The higher-end the chip, the more difficult it is to develop.

But the low-end chips used in thousand-yuan machines are much easier.

Wang Yi continued: "There are also baseband chips. The same is true. There are two versions. A 28-nanometer Wing Loong 3000 and a 40-nanometer Wing Loong 2800. The 40-nanometer Wing Loong 2800 will be taped out in the second half of the year for use in thousand-yuan machines.

.28nm Wing Loong 3000 for next year’s flagship aircraft!”

"no problem!"

Williams had no comment.

Try a simple one first to gain experience in advance, and then do a difficult one, and your success rate will naturally increase greatly.

What's more, after the 40nm trial production in the second half of the year, it will be able to be taped out in its own wafer factory, which will be even simpler.

In fact, many of Qualcomm's mid-to-low-end processors are castrated or modified from older processors, and it is impossible to develop them all from scratch.

The biggest benefit of doing this is that research and development costs are greatly reduced.

A 7-nanometer mobile phone chip, not counting other expenses, the design cost alone is 2 billion yuan!

The design cost of a 3-nanometer mobile phone chip is as high as 4 billion yuan!

It is precisely because of such high costs that there are very few companies that can develop high-end mobile phone chips.

Even if the design cost of a 28-nanometer mobile phone chip is lower, it will still cost 300 million yuan.

As for the 40-nanometer mobile phone chip, the design cost is even lower, costing 200 million yuan.

Xingyi Semiconductor launches so many mobile phone chips a year, including Kunpeng 500, 700, and 900, baseband Wing Loong 2800, and Wing Loong 3000, a total of five models!

If it were all developed from scratch, not to mention it would be time-consuming and labor-intensive. The R&D cost alone would exceed 1 billion. The subsequent tape-out costs, production line costs, etc. are all significant.

If the Kunpeng 700 is castrated, modified, and upgraded to the Kunpeng 500 or 900, the research and development costs will be much smaller.

Coupled with the cost of baseband research and development, it is estimated that 500 million can be completed, saving at least half.

Like the bomber, instead of researching many new models and failing to master each model, why not crazily modify the Sixth Master!

Ten years later, Mr. Liu will have almost no use for his 26 letters, and no one knows how powerful he is.

Wang Yi’s strategy is magic reform!

In this way, the Kunpeng 500 chip + Wing Loong 2800 baseband will be in production by the end of the year and will be used for thousand-dollar machines next year.

In this way, the launch time of the thousand-yuan phone cannot be postponed to the middle of next year. On the contrary, it must be released early next year!

Therefore, the research and development of thousand-yuan machines must be accelerated across the board.

Wang Yi had some calculations in his mind.

In 2012, the flagship Xphone 1 Pro used the 40nm quad-core processor Nvidia Tegra 3.

The 2013 Qianyuan phone uses the 40nm quad-core processor Kunpeng 500, and there are no problems at all.

Even the quad-core low-end processors from Qualcomm and MediaTek were only launched at the end of 2013, and mobile phones were not launched until 2014.

And the cost of self-research and production is even lower.

The total R&D cost of five chips is 500 million, and the R&D cost of one chip is evenly divided into 100 million.

Assuming mass production of 100 million chips, the diluted R&D cost of each chip is only 1 yuan.

As for the material cost, it depends on the wafer price.

A 28-nanometer wafer costs TSMC $3,000, which is 18,900 yuan based on the average exchange rate in 2012.

It can cut out about 700 chips, and the yield rate is almost 550. The wafer cost of a single chip is about 3 yuan.

Coupled with packaging costs, evenly distributed R&D expenses, marketing expenses, etc., the cost of a single chip is over 70.

It is precisely because of this that the costs of Qualcomm 800 and Apple A5 are around 70.

As for 40-nanometer wafers, TSMC sells them for US$2,000. Based on the same calculation of 550 wafers, the cost of a single wafer is about 23 yuan.

Fortunately, the development cost of low-end chips is low. Adding other costs such as packaging, the cost of a single 40-nanometer mobile phone chip is about 40 yuan.

However, if the Xingyi wafer factory produces it in-house, the cost will be lower.

For example, a 40-nanometer wafer costs US$2,000 per wafer if manufactured by TSMC. However, if it is produced by Xingyi Wafer Factory in-house, it is estimated to cost around US$1,500. After all, TSMC has to make money as an OEM!

1,500 US dollars, which is 9,450 RMB. If 550 chips are cut out, the cost of a single chip is 17 yuan.

Plus R&D costs, packaging costs, etc., it's about 30.

It is much cheaper than buying Qualcomm chips, and a quarter cheaper than TSMC OEM.

But unfortunately, this is just the processor, and the external baseband chip must be included.

The cost of the baseband chip is also nearly 30 yuan.

In this way, a Kunpeng costs 500 + a pterosaur costs 2800, which costs 60 yuan per set.

The cost has doubled.

No way, this is the difference between external baseband and integrated baseband.

System-on-chip (SOC) with integrated baseband, one chip includes CPU, GPU, baseband, DSP, IPS, GPS, Bluetooth, Wifi...

The technical difficulty is much higher than that of plug-in baseband, but the cost is much lower!

Wang Yi sighed and looked at Williams:

"William, if you make a 40nm SOC with an integrated baseband chip, can you be sure? It will be mass-produced before New Year's Day."

"Mass production before New Year's Day!" William frowned and thought for a while: "Chairman, we started developing the 28-nanometer integrated baseband Kunpeng 900 last year, but we couldn't make it before New Year's Day. The fastest would be the middle of next year.

"

"However, the low-end SOC with 40nm integrated baseband is much simpler. If we go all out, there is great hope before New Year's Day!"

"But in this case, the 28nm high-end chip Kunpeng 700 will not be available until after New Year's Day!"

Wang Yi nodded: "That's no problem. The second generation of Xingyi will only be released in September this year. The next generation of flagship mobile phones will be released at least half a year, so there is no rush until March next year!"

"Okay, that's no problem." Williams responded: "Then we will first make the 40-nanometer SOC Kunpeng 500 with integrated baseband and mass-produce it before New Year's Day. As for the high-end 28-nanometer Kunpeng 700 with external baseband, it will be launched next year

Mass production in February. Kunpeng 900, a 28nm high-end chip with integrated baseband, will be mass produced by June next year!”

"Okay!" Wang Yi nodded.

The release time of high-end phones must be staggered, and one model every six months will be enough.

It's a thousand-yuan phone, and it can be released in January.

The time to market can be slowed down to February.

A month's time for public opinion to ferment can actually cause a greater sensation.

After all, smartphones priced at 1,000 yuan were absolutely king in early 2013.

This 40nm SOC Kunpeng 500 with integrated baseband saves a baseband chip, and the cost of a single chip can be controlled to 30 yuan!

The cost of the thousand-yuan machine is completely reduced to 550 yuan, and the profit is very considerable.

"Looking forward to your results!" Wang Yixiao said: "But the performance cannot be too bad. At least it must have the performance of NVIDIA's quad-core chip Tegra 3!"

"This is no problem." Williams said with a smile: "Our chips have the advantage of starting late. NVIDIA and Qualcomm's chips are all ARM's V7 architecture, and we directly adopt the latest ARM V8 architecture. Architecturally,

One generation ahead of them, even if it is a 40nm low-end processor, its performance will not be inferior to that of Tegra 3."

"That's good." Wang Yi couldn't help but sigh with emotion.

Chips these days are produced one generation per year and are developing rapidly.

In the first half of 2012, the 40nm Tegra 3 was still the most powerful quad-core chip in the world.

But in the second half of the year, the 28nm Qualcomm quad-core 8064 is the most powerful, a generation ahead of the technology.

By the beginning of 2013, the performance of the Kunpeng 500 chip in Xingyi Technology's thousand-yuan phone was not much worse than that of Nvidia's flagship Tegra 3 a year ago.

There is nothing we can do about it, chip upgrades are so fast.

For example, the dual-core CPU Qualcomm 8X60 of the

After all, the 8X60 dual-core processor is at the level of Qualcomm Snapdragon 200, which is considered low-end among low-end.

Even Qualcomm's flagship processor APQ8064 in the second half of this year is only at the level of Snapdragon 600, and will become a mid-range processor next year.

Mobile phone processors are updated too quickly and are downgraded by one level every year. There is nothing we can do about it.

The Kunpeng 500 quad-core processor will be launched next year. Even if it is not as good as the 28nm quad-core processor APQ8064, it will still have to compete with the quad-core Tegra 3!

Otherwise it is simply untenable.

The Xingyi wafer fab will be fully operational next year, with an annual production capacity of 300,000 wafers. Taking into account the yield rate, it can manufacture 6.5 billion Kunpeng 500 SOCs!

These 6.5 billion self-developed and self-produced chips are not only enough for a thousand-yuan phone, but there are also 65 million more chips that can be used for xphone 1pro!

The second generation of Xingyi will be released in the second half of the year, and the dual-core Xphone 1 will be completely discontinued next year.

The quad-core Xphone 1pro will continue to be sold, but the processor will be replaced by the self-developed and self-produced Kunpeng 500 from Nvidia Tegra 3!

After all, Kunpeng 500 and Tegra 3 are both 40nm quad-core processors with similar performance, but the cost is much lower.
To be continued...