Chapter 295 - 288 Technical Route_1

Jingou Xi Qing Technology Development Zone.

This is one of SinoCore International’s contract factories and is the first to have completed the upgrade and modification of a high-yield 28nm production line.

Wang Minjiang arrived at the final product storage area led by the branch’s general manager and chief engineer, and entered a spacious room filled with the distinctive smell of PCB boards.

Several dark racks stood in the warehouse, and a worker took a tray down from one of the little compartments and handed it to Wang Minjiang.

"Mr. Wang, this is your Y1611A official release, the ’Taowu A1000’ chip. It’s been less than 20 hours since they were packaged and placed here, please take a look," the worker said.

Wang Minjiang pried out a processor roughly 6cm square and examined it carefully under the light.

The top side of the chip was completely covered with a nickel-plated copper heat spreader, and the bottom had 3900 pure gold contacts, also known as gold fingers.

The first batch of chips from Yellow River Semiconductor was still based on X32115, deeply optimized for desktop improvement. They were designed with the new "Taishan" architecture, and internally codenamed Y1611A, Y1611B, and Y1611C series.

Y stands for the first letter of Yellow River in English; 16 represents the year; the third digit is sorted according to that year’s series name, and the fourth digit is the version iteration order.

For example, the Y1601 that was given to The Great Wall computers for show—the 0 signifies the "Qiongqi" series code.

A, B, and C represent performance levels, with A being the consumer market workstation level, B is for enthusiast level, and C is for economic and practical level. In the future, other letters might be introduced, and the following numbers distinguish the specific CPU levels.

The Y1611A is a flagship-level CPU intended for the desktop market, but after its release, it will be referred to as TPU, with T standing for the initial of "ternary."

Because it is quite different from the traditional architecture in both the ternary structure and functionality division.

Originally, a computer mainly consisted of a processor, memory, and hard disk to operate, with the CPU providing direct video output. Later, to enhance graphics computing power, separate graphics cards—GPUs were introduced.

The ternary architecture took on both roles directly, providing considerable computing power for a variety of scenarios with the same architecture and design philosophy.

Compared to the mainstream CPUs on the market, the core area of the Taowu A1000 is nearly 120% larger, with a substantial rated power consumption of 350W. While this is indeed high, it actually saves electricity compared to a typical computer equipped with a separate GPU.

The Taowu A1000 has 6 cores and 18 threads, rivaling Intel’s flagship i7 6950X, and possesses strong graphics computing power, nearing the GTX 1080ti level in pure graphics computation.

To achieve ultimate performance in both areas, the Taowu A1000 is equipped with an automatic overclocking program that can release about 600W of power, reaching up to 800W under good cooling conditions.

It’s demanding but definitely worth it, and it even saves power compared to traditional systems when paired with the XOS and professional software. Not to mention the AI hardware support that mainstream manufacturers have not yet explored—this is a major bonus.

The main purpose of the Taowu A1000 is to show off in the consumer end and to sell to enthusiastic users and workstation crowds, though the latter is incidental.

The domestic ternary ecosystem has just begun to cultivate, and Huawei will soon release the world’s first ternary smartphone mate10, while Yellow River Semiconductor will officially promote the XOS system’s program editor to domestic enterprises.

The Commission of Science, Technology, and Industry for National Defense has organized the full translation of basic office software, as well as social and lifestyle apps. However, the translation of the vast entertainment software, especially games, is challenging. The commission plans to spend a year vigorously pushing the domestic software industry into ternary development.

To this end, the authorities are even determined to enact related national software legislation next year to strongly support ternary development while fending off foreign pressure.

This isn’t just about Yellow River Semiconductor; it’s a task for the entire national semiconductor hardware and software industry chain.

Therefore, Wang Minjiang had no worries about the hardware. Xinyuan Company had already produced the X32045, which is even more advanced than the system X32035, and the X32055 is in design. Each generation is more powerful than the last, and Yellow River Semiconductor just needs to squeeze out a little at a time; the X32035 alone would suffice for a decade.

After placing the TPU back, Wang Minjiang said:

"We’re not worried about the hardware, can the cost of solid-state drives be brought down?"

"This..."

The chief engineer of SinoCore felt a bit embarrassed but still told the truth.

"Binary NAND flash memory costs 2.8 yuan per GB, ternary flash memory costs 8.9 yuan per GB, we don’t see a way to reduce it at the moment."

Wang Minjiang: ...

One of the existing weaknesses of ternary computers: storage.

It was the end of 2016, and the mainstream storage mediums on the market were still mechanical hard drives, with solid-state drives being expensive and offering less capacity.

Whether it was traditional mechanical hard drives or solid-state drives, ternary computers needed to go through translation for reading, which greatly increased latency and reduced read/write speeds.

For NAND flash-based solid-state drives, the problem wasn’t severe since they were inherently fast. Despite a bit of performance loss after conversion, they were still quick, and the loss was acceptable. However, mechanical hard drives were a real headache, with sky-high latencies and already low read/write speeds becoming even worse.

Dedicated ternary particle solid-state drives did match computers with their quick response, fast read/write speeds, and long lifespan, but naturally, the price was even more extravagant.

In the era of vehicle-mounted computers, it might be acceptable to use ternary storage or solid-state drives without worrying about the cost of a few hundred or thousand yuan, but it wasn’t feasible for the consumer market.

Mainstream computers came with storage of more than 500 GB; that means spending over a thousand yuan on hard drives? Most people probably couldn’t accept that, yet the experience with mechanical hard drives was indeed quite frustrating.

Unless it was seven or eight years later, with advances in NAND flash processes or large-scale mass production of ternary storage particles, that the price of storage devices could possibly be lowered.

Wang Minjiang didn’t have a solution for this; the cost of chips in Xinyuan Company’s business was pitifully low, and they lacked relevant research.

Just as he felt somewhat at a loss, the chief engineer of SinoCore casually mentioned:

"Mr. Wang, don’t be so anxious; nothing is perfect, and the current silicon wafers aren’t really suitable for ternary anyway."

Wang Minjiang: !

"What do you mean? Are you exploring other technical routes?"

The chief engineer nodded:

"Using two binary components to simulate a ternary field-effect transistor is too wasteful. It’s clearly a compromise to adapt to the existing system; we must find a substrate with three inherent states as the material for natively supporting ternary chips.

The China Academy of Sciences has been researching this for quite some time. At present, they’re rather fond of silicon carbide, as both elements are fairly common and also relatively easy to obtain in high purity, meeting the requirement of having three states.

We’ve also got Modu City Microelectronics involved. Manufacturing chips from silicon carbide means reconstructing the production line. Currently, their plan is to lay out this path within five years and they should be bringing you in for collaboration in a few days."

After the chief engineer finished speaking, he noticed Wang Minjiang’s expression was very strange, his face contorted as if he wanted to say something but then hesitated; at last, he patted the chief engineer on the shoulder and solemnly said:

"Your spirit of innovation is excellent, Yellow River Semiconductor will definitely support you fully. Silicon carbide is good, let’s go with that."

The chief engineer of SinoCore was a bit puzzled: "Actually, it’s not that big of a deal because in recent years we’ve been working on silicon carbide IGBT power semiconductors, so it naturally occurred to us. Most ideas come from existing concepts."

Wang Minjiang: "That’s right; since there’s already a foundation for silicon carbide chips, I suggest we start with this, which should speed up the progress by a lot."