Advanced Computing in the Age of AI | Friday, March 29, 2024

HP Updates On Project Moonshot, ARM-Based Servers, Memristors 

<p><span style="font-family: verdana, geneva;">At the Server Design Summit, HP Labs researcher gives some status reports on several projects. 100-fold increase in performance per watt? Announcements imminent for Project Moonshot and ARM-based servers? …</span></p>

Hewlett-Packard Co. will reportedly make some announcements about Moonshot and its work on ARM- and Atom-based servers “within weeks.” It's working with AMD, Applied Micro, Calxeda, Cavium and Intel on the developments.

It's also making progress on the development of memristors as alternatives to server CPUs, and the technology will step in as a major force as “disruptive changes” arrive in datacenter technologies.

Those were some of the topics that HP Labs research Parthasarathy Ranganathan discussed with EEtimes reporter Rick Merritt at the Server Design Summit in San Jose. The article indicates that energy efficiency weighs heavily on HP's current agenda.

Moonshot is HP's project to create extreme low-energy server technologies. According to HP's Moonshot website, the company expects its efforts to result in 89% less energy consumption, take up 94% less space, and reduce overall costs by up to 63%.

HP has already announced an Atom-based server. In 2013, it should offer cartridges that will upgrade a server chassis for many different ARM and Atom chips.

Within three to five years, it may introduce “nanostores,” chips that combine memristors and logic to challenge the microprocessor hegemony, Ranganathan said. A memristor, as the name implies, can both maintain a safe flow of current and remember charges even when power is lost, allowing it to act as computer memory. The nanostores
are 3-D stacks that could achieve “potentially a hundred-fold increase in performance per watt,” said Ranganathan.

These improvements are needed for several reasons: Data growth is outpacing Moore's Law, networking is shifting to optical technology, and data access is increasing in complexity, he noted. Data is not only big, it's faster, is coming from many streams, and is scrutinized with deep analytics.

EnterpriseAI