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

Thanks to CTOnews.com netizen Xiao Zhan for the clue delivery! CTOnews.com January 11 news, domestic GPU manufacturer Moore Thread held a GPU cloud desktop product conference yesterday to launch new vPC cloud desktop virtualization GPU products such as MT vGPU1.0, MT GPU pass-through and MT GPU acceleration protocol coding based on Moore thread multi-function server GPU cloud desktop products, and jointly launched a number of cloud desktop solutions based on MTT S2000 Join hands to promote the landing application and deployment of Moore thread GPU products in the cloud desktop field.

According to reports, as the first innovative enterprise with independent cloud desktop GPU virtualization technology in China, Moore Thread launched the "Moore Thread Cloud Desktop GPU acceleration Co-Construction Plan" and signed strategic cooperation agreements with five key enterprises, including Lenovo Cloud, an Chaoyun, CoolStack Technology, Tingyu Technology, and Shenxin Technology. Moore Thread will provide test graphics cards, solution integration and related R & D resources, and carry out in-depth cooperation with ecological partners in terms of unique function development, market cooperation promotion and so on.

Moore Thread MTT S2000 is a data center-class multi-functional GPU product based on MUSA unified system architecture, with built-in functional engines such as graphics rendering, video codec, artificial intelligence acceleration and parallel computing. It is also compatible with mainstream CPU architectures such as X86 and ARM, as well as operating systems such as Kirin and Universe.

MTT S2000 uses 12nm process, 4096 MUSA cores, maximum video memory 32GB, single-precision computing power of the highest 12TFlops, passive heat dissipation, single-slot design to meet the high-density configuration of the data center.

The GPU supports graphics API such as OpenGL, OpenGL ES, DirectX, Vulkan, audio and video processing through FFMPEG, VA-API / DXVA compatibility, and supports H.264, H.265, AV1 multi-channel HD video codec. AI and scientific computing programs are compatible with OpenCL and Vulkan.

In terms of PC cloud desktops, MTT S2000 pioneered hardware-based MT Mesh 1.0 GPU virtualization technology, complying with the industry standard SR-IOV, and can accommodate multiple virtualized users to work remotely on each physical GPU.

The new GPU cloud desktop product based on Moore thread MTT S2000 includes MT vGPU1.0, MT GPU pass-through and MT GPU acceleration protocol coding and has the following features:

MT vGPU 1.0:

Through the self-developed virtualization driver, a physical GPU is divided into multiple vGPU, which concurrently supports the operation of multiple virtual machines. The driver supports DirectX rendering framework and DXVA codec framework in Windows to achieve vGPU hardware accelerated graphics and image rendering, video decoding, WebGL browser browsing and other functions. Based on the dynamic elastic segmentation technology provided by MT Mesh virtualization technology, it provides different configurations of vGPU for different business loads, gives full play to the function of GPU, and improves the utilization of GPU.

MT GPU through:

Through PCIe transmission technology, a complete GPU chip is transparently transferred to the virtual machine, and the drivers in the virtual machine are compatible with rendering frameworks such as DirectX,OpenGL, accelerating rendering, video codec and other functional scenarios, and compatible with Windows10, Linux, Kirin, Universe and other operating systems.

MT GPU acceleration Protocol Encoding:

Encode the display output of the cloud desktop with GPU hardware to accelerate the protocol video coding, meet the needs of multi-terminal adaptation, and improve the concurrency of the cloud desktop on a single server.

CTOnews.com learned that Moore Thread's new GPU cloud desktop product with MTT S2000 can significantly improve the performance of cloud desktop applications and provide six core values:

20 parallel routes

MTT S2000 single card can achieve parallel work of up to 20 cloud desktops with 1080p 30FPS specification while satisfying 2D / 3D picture rendering, and the number of cloud desktops increases linearly according to the number of GPU deployments.

Dynamic elastic segmentation

MTT S2000 supports vGPU and GPU pass-through virtualization, and users are free to choose between the two modes for different application scenarios. In vGPU mode, users can split GPU resources according to video memory capacity such as 1.5, 3.5, 7, 14, GB and so on. At the same time, users can recycle idle cloud desktops at any time and split the remaining GPU resources again to build new cloud desktops for other users.

Reduce CPU load

MTT S2000 can automatically take over the load of all kinds of professional software in cloud desktops, achieve high-performance 2D / 3D image rendering, and provide users with a smooth software application experience while reducing CPU load by 50%. In the live demonstration of the educational application "TELLO EDU", MTT S2000 can increase the FPS from 10 frames to 60 frames, a six-fold improvement in performance.

Video hardware decoding

MTT S2000 can provide cloud desktop users with 1080p video hardware decoding capability, support H.264 and other common decoding formats, and continue to support massive coding formats such as AV1, H.265, H.263, VP9 / VP8, MPEG-1/2/4, VC1, AVS / AVS+/AVS2 through software upgrades, with a wider range of adaptability.

When playing 1080p video on a 2-core 4GB virtual machine, MTT S2000 reduces CPU occupancy from 100% to an average of 20%, and continues to provide a smooth video frame rate when the host has other calculations.

Office application acceleration

MTT S2000 enables hardware codec of video streams in many mainstream video conferencing software, reducing CPU load by 50%, leaving room for more office software and background processes to run at the same time. In addition, vGPU based on MTT S2000 can support WebGL applications under cloud desktops, greatly improve the experience of 3D rendering scenes, save more than 70% of CPU computing resources, improve the response speed of cloud desktops to concurrent computing, and effectively increase the number of concurrent paths of CVMs.

Application adaptation and extension

Moore thread vGPU can meet the needs of light 3D real-time rendering of educational software such as the "new curriculum reform", provide a smooth interactive experience, reduce the consumption of one-way CPU computing resources, and increase the number of concurrent cloud desktop services. At present, Moore Thread has worked with partners to complete the complete adaptation of 19 new curriculum reform (education) applications and 60 common Windows applications, and Moore Thread will continue to work together with ecological partners to complete more commonly used and professional software adaptation work.

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.