The Intel company consists of many departments that specialize in certain areas. The most advanced and often experimental solutions are developed by Intel Laboratories or Intel Labs. The company has many research laboratories around the world, which allows it to work more closely with universities, government institutions and companies in different countries.
It is noteworthy that it was Pat Gelsinger, the current head of Intel, who brought disparate labs under the auspices of one entity almost 30 years ago and established a teamwork between them. During that time, many technologies have left the walls of Intel Labs and become so familiar that you don’t even notice them. It’s hard to imagine that they once had experimental company status and went through multiple stages of audits assessing their prospects for development, implementation, scaling, and the overall cost of development.
The last point is the “sickest”, as for most other research labs in any field. Software development can be completed in one to two years. Microarchitecture solutions require more time. For example, the VT-x virtualization technology was “cooked” in the laboratory for about three to four years, and the entire path from the idea to the end products took about five to six years. The most complex developments obviously concern solutions in “silicon”.
For example, among the most notable products initiated and later developed by Intel along with other giants of the IT industry are the USB and PCI Express standards, which have now merged into Thunderbolt thanks to Intel Labs specialists. There were also completely unusual solutions: Intel Labs developed a computer system that allowed the famous theoretical physicist, cosmologist and astrophysicist Stephen Hawking to communicate with the world, despite the loss of speech due to illness. Subsequently, the development open-source Platform.
Currently, the key areas for Intel Labs are: artificial intelligence, data security, technologies for data centers and 5G infrastructure, new approaches to data processing, increasing the efficiency of software and hardware development, custom and specialized chips, and of course , new microarchitectures for future processors. But that’s not all – Intel Labs employees have no intention of divulging all secrets.
In the field of new computing, quantum systems are one of the main areas of development. Here, Intel Labs develops both software and hardware. The company focuses on the development of quantum processors based on spin qubits. She considers this direction to be the most promising due to the ease of scaling and easy integration into existing CMOS solutions.
But a quantum processor alone cannot function under normal conditions – it needs extremely low temperatures, and these must be maintained with great accuracy. For this reason, Intel Labs develop cryogenic control systems that, in addition to accuracy, also have high efficiency and reliability. In addition, the company pays much attention to the software component and works on quantum compilers and various algorithms for quantum computing.
Another important area for Intel is neuromorphic computing, which will help make AI systems more like real human thinking. The “traditional” AI approach requires deep learning on huge data sets, resulting in huge energy costs, while at the same time limiting such systems to only a given area of “knowledge” – essentially an AI model for image recognition, cannot process speech. Neuromorphic computing is much more efficient.
You (self-)learn faster and can already “pull up” external sources during the work process. In addition, there are many other uses for them, including combinatorial optimization, graph search, thermal diffusion modeling, and more. In all cases, the neuromorphic processor turns out to be dozens of times more productive than chips with traditional architectures, while consuming much less energy. One of the results of Intel Labs’ work in this direction is the Loihi architecture and chips based on it.
Another interesting area of development from Intel Labs is full homomorphic encryption (FHE), which allows you to protect information much more efficiently. This technology allows you to perform calculations directly on encrypted data without the need to decrypt and reverse the encryption. According to Intel, the developed approach is 1,000 times faster than the existing ones. FHE will complement or replace current solutions such as the Intel Software Guard Extensions (SGX) instructions, also created by Intel Labs.
Both FHE and SGX are important for confidential calculation and especially for federated training of AI models based on many datasets from disparate sources that do not want their information to be visible to the rest of the participants in the process. For example, the world’s largest database of medical data for training specialized AI was created. In this case, it is imperative that the entire system and its users do not have access to sensitive information such as patients’ personal data.
Another activity of Intel Labs is silicon photonicson which the company has been working for more than ten years. Simply put, it is the integration of laser transceivers directly into chips for ultra-fast data transmission with minimal power consumption. This is necessary because more and more energy is being used for data transmission using the classic electrical route and the capacity of the channels is reaching its limits, including the physical ones – modern chips already require several thousand contacts. As a first step, it is planned to connect all chips (CPU, GPU, FPGA, ASIC, SSD etc.) inside servers and inside server racks directly to “optics”.
Finally, another development from Intel Labs that recently made its debut in the finished product should not go unmentioned. We are talking about ASIC for mining called Bonanza Mine. We have previously spoken in detail about the chip itself and the miner based on it. And here we just note that Bonanza Mine has turned out to be an extremely compact and efficient chip. In terms of specific energy consumption, it is one of the best on the market. And judging by the large orders for these chips, Intel was able to offer an attractive price for new items.
This is not an exhaustive list of Intel Labs developments. The company isn’t ready to talk about many of them in detail. And not only because these are still secret studies on which the future of the company will depend, but also because not all of them will see the light of day soon. For example, the head of Intel Labs Rick Uhlig (Rich Uhlig) mentioned the study of the possibility of combining individual blocks inside the chips with … wireless. It turns out that over short distances this can be more profitable in every way than wiring thousands and thousands of connections in the interposer.