AMD and Sailingsai created a new AI inference world record

Why fill in in California emperorSai LingsaiOn developer forum, our CEO Victor Peng andAMD CTO Mark Papermaster greeted “ jointly Jinisi ” . Do not want to misunderstand, this is not to hold the beer in bottle, write in the world record that this getting on however.

 

AMD and Sailingsai are announced, processor of dragon of clouds of both general AMD and brand-new Sai Lingsai Alveo blocks union quickly, in order to realize inference of AI of high-powered real time. In the meantime, they announced record of a world: Every second 30, the inferential handling capacity of 000 pieces of pictures.

 

The system with this colourful Jing making a person will appear today in the Alveo ecosystem a group of people of same interest of forum, it was used have an industry to precede PCIe joins dragon of 2 AMD clouds 7551 processor, and 8 brand-new issuance Sai Lingsai Alveo U250 gets stuck quickly. Inferential process thinks of machine study to cover a drive by contest spirit, the person that should cover a get out of the way to send can optimize the inference that deploy quickens, support those who include TensorFlow inside a variety of machine study frameworks. This test reachs on GoogLeNet as a result () of a data set moved below Int8 precision, this is current use extensively roll accumulate nerve network (abbreviate is CNN or ConvNet) .

 

Sai Lingsai

AMD and Sailingsai are being calculated have on gradual progress and framework of different compose system wish jointly scene, and have long-term technology cooperate. Both sides is aimed at processor of AMD clouds dragon and Sai Lingsai FPGA to optimize drive jointly, debug each other to operate a gender. In the meantime, both sides also is mixed in the cache consistency of accelerator with the other partner inside the industry interrelate the respect undertakes cooperative (CCIX alliance is dedicated at multiprocessing implement the cache consistency between and memory share) .

 

Processor of AMD clouds dragon is the perfect platform that handles load of artificial intelligence and high-powered computation work quickly. Clouds dragon processor has the passageway of 8 Cheng of 32 core, 64 lines, memory, memory that every insert groove to be as high as 2TB, 128 PCIe passageway, and of the class of hardware of X86 server ground floor of the initiate inside course of study embedded data security, remarkable memory ability, bandwidth and much core handle common in AI and high-powered computation memory effectively concentrated model the job is laden. Had clouds dragon processor, client can more apace is collected and analyse many data set, help them accelerate the processing speed of complex issue significantly.

 

     Sai Lingsai and AMD very the future that values bilateral technology collaboration. Bilateral product route chart matchs each other, processor of high-powered AMD clouds dragon and graphical processor and Sai Lingsai quicken platform, include Alveo to quicken card and coming Versal product combination can perfect cooperate.

 

Accordingly, let us raise Jinisi beer, beatific AI inference and the happiness that platform of different form calculation innovates did not come.

Add a Comment

Your email address will not be published. Required fields are marked *