Shulou(Shulou.com)11/24 Report--

CTOnews.com November 3 news, today, Moore Thread 2022 autumn conference was held in Beijing Zhongguancun National Independent Innovation demonstration Zone. At the event, Moore Thread unveiled MTTS80, the first graphics card that supports Windows environment and DirectX graphics interface, a new multi-function GPU chip Chunxiao, MTT S3000 for server applications, and MCCX, a meta-computing all-in-one machine. This is a new multi-functional GPU product released by Moore Thread after a lapse of seven months.

Moore Thread first released the second multi-function GPU chip "Chunxiao" based on MUSA architecture. Compared with "Suti" released in March this year, Chunxiao's four built-in computing engines have been fully upgraded to support graphics and image rendering, 8K video codec, AI training and reasoning, general computing, GPU virtualization, scientific simulation computing and other functions at the same time.

The most important product is the first "national trend graphics card" MTT S80, which can play games freely. It is based on the multi-function GPU chip "Chunxiao". It integrates 22 billion transistors and has a built-in MUSA architecture, general computing core and tensor computing core, which can support computing accuracy such as FP32, FP16 and INT8.

In terms of core parameters, MTT S80 has 4096 MUSA cores, main frequency 1.8GHz, 16GB GDDR6 video memory, video memory bit width 256bit, uses the latest PCIe 5.0 DP1.4a 16 interface, and provides 3 DP1.4a and 1 HDMI 2.1 interface support, up to 4 screens 8K 30 frame output.

The most special thing about MTT S80 is that it is the first GPU in China to support Windows environment and DirectX graphical interface. Moore Thread said at the press conference that the Windows driver of MTT S80 has a built-in MUSA DirectX Driver module, and has adapted dozens of mainstream games such as Diablo 3, League of Legends and Crossing the Line of Fire, providing full-featured support based on PC scenes.

At the press conference, Moore Thread demonstrated the game of Diablo 3 on the spot, and the game experience was smooth.

Another new product is the server-oriented MTT S3000. It is also based on "Chunxiao" and can support mainstream graphics and computing interfaces such as DirectX, OpenGL, OpenGL ES, Vulkan, OpenCL, and is compatible with CUDA. It can provide general intelligent computing support for AI reasoning and training, cloud games, cloud rendering, video cloud, digital twins, digital content creation and other scenarios.

In terms of core parameters, the MTT S3000 series contains 4096 MUSA stream processing cores and 128dedicated tensor computing cores. The transistor scale is 22 billion, the operating frequency is 1.9GHz, and the video memory bit width is 256bit. it is equipped with 32GB GDDR6 video memory, and supports FP32, FP16, INT8 and other computing accuracy, of which FP32 computing power can reach 15.2TFLOPS.

For enterprise server products, what is more important is its compatibility with algorithm models and application deployment. MTT S3000 and its supporting software and hardware products can cover the whole process from algorithm model to application deployment, and can provide AI users with a friendly and rich package of solutions. At the algorithm level, Moore thread can not only support user-defined algorithm models, but also provide a rich pre-training model library, which can empower the majority of AI developers and improve development efficiency. In addition, the MTT S3000 also provides a CUDA ON MUSA compatible solution, and the code developed on the CUDA can be seamlessly migrated to the MTT S3000.

GPU is the key computing power needed by the current data center, and the application of cloud native technology is becoming more and more common in cloud computing, cloud desktop, cloud games and other fields. To this end, Moore Thread released a series of native GPU cloud solutions based on Moore Thread's innovative MT Mesh 2.0. You can automatically allocate GPU computing and video memory resources according to the cloud-centric application load to achieve GPU computing flexibility and scaling. You can split a GPU into multiple containers or virtual machines at will, or you can support a container or virtual machines to schedule multiple GPU.

At the press conference, Moore Thread demonstrated the use of MTT S3000 cloud GPU computing full frame to play "Arena of Valor".

Finally, Moore Thread also released the MTVERSE metacosmos platform specially built for meta-cosmic applications and a number of software and hardware products, including the MCCX metacomputer based on the MTT S3000. Based on the computing power of Moore thread MUSA GPU cluster, MTVERSE provides users with computing infrastructure and services, including big data, AI training and reasoning, graphic rendering and physical simulation, to further simplify the difficulty of application and solution development and shorten the development cycle.

It is undoubtedly a long and difficult process for domestic GPU to be compatible with the old software ecology. Its difficulty lies not only in designing game graphics cards, but also in how to develop drivers. As strong as intel, it also encountered setbacks in driver development when entering the independent graphics market. Due to the strong professionalism of computer graphics, there are few people in the world who can understand the development of Windows drivers, most of which are concentrated in western countries, and there are only a handful of professionals in China. Domestic GPU companies in the start-up stage need to quickly launch market-oriented GPU products, but the problem they face lies in the lack of talent and team honing in key areas such as chip design, underlying driver development and so on. So developing a generic GPU is by no means easy.

Independent innovation is a very difficult road, but it is also a road that we have to take. Today we are fortunate to see that Moore threads have taken the first step towards compatibility with mainstream interfaces. But we should also look at it rationally, and we cannot expect it to rise to the top in one step. When evaluating GPU products and their performance, we should not be divorced from the application environment and the background of the times, but should give maximum encouragement and tolerance. Of course, the Moore thread itself should continue to follow up, do a good job of driver updates, environment and application adaptation, and localization services.

Last. The consumer-oriented graphics card product MTT S80 will be limited to JD.com during the double 11 period. If you are interested in domestic GPU products, you might as well pay attention to it.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.