In October 2020, Nvidia revealed plans to begin developing Cambridge-1, UK’s fastest supercomputer. The project would cost an estimated £40 million, or $51.7 million. As we all know, the COVID-19 took a stronghold in 2020. With the global pandemic still in full effect, Nvidia was head-on with a list of likely challenges. For one, the ability to remotely manage the implementation of a supercomputer from the other side of the Atlantic. So far, only 20 weeks after it was first revealed, the Cambridge-1 entered its first stages of operation. The time frame of building the supercomputer has never been done, let alone during a global health crisis. In comparison, most supercomputers that are currently on the Top500 list took a couple of years from concept planning to the final build. Cambridge-1 took only five months.
Currently operating out of one of data center provider Kao Data’s buildings in Cambridge, the supercomputer is enduring final tests before scientists can start using it to focus on healthcare research. The system will feature 80 Nvidia DGX A100 systems combined, for 400 petaflops of ‘AI compute,’ or eight petaflops of standard Linpack performance. That puts it 29th on the Top500 list of the world’s most powerful supercomputers and among the top three most-energy-efficient machines in the Green500.
Meet Summit, the IBM Supercomputer.
How was Cambridge-1 completed so quickly?
When Cambridge-1 was first revealed, data center host Kao Data was already more than six months into applying strict operational COVID-19 measures. Implementing these safety measures ended up working in their favor as the fewer people there who are not necessary, the more effectively the job can get done. The small team on the ground, worked in collaboration with the Nvidia remote team, achieving the result that was required. This eliminated the need to physically send Nvidia employees to Cambridge. Instead, Nvidia engineers used a technique called computational fluid dynamics to accurately measure their allotted space in Kao Data’s building. Once their space was correctly modeled, Nvidia was able to decide where they wanted to place the servers and computer racks that constitute the building blocks of the supercomputer.
Cambridge-1 was designed to expand across three rooms in the data center, all fitted with separate power and HVAC systems. Each room is outfitted with two rows of 12 refrigerator-sized racks, and thousands of fiber optic cables connecting the systems, set up like horizontal ladders on top of the racks. To keep a closer eye on the supercomputer project, Nvidia settled on basic telepresence robots – think an iPad on wheels – to patrol the facility checking out faults. The same robot had been used in the past during the production of another one of Nvidia’s supercomputers, Selene. Selene is similar to Cambridge-1 in that it extends across several rooms, but the difference is there was always an Nvidia employee on site for the build.
Nvidia Takes Shots at Crypto Miners – Limits GPU Hash Rates
Technology has allowed us to do more than ever before from a remote location. However, building a supercomputer is deemed to be a tough task to do remotely. Cambridge-1 had to be physically put together by a team that was being directed from afar. The refrigerator-sized racks that make up a supercomputer consist of smaller computers, each of which has ten fiber optic cables sticking out. Bundles of hundreds of fiber optic cables were connected, pre-packaged, and shipped to the data center, where engineers were tasked with plugging one end into the servers and another into the network switches. The easy plug-and-play method was the key to building Cambridge-1 at such an alarming rate.
By the time Nvidia holds its annual GPU Technology Conference in mid-April, the first research results are expected from some of the early projects run on Cambridge-1. Nvidia has already announced partnerships with four healthcare organizations, which are set to be granted access to the device for medical research. With the extra computational power enabled by the supercomputer, scientists will be able to solve data-based problems that were previously difficult to harness, such as better diagnosing patients and identifying appropriate treatments. They are hopeful that the device can lead to breakthroughs in medical research and possible new drug discoveries.
