Days to Seconds

For today’s design challenges the use of parallel computing to accelerate simulation performance is a must. I ask myself what would be a dream scenario for circuit design decision makers - what-if you could predict circuit performance at the click of a mouse? Imagine the opportunities, maybe a new customer interaction model will emerge, perhaps a new cost model will appear, possibly a new delivery model will arise, perhaps a new design methodology - the possibilities are endless. But is this dream realistically achievable?

What if?

What-if we have a new companion to the existing super-fast simulators that makes this dream a reality and gives you a new and unique lens into design decision making? Simulation technology has already achieved significant performance improvements over the last decade. This is being further accelerated by the use of cloud computing to run thousands of simulation tasks in parallel. However, getting results at the click of a mouse remains elusive. What new companion technology can be added to fast simulation technology and cloud computing to change the landscape? Results show that supplementing the existing simulators with neural network technology makes the dream a reality.

Is the technology available yet?

The question then is has neural network technology matured enough to step up to the task? My answer to this question is yes. My colleague Muhammad Kasim’s blog discusses a case where Machine Discovery’s neural network technology predicted the output of more than 250 simulation runs in less than a second using a single GPU, versus an original cost of 1100 CPU hours per simulation run - 1 CPU sec vs 1100 CPU hours. With this technology combination the immediate question changes from ‘Find out later what happens should I make these changes?’ to ‘Find out now what happens if I make these changes?’

One can now use existing super-fast simulators to analyse the design accurately and use cloud computing to run hundreds of simulations in parallel to collect the data needed to train the neural network to then predict the circuit performance with high accuracy at the click of a mouse.

So what new horizon changes are now possible?

Having instant prediction capability on top of simulation brings new opportunities to product development leadership who seek to implement transformations that differentiate their companies’ offerings. Effectively, the engineering design team receives more data much sooner, which they can then apply to the task of making the next design exploration choices.

In my experience the impact of these seismic, several orders of magnitude speed improvements is not just getting new data sooner; it is primarily the new differentiation options made viable due to expanding the scope and variety of outcomes available to be assessed. Such a big quantitative change leads to a qualitative change - you are not just increasing the speed at which you get feedback on changes, you can now do things differently and do different things.

In summary

This is a perfect storm - a combination of advanced simulation technology, cloud computing and neural network technology to make the click of a mouse dream a reality.

Disruptive new design methodologies and customer engagement models are made possible by approaching problem solving differently – all because waiting for reliable and informative results is no longer the bottleneck. Indeed, a perfect storm is in the making.

This provides a brand new lens due to being able to slash the risk in data driven decision making thus enabling the realisation of the imagined future and thus delivering a step change on achievable ROI levels.