Skip to main content

Tengyun S2500, the strongest server CPU released by Tengyun, supports 8-channel 512 cores

According to the latest forecast by Million Insights, a market research organization, the global data center network market will reach US$40.9 billion by 2025 and achieve a compound annual growth rate of 11.0% from 2019 to 2025. Obviously, this will be a huge opportunity for server chip manufacturers.

Although the current server chip market is dominated by the X86 architecture, accounting for nearly 99% of the market share, in recent years, in Arm and its ecological partners (including Amazon, Huawei, Feiteng, Ampere, Marvell and other Arm server chip manufacturers) Thanks to the hard work, the Arm server market is gradually emerging.
Especially in the Chinese market, under the general trend of domestic independent control and localized substitution, the market opportunities of new infrastructure are superimposed, and the domestic Arm server market is accelerating.
On the morning of July 23, domestic CPU manufacturer Feiteng officially announced the three newly-named product lines, and at the same time released a new generation of scalable multi-channel server chip compatible with Arm instruction set-Tengyun S2500 series.
Three product lines enable new brands
Feiteng has always been a relatively low-key domestic CPU manufacturer. In terms of the company's product naming, it has always been the thinking of a man of science and technology.
We all know that Feiteng’s past chip naming basically started with "FT", followed by a four-digit number, sometimes followed by an English letter "A" and a number (the number usually represents the number of cores). Such a naming method is simply too technical, and it is even a bit confusing in the eyes of outsiders, which is totally unfavorable for product promotion.

For example, at the end of 2012, Feiteng launched the first 16-core general-purpose CPU, model FT1500; in 2014, Feiteng launched a new generation of general-purpose CPUs, including FT-1500A/4 for desktops and embedded systems, and FT-1500A for the server market. /16; In May 2016, Feiteng launched its first 64-core Arm server chip, model FT2000; in September 2016, Feiteng launched a dual-core CPU for desktop and embedded fields, model FT2000A/2; 2017 March 3 In June, Feiteng launched the second-generation 64-core Arm server chip, model FT-2000+/64; in June 2019, Feiteng launched a new generation of desktop quad-core processor FT-2000/4.

How about, judging from the product naming above, is it a bit dizzy? If you introduce it with your mouth, it may make others dizzier. It can be said that the naming of Feiteng's embedded CPU, desktop CPU and server CPU is completely unclear. This is indeed very inconvenient to understand and remember, and it is not easy to promote.
In contrast, Huawei uses different brands to differentiate its chips for different application areas. For example, the processor for mobile phones is called "Kirin", the processor for servers is called "Kunpeng", and the WiFi chip is called "Lingxiao." ", the baseband chip is called "Balong", the base station chip is called "Tiangang", and the smart screen chip is called "Honghu"...... Not only is it catchy and easy to remember, but it also has a good sense of meaning and picture.
I remember that when I visited the Feiteng Beijing office more than a year ago, the author suggested to an executive of Feiteng that the product line of Feiteng should be renamed and different brands should be used for a clearer distinction. Since the company name is "Fei Teng", the author suggested at that time to use three different types of birds, such as phoenix, peacock, and flying swallow, for the three existing product lines. Of course, this is just my personal opinion.
Unexpectedly, after a lapse of more than a year, at this press conference, Feiteng finally let its three product lines abandon the man-style naming method of science and engineering, and launched a new Chinese brand name: high-performance server CPU—— Feitengyun S series, high-performance desktop CPU-Feiteng Tengrui D series, high-end embedded CPU-Feitenglong E series. Compared with the previous naming method of FT-XXXA/XX, it instantly rises up, is there anything wrong? This also reflects a more "high-end" change in Feiteng's brand thinking.

Dou Qiang, general manager of Feiteng, said: "This year Feiteng will carry out a comprehensive brand upgrade of the three product lines to further condense core values ​​and create a new experience. In the future, Tengyun, Tengrui, and Tenglong will go hand in hand, and the three lines will fly together to promote Riding the wind and waves, all walks of life accelerate their progress."
Tengyun S2500 series released: up to 8 channels 512 cores
For server chips, performance is critical. For server chips, the most direct way to improve performance is to increase the number of cores and scalability, that is, multiple CPUs are directly connected to achieve large-scale CPU cluster processing.
The Tengyun S2500 series released by Feiteng adopts a 64-core architecture, which can support up to 8 direct connections, and can achieve up to 512 cores to work together.

In terms of specific parameters, the Tengyun S2500 processor is based on a 16nm process, integrates 64 FTC663 cores, clocked at 2.0-2.2GHz, on-chip also integrates 64MB three-level Cache, supports 8 DDR4-3200 storage channels. The area of ​​the whole chip is close to 400 square millimeters, the package size is 65mm×65mm, and the power consumption is 150W.

Although Tengyun S2500's single-channel chip performance is not much improved compared with FT2000+, it has made a qualitative leap in scalability, adding 4 direct connections, total bandwidth 800Gbps, and supporting 2 channels, 4-way and 8-way direct connection, the Cache coherency protocol extends from a single-way 64 core to 8-way 512 core. That is, a server can support up to 8 S2500 chips (512 cores in total) directly connected to form a multi-channel server.

Tengyun S2500 not only continues Feiteng’s previous high-end chip parallel system-on-chip (PSoC) architecture, data-friendly large-scale consistent storage architecture, and hierarchical two-dimensional Mesh interconnection network three advantages, but also adds Large-capacity shared L3Cache, multi-port high-speed and low-latency direct connections, memory mirroring storage reliability enhancement technology, and application-oriented security enhancement technology, can defend against ghost and fuse side-channel attacks, and the security and credibility of the product are further improved.
In terms of specific performance, according to the test data of Feiteng Research Institute, the performance of the server based on Tengyun S2500 has a SPECint score of 1000+ for two channels and a SPECint value of 1800+ for four channels, respectively reaching the original FT. -2000+ single-socket server twice and 3.5 times; in terms of distributed database performance, the tpmC value of the Tengyun S2500 dual-socket server reaches 98000, and the tpmC value of four-socket reaches 176000, respectively reaching FT-2000+ single-socket server In terms of cloud desktop support, the Tengyun S2500 dual-socket server can support 70 virtual machines, and the four-socket server supports 140 virtual machines, which are 2.5 times and 5 times that of the FT-2000+ single-socket server. Times.

"Compared with the FT-2000+ single-socket server, the performance of the multi-socket server based on Tengyun S2500 has been greatly improved." Dou Qiang emphasized: "This multi-socket server chip (Tengyun S2500) is a high-end chip complemented by Feiteng. The last piece of the server CPU product, and continuous innovation and development in the architecture."
In terms of specific server product forms, the Tengyun S2500 server mainly includes dual, four and eight-way servers, supporting up to 64 DDR4 memory modules, a maximum memory capacity of 4TB, 8-way PCIe 3.0 accelerator card, 24x2.5 SSD storage, 40GEA Network and so on.

It is understood that the current partners of Feiteng have launched the development of two-way, four-way and eight-way servers based on Tengyun S2500, and it is expected to be available on a large scale in the fourth quarter of this year.
Dou Qiang said, "After testing, the eight-way server based on Tengyun S2500 is currently the highest performance domestic server system."
Facilitate new infrastructure construction and push Tengyun S2500 to be fully popularized in multiple fields
On March 4 this year, at a meeting of the Standing Committee of the Political Bureau of the Communist Party of China Central Committee, the decision-makers emphasized the need to accelerate the progress of major projects and infrastructure construction that have been clearly defined by the national plan, which requires speeding up 5G networks, data centers, artificial intelligence, and industrial The progress of the construction of seven new infrastructures including the Internet. This also makes the "new infrastructure" quickly become the focus.
Subsequently, the National Development and Reform Commission clarified the definition of "new infrastructure", which includes three aspects: one is the construction of information infrastructure; the second is the construction of integrated infrastructure; the third is innovative infrastructure. And these three aspects of infrastructure construction are inseparable from the support of core chips.

Feiteng General Manager Dou Qiang also said: “Information technology is the core support of the new infrastructure. The new infrastructure is expected to lead the information industry into a new round of development cycle and detonate the entire industry. As a state-owned technological innovation enterprise, a leading domestic independent company As a core chip provider, Feiteng will take advantage of the national policy to devote itself to serving the transformation and upgrading of information technology with domestic computing power and serving new infrastructure."
In this regard, Feiteng will also rely on Tengyun S2500's five core capabilities of high scalability, high performance, high security, high reliability, and high energy efficiency to empower cloud computing, big data, edge computing, 5G, AI, blockchain, etc. Technology has been widely used in data centers, cloud computing platforms, virtualization, Internet of Things and other fields, including government affairs, digital cities, telecommunications, finance, energy, transportation, industrial manufacturing and many other industries.

"Tengyun S2500 will help PC, server, cloud computing, data center and other partners to achieve transformation and upgrading, and share the new infrastructure dividend. In the second half of this year, we will promote and apply in key areas, and achieve full popularity in more areas next year." Dou Qiang said with confidence.
Strengthen safety and help Xinchuang industry
In the Xinchuang industry, which has higher requirements for independent control and information security, the "Level Protection 2.0 National Standard" officially implemented in December 2019 requires the full use of safe and reliable products and services to ensure the security of the critical infrastructure. Trusted computing should start from the underlying basic hardware such as chips, prevent problems before they occur, and effectively establish a network security "immune system" without destroying the software code logic.
According to the data, the state has asked all provinces and ministries and commissions to strive to complete the annual procurement tasks of Xinchuang before the end of July 2020. Adjust the replacement ratio between 2021 and 2022, and strive to complete the project implementation by June 2022. This also means that most provinces and ministries will complete the three-year procurement task before the end of 2021. This will undoubtedly be a huge opportunity for domestic CPU chip manufacturers.

At present, Feiteng has built a safe and reliable domestic computing system. First, at the chip design level, security-related functional modules are integrated, chip security features are improved, and the PSPA security architecture standard for Feiteng processors is formed. At the Feiteng Ecological Partner Conference on December 19, Feiteng officially released the PSPA 1.0 standard, and released the standard technical white paper and related technical specifications. This is the first time that a domestic CPU company has released a CPU-level safety architecture standard, which has realized the bottom-up intrinsic safety of domestic computer systems. The second is to build an active immune trusted computing platform, an endogenous immune anti-corruption subsystem, and an independent high-performance green computing platform with balanced performance and outstanding energy efficiency at the level of system solutions.
The Tengyun S2500 launched by Feiteng this time also adopts application-oriented security enhancement technology, which can defend against ghost and fuse side-channel attacks, and the security and credibility of the product are further improved.
Tengyun S2500 assists, Feiteng CPU shipments will exceed 1 million pieces this year
After nearly 20 years of technology accumulation, Feiteng has formed a complete pedigree of diversified computing power products. It is the CPU manufacturer with the most complete pedigree among general-purpose CPUs in China, including high-performance server CPUs, high-performance desktop CPUs, and high-end embedded CPUs. Provide core computing power support for all types of devices from end to cloud.
According to Feiteng, 200,000 chips have been applied to its previous FT-1500A series of chips. The current main products are FT-2000+/64 for servers, FT-2000/4 for desktop terminals, and FT-2000A for embedded devices. /2. Under the influence of the epidemic in the first half of this year, it still achieved growth against the trend, with revenue of 300 million yuan.

With the launch of Tengyun S2500, Feiteng is more confident about the growth in the second half of the year. According to Feiteng’s forecast, the annual shipments this year will reach more than 1 million pieces, a four-fold increase compared to 2019, and the annual revenue will reach 1 billion.
Future product planning and ecological construction
In addition to the Tengyun S2500 series released this time, Feiteng also announced the follow-up plans for the three series of Tengrui and Tenglong products at this conference.

Tengyun S series has planned two core products, Tengyun S5000 and Tengyun S6000. Among them, Tengyun S5000 will use the 7nm process and support PSPA1.0 security architecture, and will be released in the third quarter of 2021; the Tengyun S6000 series will use 5nm process, the overall performance will be doubled, and support PSPA2.0 security architecture.
The Teng Rui D series also plans two core products, Teng Rui D2000 and Teng Rui D3000. Among them, the Teng Rui D2000 series will be based on the 14nm process and support the PSPA1.0 security architecture and will be released in the fourth quarter of this year; the Teng Rui D3000 is also based on the 14nm process, but the single-core performance will be doubled by then and support the PSPA2.0 security architecture.
Tenglong E series also has two core products, Tenglong E2000 and Tenglong E3000. Among them, the Tenglong E2000 will be based on the 14nm process and support the PSPA1.0 security architecture, and will be released in the second quarter of 2021; the Tenglong E3000 will also be based on the 14nm process, while integrating rich I/O interfaces and supporting the PSPA2.0 architecture.
"In terms of the strategy of technological innovation, we adhere to the 16-character policy. From end to the cloud, it is to design processor chips that can meet the needs of different devices. On-demand customization is to customize chips according to the needs of different users in different industries. Safe and reliable, Independence does not mean security. Security must be superimposed on autonomy. Open cooperation and hope to accomplish the goal of technological innovation with partners." Dou Qiang said.
For domestic CPU manufacturers, good products are not enough. They also need a rich ecosystem to succeed.
According to reports, Feiteng has cooperated with many domestic manufacturers and has more than 1,000 partners. It has developed more than 900 complete machine products in 6 categories, and more than 2,400 software and peripherals have been adapted and are being adapted.
It is worth mentioning that in April of this year, Feiteng also worked with ecological partners to open industry solutions, and released 4 categories and more than 80 industry joint solutions, covering Xinchuang, telecommunications, finance, energy, transportation, medical, Digital city, industrial manufacturing and other industry fields, these standardized, practical, and systematic industry joint solutions have achieved "out of the box


Popular posts from this blog

AMD's GPU technology enters the mobile phone chip market for the first time

In addition to the release of the Exynos2100 processor, Samsung also confirmed a major event at this Exynos event, that is, the custom GPU that they have worked with AMD for many years will soon appear and will be used on the next flagship machine. The current Exynos2100 processor uses ARM’s Mali-G78GPU core with a total of 14 cores, so the GPU architecture developed by Samsung will be the next Exynos processor, and the GPU will be the focus. This is probably the meaning of Exynos2100’s GPU stacking. The key reason. Dr. InyupKang, president of Samsung’s LSI business, confirmed that the next-generation mobile GPU in cooperation with AMD will be used in the next flagship product, but he did not specify which product. Samsung is not talking about the next-generation flagship but the next one, so it is very likely that a new Exynos processor will be available this year, either for the GalaxyNote21 series or the new generation of folding screen GalaxyZFold3. In 2019, AMD and Samsung reached

Apple and Intel want to join the game, what happened to the GPU market?

Intel recently announced that it will launch Xe-LP GPU at the end of this year, officially entering the independent GPU market, and will hand over to TSMC for foundry. At the 2020 WWDC held not long ago, Apple also revealed that it is possible to abandon AMD's GPU and use a self-developed solution based on the ARM architecture. It will launch a self-developed GPU next year. What happened to the GPU market? Why are the giants entering the game?    Massive data calls for high-performance GPU    Why has the demand for GPUs increased so rapidly in recent years? Because we are entering an era where everything needs to be visualized. Dai Shuyu, a partner of Aiwa (Beijing) Technology Co., Ltd., told a reporter from China Electronics News that visualization requires a large amount of graphics and image computing capabilities, and a large amount of high-performance image processing capabilities are required for both the cloud and the edge.    Aiwa (Beijing) Technology Co., Ltd. is an enterp

NVIDIA officially launches RTX 30 series mobile graphics cards

In the early morning of January 13, NVIDIA officially launched the RTX30 series of mobile graphics cards at the CES2021 exhibition. Ampere-based GPUs have also reached the mobile terminal, mainly including RTX3080, RTX3070 and RTX3060 models. In addition to improving game performance, the RTX30 series of mobile graphics cards have twice the energy efficiency of the previous generation, and support the third-generation Max-Q technology, mainly supporting DynamicBoost2.0 dynamic acceleration technology, WisperMode2.0 noise control, ResizableBAR (similar to AMD’s SAM technology) and DLSS. The third-generation Max-Q technology uses AI and new system optimization to make high-performance gaming laptops faster and more powerful than ever. These technologies include: ·DynamicBoost2.0: The CPU and GPU powers of traditional gaming notebooks are fixed, while games and creative applications are dynamic, and the requirements for the system will vary with the number of frames. With DynamicBoost2.0,