On April 14th, Tencent Cloud launched a brand new technology of High-Performance Computing Cluster aimed toward large-scale mannequin coaching. The total efficiency has been improved 3 times in comparison with the previous.
This cluster makes use of Tencent Cloud’s self-developed servers, geared up with the newest technology of NVIDIA H800 GPUs. The servers are linked by 3.2T interconnect bandwidth, offering high-performance, high-bandwidth and low-latency cluster computing energy for large-scale mannequin coaching, autonomous driving, scientific computing and different functions.
In October final yr, Tencent accomplished the primary trillion-parameter AI mannequin coaching – the Hunyuan NLP mannequin. The coaching time was diminished from 50 days to 11 days. If primarily based on the brand new technology cluster, the coaching time shall be additional shortened to 4 days.
SEE ALSO: Tencent Cloud and Singapore-Based Strange Universe to Develop Web3 Business Metaverse Platform
Currently, Tencent Cloud’s distributed cloud-native scheduling has a complete scale of over 150 million cores and offers 16 EFLOPS of clever computing energy.
Tencent Cloud’s new technology cluster will serve not solely large-scale mannequin coaching but in addition be utilized in numerous eventualities, together with autonomous driving, scientific computing, and pure language processing.
Sign up right now for five free articles month-to-month!
…. to be continued
Read the Original Article
Copyright for syndicated content material belongs to the linked Source : Pandaily – https://pandaily.com/tencent-cloud-launches-new-high-performance-computing-cluster/