A report citing SemiAnalysis measurements said that Ascend 950PR in 1055 groups of commonly used large-model matrix calculations had an actual speed of approximately H100 of 43%. The report said that DeepSeek is improving efficiency through software adaptation, while large-scale training still faces coordination and stability challenges.
The report says DeepSeek and Huawei are incorporating model-training requirements into next-generation Ascend chip design and plan to test pre-training using Ascend 950DT. The report also says Huawei plans to prioritize supplying Ascend 960DT。
0G announced the launch of its Compute Finance (ComFi) system centered on AI computing power as its core asset; it said Ascend is now live and Infinite AI is scheduled to launch on September 29.
