Fenghua 2 high-performance domestic desktop GPU released for the first time, contract size reaches 500 million yuan

Publisher:EE小广播Latest update time:2022-08-11 Source: EEWORLD Reading articles on mobile phones Scan QR code
Read articles on your mobile phone anytime, anywhere


Sixteen years of accumulation, Fenghua GPU continues to reach new heights


Rome was not built in a day, and the door of opportunity for domestic GPUs is by no means created overnight. The successful development of "Fenghua 2" is the result of 16 years of long-term cultivation and craftsmanship by the Innosilicon team. Mao Mingming, vice president of Innosilicon Technology, introduced, "For 16 years, Innosilicon Technology has been low-key and pragmatic, daring to be the first, and overcoming difficulties. The team has been sitting on the bench for ten years and has more than 200 advanced process tape-outs, setting a record in the industry. The company has remained profitable despite major investment in GPU development and does not need any financing to maintain it. It can be said that we did not spend a penny of investors' money on GPUs, because Innosilicon technology is used behind more than 6 billion advanced SOC chips of well-known companies around the world, and our income can support our healthy and organic development. Innosilicon is a company that creates core value through its own hard technology. Innosilicon used to be a behind-the-scenes hero. In to B's circle is very famous, providing empowerment for major semiconductor foundries and upstream and downstream customers . Driven by demand and combining years of accumulated advantages, Innosilicon has opened up mountains and forests in the global market through customized mass production and has achieved fruitful results. In 2021, Innosilicon released the data center server GPU 'Fenghua No. 1', which realized multi-user 1080P and multi-channel hardware virtualization, and various high-definition encoding/decoding/transcoding functions for the first time in the field of cloud applications and data centers. In the past six months, Innosilicon has continued to innovate and pursue success, completing a full set of kernel underlying technology optimization and ecological adaptation, and launching a low-power version of desktop products as expected. "

image.png image.png image.png image.png

▲The GPU exhibition area was very popular


The "Fenghua 1", which has entered industrialization, demonstrated its smooth operation of Honor of Kings in a multi-channel 1080P cloud gaming scenario at the scene, which amazed many young audiences. The person in charge of (China Unicom) Xiaowo Technology said, "The layout of domestic cloud games has always been completely dependent on international mainstream graphics cards, which are always lacking in high cost performance and customization. Seeing such a domestic GPU that can support multi-channel virtualization is really shocking and worth the trip!"


image.png

▲Fenghua 1 can support 16-channel 1080P cloud games at the same time


The continuous evolution of high-tech self-developed technology is the key to GPU iteration capabilities and core competitiveness. He Ying, GPU project director of Innosilicon Technology, said, "Modern GPUs cover a wide range of areas and have many technical barriers in development and application. To come up with a graphics GPU that is benchmarked against the industry's advanced level and is smoothly compatible with various applications, buying a GPU IP will not achieve anything. Innosilicon's one-stop IP and full set of system architecture customization services provide innovative power for the rapid iteration of Fenghua GPUs: breaking the memory wall, the world's strongest GDDR6/6X Combo IP, HBM3/2e IP, LPDDR5/5X/DDR5 IP and other high-speed storage solutions; 32/56G SerDes (PCIe6/5/4) high-speed communication interface, UCIe Chiplet interconnection solutions, and autonomous GPU core evolution. Therefore, like Apple, Innosilicon has continuously optimized the GPU core based on the initial architecture license. It has now completely and independently mastered the GPU core evolution architecture diagram, has its own GPU architecture control, and is capable of releasing two GPUs every year in the next three years. For example, Fenghua 2 can support APIs above DX11/OpenGL4.2, which are not available in the original GPU architecture. "


Dr. Yang Xile, Chief GPU Scientist of Innosilicon Technology, who has been engaged in leading architecture development for the world's top GPU companies for more than 25 years and holds 125 international patents in the core field of 3D graphics, shared, "Following the release of the first GPU 'Fenghua 1' in 2021, Innosilicon quickly launched a new low-power and high-performance GPU 'Fenghua 2' with independent technical characteristics. This is because we have a very excellent technical team, from the GPU architecture R&D team to the GPU hardware and software development team, high-end DDR and bus design and experienced back-end team, etc. We work together to overcome difficulties and make rapid progress. Making GPUs is a very complicated thing, but the Innosilicon team has never been afraid of difficulties. Targeting the markets such as office desktops, in-vehicle smart cockpits, cloud games, and cloud computing, we continue to launch new GPU products, design and evolve advanced GPU architectures, achieve top rendering performance, while taking into account low power consumption and highly scalable designs to meet the different needs of customers. In the future, several Fenghua GPUs will support the most advanced graphics processing ecosystems such as ray tracing, and further support Direct X12, OpenGL4.6, Vulkan and other frameworks; in addition, in the future, Innosilicon will also develop high-performance GPGPU to meet a wide range of application fields such as AI training and reasoning, image processing, general computing, scientific and engineering computing, etc. Innosilicon has a first-class team of experienced experts in the GPU field. We are confident that we can accept all challenges, and we are confident that we can meet the international top level and open up a new future for domestic GPUs. "


How can domestic GPUs achieve win-win results in the digital wave?


With the surge in demand for 5G new infrastructure, medical imaging, intelligent manufacturing, autonomous driving, smart cockpits, cloud computing, and localization of smart terminals, GPU applications continue to expand, and China's GPU market size and growth potential are huge, currently approaching 50 billion yuan. In this regard, experts participating in the roundtable forum said, "The more the digital economy develops, the higher the performance requirements of the GPU on the application side. Performance mismatch, poor applicability, imperfect functions, and immature system ecology are obviously the pain points of domestic products competing in the market, and are also the reasons why terminal companies and users dare not and are unwilling to use domestic GPUs. However, the Fenghua series GPU dares to hit the pain points directly, and is the first to break the tight spell. With leading products that are precisely customized and continuously iterated, it takes advantage of functional customization, timely response, and product versatility to stand on the top of the wave." In terms of how to jointly build a new ecology and promote the process of localizing GPUs, the experts at the roundtable forum unanimously agreed that industry upstream and downstream partners, academic institutions, and research institutes need to work together and build an ecology to achieve long-term win-win results. It is a long process for domestic GPUs to go from being usable to being easy to use. It takes time to promote, plan, optimize, and improve, and more importantly, it requires user support, tolerance, and error correction.

image.png

▲The atmosphere was warm and the venue was packed with guests


Facing everyone's high expectations for the Fenghua series GPUs, Mr. Ao Hai, President of Innosilicon Technology, said, "Innosilicon's mission is to bring Fenghua GPUs into thousands of households and empower smart life. 'Fenghua 2' is the beginning of Innosilicon's empowerment of the desktop ecosystem, and will kick off the Fenghua series GPUs' entry into the mass market. The powerful iteration capabilities of advanced processes and the team's ability to overcome difficulties are Innosilicon's inherent advantages. We rely on our own IP and the support of global foundry partners to continuously evolve advanced technologies to ensure continued performance leadership, supply chain security and superior cost-effectiveness to meet users' customization needs. At present, the research and development of 'Fenghua 3' has been basically completed, and it will provide domestic users with cool performance such as super computing power and ray tracing; GPGPU and cockpit products are also being defined with customers, and more customized computing product lines will follow in the future. Relying on research and development, based on the market, looking at the world, we are determined to empower the domestic GPU industry chain!"


*For more information about Fenghua GPU, please visit the official website; our customers’ success is our success, and Innosilicon will serve you wholeheartedly.


[1] [2]
Reference address:Fenghua 2 high-performance domestic desktop GPU released for the first time, contract size reaches 500 million yuan

Previous article:Renesas launches HMI MPU RZ/A3UL supporting RTOS and fast boot
Next article:Mir: First-person experience of the domestic processor T507-H development board

Recommended ReadingLatest update time:2024-11-16 12:48

Innovative distributed functional safety mechanism based on GPU device behavior protects intelligent driving
With the rapid improvement of the intelligence of automobiles, a large number of new processors and system-on-chips (SoCs) have been widely introduced into vehicles. Whether in driving or cockpit scenarios, whether using domain controller mode or emerging central control unit mode, all are considering adding more in
[Automotive Electronics]
Innovative distributed functional safety mechanism based on GPU device behavior protects intelligent driving
Hanbo Semiconductor received 1.6 billion yuan in financing and will increase investment in the development of graphics GPU product lines
Recently, Hanbo Semiconductor received 1.6 billion yuan in B1 and B2 rounds of financing, led by Alibaba Group, PICC Capital, Matrix Partners and 5Y Capital, and followed by China Life Science and Technology Innovation Fund, Mirae Asset, Cornerstone Capital, Muhua Science and Technology Innovation Fund (with Tsinghua
[Mobile phone portable]
NVIDIA Unveils New Ada Lovelace RTX GPUs for Designers and Creators
RTX 6000 GPUs to deliver 2-4x performance boost to enterprise workflows; global manufacturers to begin shipping in Q4 SANTA CLARA, Calif., Sept. 20, 2022 /PRNewswire/ -- NVIDIA today unveiled the NVIDIA RTX™ 6000 workstation GPU, based on the new NVIDIA Ada Lovelace architecture, usher
[Industrial Control]
NVIDIA Unveils New Ada Lovelace RTX GPUs for Designers and Creators
Samsung may launch Exynos 2200 on November 19, with support from AMD GPU technology
      Foreign media said that Samsung Electronics may officially unveil the new generation "Exynos 2200" on November 19, but some people believe it will be Exynos 1250.   Samsung Exynos posted a mysterious post on Instagram, saying that "On November 19, everything changes," and that it will show us what "future entert
[Mobile phone portable]
Foreign media reveals the performance of Kirin 9000 and Snapdragon 888: GPU is tied
       Foreign media Anandtech showed the performance of Snapdragon 888. Judging from the actual test results, it has made great improvements compared to the previous generation Snapdragon 865, such as a 25% increase in CPU performance and a 35% increase in GPU performance.   From more test details, in the GFXBench te
[Mobile phone portable]
Foreign media reveals the performance of Kirin 9000 and Snapdragon 888: GPU is tied
Imagination GPUs provide efficient acceleration for edge intelligence
As an important partner of ESWIN Computing, Imagination was invited to attend the "2024 ESWIN Computing Developer Partner Conference" held in Yizhuang, Beijing on September 10. Experts from Imagination's UK headquarters delivered a speech at the conference entitled "Welcoming the arrival of edge-based generative
[Network Communication]
Imagination GPUs provide efficient acceleration for edge intelligence
Actions speak louder than words: the next generation of domestic GPUs is in full swing
A few years ago, AMD, Qualcomm and Intel entered the field of artificial intelligence one after another. Faced with the situation that the field of artificial intelligence chips has been occupied by large companies, "Leather Jacket Master" Huang Renxun said to the media, if this market and industry have developed well
[Mobile phone portable]
Latest Embedded Articles
Change More Related Popular Components

EEWorld
subscription
account

EEWorld
service
account

Automotive
development
circle

About Us Customer Service Contact Information Datasheet Sitemap LatestNews


Room 1530, 15th Floor, Building B, No.18 Zhongguancun Street, Haidian District, Beijing, Postal Code: 100190 China Telephone: 008610 8235 0740

Copyright © 2005-2024 EEWORLD.com.cn, Inc. All rights reserved 京ICP证060456号 京ICP备10001474号-1 电信业务审批[2006]字第258号函 京公网安备 11010802033920号