Suiyuan chip computing power
Publish: 2021-04-14 21:14:49
1. In December 2020, Inspur edge computing server ne5260m5 announced that it will complete compatible adaptation and deep optimization for the newly released AI reasoning acceleration card "yunsui I10" of Suiyuan technology. It can support 2-4 pieces of yunsui I10 according to customers' needs and provide support for edge AI reasoning applications, such as automatic driving, intelligent transportation, etc recognition, predictive maintenance of precision instruments, intelligent transportation, etc recognition and so on AI instrial quality inspection and other scenarios provide powerful computing power support, and the maximum performance can reach 70 trillion times per second (70tflops). The performance of high-precision mainstream reasoning model achieves the instry-leading, and provides more flexible acceleration options for edge environment. At present, the wave server ne5260m5 based on yunsui I10 will further promote the instrialization of AI to edge scenes, and drive the digital and intelligent transformation of various instries.
2. A few days ago, Inspur server ne5260m5 announced the completion of compatible adaptation and deep optimization for the newly released AI reasoning acceleration card "yunsui I10" of Suiyuan technology, providing powerful computing support for edge AI reasoning applications such as automatic driving, intelligent transportation, etc recognition, predictive maintenance of precision instruments, AI instrial quality inspection, etc. Inspur server ne5260m5 is small in size, but it supports a variety of heterogeneous acceleration cards, and each mole has a good heat dissipation effect, which can well adapt to the edge of harsh high temperature and humidity environment. Based on the highly adaptable combination of software and hardware, Inspur server ne5260m5 can carry 2-4 yunsui I10, which can provide up to 70 trillion times of single precision computing power per second and 280 trillion times of semi precision computing power per second, providing more flexible acceleration options for edge environment.
3. The full name of logic device is: programmable
Logic
device
that is,
PLD. PLD is proced as a kind of general integrated circuit, and its logic function is determined according to the user's device programming. The general PLD has a high degree of integration, which is enough to meet the needs of general digital system design
in this way, a digital system can be "integrated" into a PLD by the designer himself, without having to ask the chip manufacturer to design and manufacture special integrated circuit chips
the difference between PLD and general digital chip is that the internal digital circuit of PLD can be planned and decided after leaving the factory, and some types of PLD can be changed and changed again after the planning decision. However, general digital chip has decided its internal circuit before leaving the factory, and can not be changed again after leaving the factory. In fact, general analog chip and mixed signal chip are the same, It is impossible to adjust and repair the internal circuit after leaving the factory.
Logic
device
that is,
PLD. PLD is proced as a kind of general integrated circuit, and its logic function is determined according to the user's device programming. The general PLD has a high degree of integration, which is enough to meet the needs of general digital system design
in this way, a digital system can be "integrated" into a PLD by the designer himself, without having to ask the chip manufacturer to design and manufacture special integrated circuit chips
the difference between PLD and general digital chip is that the internal digital circuit of PLD can be planned and decided after leaving the factory, and some types of PLD can be changed and changed again after the planning decision. However, general digital chip has decided its internal circuit before leaving the factory, and can not be changed again after leaving the factory. In fact, general analog chip and mixed signal chip are the same, It is impossible to adjust and repair the internal circuit after leaving the factory.
4. This is the original factory. Now it is a very good place, and the development there is also very good.
5. Inspur information's positioning and boundary in metabrain ecology focus on AI computing platform, which helps left-hand and right-hand partners innovate through advanced AI computing platform. For example, Inspur information's first AI computing system MX-1, which supports OAM acceleration specification, has well supported Cambrian, Suiyuan, Intel, Habana and other leading AI chips at home and abroad. AI chip companies only need to focus on
making chips, and computing power systems can be safely handed over to Inspur. At the same time, Inspur also provides resource scheling capability for intelligent computing platform, optimizes AI resource management and scheling reasonably, and provides integrated proct solution capability from AI training to AI model deployment, and from cloud to edge reasoning model deployment. At present, metabrain ecology has formed nearly 100 leading AI solutions covering finance, energy, transportation, communication and other instries, which are widely applied in the fields of intelligent finance, intelligent transportation, intelligent driving and so on.
making chips, and computing power systems can be safely handed over to Inspur. At the same time, Inspur also provides resource scheling capability for intelligent computing platform, optimizes AI resource management and scheling reasonably, and provides integrated proct solution capability from AI training to AI model deployment, and from cloud to edge reasoning model deployment. At present, metabrain ecology has formed nearly 100 leading AI solutions covering finance, energy, transportation, communication and other instries, which are widely applied in the fields of intelligent finance, intelligent transportation, intelligent driving and so on.
Hot content