Increase Computing Power With Advania HPCFlow | Diabatix

February 14, 2021
Lieven Vervecken

In the tech world, thermal design has always been a cornerstone for innovation. Today, that innovation has resulted in size conscious products that deliver more, faster, to maintain their competitive edge. This has resulted in more power being packed in today‘s smaller gadgets, which means that they generate more heat than their predecessors did. The increase in heat generation demands more robust and efficient cooling to ensure product reliability and efficiency.

While thermal design has come a long way in the past years, the typical design process remains a long, trial and error based one, which is both resource and capital intensive.

To address this challenge, Diabatix, a Belgian company, has developed a software that leverages AI to create advanced thermal designs in a fraction of the time. Diabatix’s AI, dubbed ‘X’, is capable of autonomously designing high density heat sinks and cold plates through a generative process similar to topology optimization, but with the added benefit of manufacturability. Diabatix is empowering engineering companies all around the world to ‘not sweat the cooling’ and simply innovate.

“Cooling remains one of today’s biggest roadblocks towards developing innovative tech. We’re bringing the wrecking ball” - Lieven Vervecken, Diabatix CEO

Diabatix’s thermal designs are applied in cases where cooling is absolutely critical, enabling much of the high powered tech that many OEMs and Tier 1 suppliers currently develop. In particular, Diabatix designed heat sinks and cold plates find themselves in e-motors, EV batteries, power electronics, and the like, as they boast powerful performance improvements over conventional thermal solutions.

To power their robust service offering, Diabatix requires great super computing power. Subsequently, as the company is growing fast, flexibility, scalability, and management of that infrastructure was an important factor that Diabatix needed to address in order to ensure operational excellence.

“With artificial intelligence, the computing load can be high, especially when we’re talking about generative design, which is a demanding and fast iterative process. The computing power that our AI needs for its daily design work is equivalent to around 2 000 laptops”, comments Vervecken.

With the constant development of the engineering design algorithm that powers Diabatix’s AI, their service offering continues to grow, resulting in an increased demand for computing power in the years to come. “On top of our staple thermal design services for liquid and air cooling, we have put our virtual engineer to work for conduction and natural convection as well. We won’t be stopping there either”, states Vervecken.

For its computing needs, Diabatix found a great fit in the HPC & AI services provided by Advania Data Centers (ADC). Specifically, in the HPC & AI as a Service offering, HPCFLOW, which enables Diabatix to use a fully managed and supported HPC & AI platform to power their design work and scale as needed, when needed.

“Diabatix is a perfect example of a highly innovative company that benefits from the way we deliver HPC & AI specific computational power.” States Gisli Kr., Chief Commercial Officer of Advania Data Centers. “Their fast growth means that it does not make sense for them to acquire infrastructure to operate themselves, and why should they spend time and effort on operating infrastructure, that would only derail their focus from their mission, to elevate cooling components world wide.”

ADC’s expert HPC & AI engineering teams support Diabatix’s workload with the goal of ensuring operational excellence, resulting in innovation as well as satisfied customers all around.

Follow us on LinkedIn for our latest news.

Get started with ColdStream
Discover how our generative design software will help you during every phase of the cooling design process - from optimizing first designs to virtual testing and detailed analysis.
Book a demo

Continue reading

Continue reading