A research team from National Yang Ming Chiao Tung University (NYCU) has developed a technology that can significantly improve heat dissipation in multicore processors.
Laptops, mobile phones and computer servers commonly have multicore processors, the university said in a statement.
As the number of cores increases, the challenges of connectivity within the network-on-chip (NoC) also rise, it said.
Photo courtesy of National Yang Ming Chiao Tung University
As the core clock speed increases, so do temperatures, which affect operating performance and reliability, it said.
The research team of members of the NYCU Cerebral and Reliable System on Chip Laboratory led by NYCU electrical engineering associate professor Chen Kun-chi (陳坤志) proposed a more cost-effective mechanism that can accurately predict temperatures in NoCs and apply dynamic temperature management via adaptive reinforcement learning technology, greatly improving the performance of a system’s temperature management, the university said.
The paper, “Adaptive Machine Learning-Based Proactive Thermal Management for NoC Systems,” was published last year in the international journal IEEE Transactions on Very Large Scale Integration Systems, the university said.
The team won the journal’s Best Paper Award, the first Taiwanese to receive the honor in the past 30 years, it said.
The machine-learning temperature management technology adopts least-mean-square adaptive filtering theory optimization, which improves temperature prediction accuracy as a processor undertakes different tasks, Chen said.
It uses an adaptive reinforcement learning method to adjust the throttling ratio via real-time feedback from temperature, predicted temperature and system throughput, with the goal to achieve optimal thermal management while maximizing performance, he said.
The research showed that, compared with traditional methods, the adaptive learning method significantly reduces temperature prediction errors and improves system performance, he said.
South Korean K-pop girl group Blackpink are to make Kaohsiung the first stop on their Asia tour when they perform at Kaohsiung National Stadium on Oct. 18 and 19, the event organizer said yesterday. The upcoming performances will also make Blackpink the first girl group ever to perform twice at the stadium. It will be the group’s third visit to Taiwan to stage a concert. The last time Blackpink held a concert in the city was in March 2023. Their first concert in Taiwan was on March 3, 2019, at NTSU Arena (Linkou Arena). The group’s 2022-2023 “Born Pink” tour set a
CPBL players, cheerleaders and officials pose at a news conference in Taipei yesterday announcing the upcoming All-Star Game. This year’s CPBL All-Star Weekend is to be held at the Taipei Dome on July 19 and 20.
The Taiwan High Court yesterday upheld a lower court’s decision that ruled in favor of former president Tsai Ing-wen (蔡英文) regarding the legitimacy of her doctoral degree. The issue surrounding Tsai’s academic credentials was raised by former political talk show host Dennis Peng (彭文正) in a Facebook post in June 2019, when Tsai was seeking re-election. Peng has repeatedly accused Tsai of never completing her doctoral dissertation to get a doctoral degree in law from the London School of Economics and Political Science (LSE) in 1984. He subsequently filed a declaratory action charging that
The Hualien Branch of the High Court today sentenced the main suspect in the 2021 fatal derailment of the Taroko Express to 12 years and six months in jail in the second trial of the suspect for his role in Taiwan’s deadliest train crash. Lee Yi-hsiang (李義祥), the driver of a crane truck that fell onto the tracks and which the the Taiwan Railways Administration's (TRA) train crashed into in an accident that killed 49 people and injured 200, was sentenced to seven years and 10 months in the first trial by the Hualien District Court in 2022. Hoa Van Hao, a
RSS