Lifeboat Foundation

AI processing can take a huge amount of computing power, but by the looks of this latest joint project from the Jülich Supercomputing Center and French computing provider Eviden, power will not be in short supply.

“But can it run Crysis” is an old gag, but I’m still going to see if I get away with it.

China Telecom claims it has built the country’s first supercomputer constructed entirely with Chinese-made components and technology (via ITHome). Based in Wuhan, the Central Intelligent Computing Center supercomputer is reportedly built for AI and can train large language models (LLM) with trillions of parameters. Although China has built supercomputers with domestic hardware and software before, going entirely domestic is a new milestone for the country’s tech industry.

Exact details on the Central Intelligent Computing Center are scarce. What’s clear so far: The supercomputer is purportedly made with only Chinese parts; it can train AI models with trillions of parameters; and it uses liquid cooling. It’s unclear exactly how much performance the supercomputer has. A five-exaflop figure is mentioned in ITHome’s report, but to our eyes it seems that the publication was talking about the total computational power of China Telecom’s supercomputers, and not just this one.

“The memory requirements for PRIYA simulations are so big you cannot put them on anything other than a supercomputer,” Bird said.

TACC awarded Bird a Leadership Resource Allocation on the Frontera supercomputer. Additionally, analysis computations were performed using the resources of the UC Riverside High-Performance Computer Cluster.

The PRIYA simulations on Frontera are some of the largest cosmological simulations yet made, needing over 100,000 core-hours to simulate a system of 3072³ (about 29 billion) particles in a ‘box’ 120 megaparsecs on edge, or about 3.91 million light-years across. PRIYA simulations consumed over 600,000 node hours on Frontera.

Tesla is set to invest $500 million into building a Dojo supercomputer at its New York Gigafactory, as announced this week.

Tesla is gearing up to build its next-generation Dojo supercomputer at its Gigafactory in Buffalo, New York, as part of a $500 million investment announced by the state’s governor on Friday.

The Dojo supercomputer is designed to process massive amounts of data from Tesla’s vehicles and train its artificial intelligence (AI) systems for autonomous driving and other applications. It is expected to be one of the most powerful computing clusters in the world, surpassing the current leader, NVIDIA.

Cryptocurrency is usually “mined” through the blockchain by asking a computer to perform a complicated mathematical problem in exchange for tokens of cryptocurrency. But in research appearing in the journal Chem a team of chemists has repurposed this process, asking computers to instead generate the largest network ever created of chemical reactions which may have given rise to prebiotic molecules on early Earth.

This work indicates that at least some primitive forms of metabolism might have emerged without the involvement of enzymes, and it shows the potential to use blockchain to solve problems outside the financial sector that would otherwise require the use of expensive, hard to access supercomputers.

“At this point we can say we exhaustively looked for every possible combination of chemical reactivity that scientists believe to had been operative on primitive Earth,” says senior author Bartosz A. Grzybowski of the Korea Institute for Basic Science and the Polish Academy of Sciences.

Give people a barrier, and at some point they are bound to smash through. Chuck Yeager broke the sound barrier in 1947. Yuri Gagarin burst into orbit for the first manned spaceflight in 1961. The Human Genome Project finished cracking the genetic code in 2003. And we can add one more barrier to humanity’s trophy case: the exascale barrier.

The exascale barrier represents the challenge of achieving exascale-level computing, which has long been considered the benchmark for high performance. To reach that level, however, a computer needs to perform a quintillion calculations per second. You can think of a quintillion as a million trillion, a billion billion, or a million million millions. Whichever you choose, it’s an incomprehensibly large number of calculations.

Continue reading “Supercomputer uses machine learning to set new speed record” »

New advancements in technology frequently necessitate the development of novel materials – and thanks to supercomputers and advanced simulations, researchers can bypass the time-consuming and often inefficient process of trial-and-error.

The Materials Project, an open-access database founded at the Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) in 2011, computes the properties of both known and predicted materials. Researchers can focus on promising materials for future technologies – think lighter alloys that improve fuel economy in cars, more efficient solar cells to boost renewable energy, or faster transistors for the next generation of computers.

Artificial intelligence has progressed from sci-fi fantasy to mainstream reality. AI now powers online tools from search engines to voice assistants and it is used in everything from medical imaging analysis to autonomous vehicles. But the advance of AI will soon collide with another pressing issue: energy consumption.

Much like cryptocurrencies today, AI risks becoming a target for criticism and regulation based on its high electricity appetite. Partisans are forming into camps, with AI optimists extolling continued progress through more compute power, while pessimists are beginning to portray AI power usage as wasteful and even dangerous. Attacks echo those leveled at crypto mining in recent years. Undoubtedly, there will be further efforts to choke off AI innovation by cutting its energy supply.

The pessimists raise some valid points. Developing ever-more capable AI does require vast computing resources. For example, the amount of compute used to train OpenAI’s ChatGPT-3 reportedly equaled 800 petaflops of processing power—on par with the 20 most powerful supercomputers in the world combined. Similarly, ChatGPT receives somewhere on the order of hundreds of millions of queries each day. Estimates suggest that the electricity required to respond to all these queries might be around 1 GWh daily, enough to power the daily energy consumption of about 33,000 U.S. households. Demand is expected to further increase in the future.