Silvaco Nanometer Newsbyte 
Graham Bell
Graham is Sr. Director of Marketing at Silvaco Inc. An experienced semiconductor-design marketing strategist and avid blogger before joining Silvaco, Graham previously was VP of marketing at Uniquify, a fabless SoC product and IP company, and at Real Intent, a verification software company. In his … More » Machine Learning in Silvaco EDA SoftwareAugust 1st, 2019 by Graham Bell
In the following video, Dr. Firas Mohamed, VP & GM, Machine Learning & Flow Optimization Division and GM, Silvaco France talks with Graham Bell about Machine Learning technologies deployed in Silvaco EDA tools at the SEMICON West 2019, July 9 – 11 at Moscone Center in San Francisco. A transcript of the video is also below.
Hi, this is Graham Bell with Silvaco. I'm speaking with Firas Mohamed, Vice President of Machine Learning Technologies and General Manager of Silvaco France. Hi Firas. FM: Hi Graham, nice to be with you today. GB: Firas, off camera, we were talking a little bit about some of the different applications for machine learning. What's one application that you would like to talk about with our audience? FM: So I would like to talk about our first solution which is called TechModeler and TechModeler is really based on advanced machine learning algorithms and dedicated to what to what we call compact modeling which is the modeling of transistors and specially of emerging technologies like O-TFT or organic electronics. In these technologies modeling time may take years. And with TechModeler, this can be really be reduced to days or Hours. GB: What are some of the other technologies that have been applied in our tools? FM: So I'd like to talk about VarMan. VarMan is based on an advanced statistical algorithm mixed with also a unique machine learning technology. And with that, to verify all these design applications at advanced nodes where variations are impacting the design and especially the yield and may need what you call statistical verification based on Monte Carlo analysis where they need to do a thousand, hundred-thousand or even billions of runs and this cannot be possible [or practical]. VarMan based on this machine learning algorithm is able to drastically reduce these analysis times or even number of runs to hundreds or a few thousand instead of billions of runs. This is done while preserving high accuracy and VarMan is really today proven on the most advanced nodes and technology down to 7 nanometers. GB: Firas, thanks for speaking with us today, FM: Thank you Graham. It was a pleasure. Category: Silvaco |
|
|
|||||
|
|
|||||
|
|||||
