Legacy CAE tools are too costly, risky, and slow, and legacy CAE does not scale, according to Ian Campbell, CEO of OnScale. His company’s solution is to provide CAE tools using a “solver-as-a-service” (a specialized version of software as a service, or SaaS) subscription model, with companies paying for the CAE capabilities they need, when they need them.
With traditional CAE, Campbell said in a recent phone interview, companies have a fixed supply of software licenses and hardware. Periods during the design cycle that don’t require all the available CAE resources (CAE supply exceeds demand) represent wasted R&D budget, while periods that that could take advantage of resources that aren’t available (CAE demand exceeds supply) represent wasted R&D time.
“Engineers designing next-generation products are experiencing billion-dollar pains because they are using legacy CAE tools that no longer cut it,” he said. A large engineering firm can face costs exceeding $1 million per year, he said, adding that the costs put access to world-class CAE and high-performance-computing (HPC) out of reach for startups.
Risk is also an issue. “If engineers don’t have access to the tools they need, they’re shooting in the dark,” Campbell said. Without adequate tools, they can’t optimize a new design for performance and cost, and they can’t predict potentially show-stopper issues, he added. Consequently, projects can die in the engineering phase, or products that make to manufacturing can end up being obsolete on day 1.
Finally, he cited time. Engineers have too little of it, and they don’t want to waste what they do have on their CAE tools. He noted that when he started force-sensing solutions provider NextInput in 2012, “…the design cycle for a typical new IoT sensor was 18 months. Now, many leading companies introduce new products on a six-month rotating basis.”
With its SaaS model, OnScale provides access to multidomain, RAM-efficient, MPI- and HPC-compatible solvers hosted on the AWS scalable HPC platform. Target customers, Campbell said, include biomedical device OEMs, ADAS device OEMs, IoT device OEMs, and 5G RF device OEMs, with each OEM segment representing a $1 billion TAM. The company offers fully coupled transient multiphysics capabilities for thermal, piezoelectric, structural, electrostatic, and acoustic applications, and it supports anisotropic, nonlinear, erosion, plasticity, and time-dependent advanced materials models. The approach, he said, offers “limitless” compute power with fully secure end-to-end encryption, with AWS handling security. OnScale, he said, uses 100% AWS datacenters in each customer’s country.
To use the OnScale approach, he explained, you create models using a local copy of OnScaleLab or import models from a MATLAB library, for example. You then run simulations in the OnScale Cloud. Finally, you download and extract engineering results for analysis and visualization.
OnScale charges customers based on the actual solver time they use, measured in core-hours (using five processor cores for 10 hours would equal 50 core hours). OnScale Cloud is offered in three monthly subscription levels:
- Free: Any engineer can begin using OnScale for free and receive 10 core-hours per month without commitment. Ten core-hours is sufficient to perform many simulations of simple devices. Additional on-demand core-hours can be purchased with a credit card for $10 per core-hour.
- Professional: The Professional OnScale Cloud subscription is $300 per month and includes 50 core-hours, which opens up simulation of more complicated designs and parametric design studies. Additional on-demand core-hours can be purchased for $9 per core-hour at the Professional subscription level.
- Team: The Team OnScale Cloud subscription is $1,000 per month and includes 200 core-hours, enough for a small engineering team to optimize next-generation devices. Additional on-demand core-hours can be purchased for $7 per core-hour at the Team subscription level.
In addition to these subscription levels, OnScale also provides discounts for annual subscriptions and flexible subscription programs for small, medium, and large enterprises.
Campbell presented several examples, including a 5-DoF simple mechanical model and a 20-DoF 33-cell MEMS PMUT array, for which he said the OnScale Cloud approach outperformed traditional CAE systems.
He concluded by noting that OnScale’s advanced solvers enable huge MDoF models and passive parallel optimization studies and stand poised to drive a new wave of innovation in 5G, IoT, and other applications.
See also “OnScale Cloud solver-as-a-service now generally available.”
