Latest chips publications
UK chipmaker Arm is migrating the vast majority of its workloads, including those used to test semiconductor designs, to Amazon Web Services, bringing about a dramatic reduction of its data centre footprint.
By switching workloads from its own infrastructure to instances hosted in Amazon’s hyperscale data centres, the semiconductor giant will reduce its on-premises infrastructure by 45 percent, and slash on-premises compute to a fifth of previous levels.
The Government is examining the sale of UK-based computer chip design firm Arm to US tech giant Nvidia, with Boris Johnson taking a personal interest in the deal, his spokesman said.
Earlier on Monday, it was announced that the firm is to be bought by the US graphics chip maker in a deal worth 40 billion dollars (£31.2 billion).
After five years of research and development, IBM has taken the wraps off its first 7nm processor, which is packed with features to cater to the demands of modern-day hybrid cloud computing.
The tiny chip is the tenth iteration of IBM’s flagship Power processor range and is destined for enterprise data centres by the second half of 2021.
7-nanometre fabrication is the cutting edge of bitesize chip manufacturing and affords major gains in power efficiency and performance compared to the 14-nanometre process, like that used for IBM’s current-generation Power9 semiconductor. IBM said Power10 will deliver a three-fold improvement in capacity and processor energy efficiency compared to Power9.
Cisco has entered the silicon routing space with the launch of what the company claims is the industry’s first universally adaptable networking chip architecture.
The networking giant said the new single-chip architecture, dubbed Silicon One, will handle data far better than processors currently found in networking and switching gear, as well as providing the versatility needed to support complex new technologies like 10G and adaptive cyber security.
Nvidia and Google are producing some of the world’s fastest hardware for training machine learning models The latest AI hardware from Nvidia and Google have set new records for the time taken to train machine learning models, according to new AI benchmarking results from the MLPerf consortium. MLPerf is a benchmarking suite that measures the… Read More