. Deep Blue's technology directly skips this step. Deep Blue's technology has far surpassed our understanding of lithography machines. Deep Blue's technology may be several generations ahead of existing technologies."

"How come the big brother upstairs knows so much?"

"My job is to develop photolithography machines. The news about the photomask came from insiders. Deep Blue's technology is so far ahead that the boss is considering stopping the development of photolithography machines."

麒麟925采用的是现在最主流的大小核心设计，8颗核心里面有4颗是超大核心，剩下的是2颗中等核心和2颗小核心。

大核心的频率为3.2G频率，1级缓存为56K，2级缓存为128K，2颗中核心的频率为2.4Gz，1级缓存32K，2级缓存也是32K，2颗小核心的频率为2.0Gz，1级缓存32K，2级缓存也是32K，8颗核心共享的3级缓存8MB缓存。

Kirin 925 is nearly 3% faster than Deep Blue 30 in terms of CPU performance.

The GPU used is their self-developed GPU, but the architecture is the Kongxuan architecture authorized by Deep Blue, and its performance is 100% faster than Kongxuan MB40.

In general, Mate 7 is much faster than Deep Blue 3 in terms of CPU and GPU.

Of course, they are able to develop a CPU with more powerful performance than Deep Blue 3 because it is built on a 14nm process.

Under the same technology, their R&D capabilities are not as good as Deep Blue's. Even Qualcomm, Apple and even ARM cannot compare with Deep Blue.

Deep Blue's next generation mobile phone CPU will definitely be much more powerful than theirs.

"We have equipped the Mate 7 with 16GB of Qiongqi memory and 256GB of Taoji memory. I believe this is the first time that everyone has heard of this name, but you should be familiar with this memory. This memory is the memory developed by Deep Blue Semiconductor." Yu Dazui began to introduce the memory parameters.

The audience is not unfamiliar with Deep Blue Foundry, so Yu Dazui did not give much introduction. After all, Deep Blue has been using this kind of memory for a long time.

It was advanced for him, but it had been used for who knows how long for Deep Blue. (End of this chapter)