( ESNUG 596 Item 2 ) ------------------------------------------------------- [03/21/24]
Subject: Jean-Marie on Siemens' new AI-powered Solido Design Environment launch
The live DAC'23 Troublemakers Panel
Cooley: Jean-Marie. Siemens EDA is launching the Solido Design
Environment against the Cadence Virtuoso ADE monopoly (or
near monopoly.)
Are you effing crazy?
[ audience laughter ]
And before you even get a chance to answer, right now, my
follow-up questions are: How many users do you have?
Are they actually swapping out Virtuoso ADE and using
Solido Design Environment instead? Can circuit designers
do schematic capture in this? Or is it purely a circuit
simulation environment?
Jean-Marie: Yes, yes, yes to all. So, we're not crazy. Customers are
always looking for alternatives.
So, we launched Solido Design Environment today here at DAC'23.
This is a continuation from Solido that we bought in 2017.
Our Solido Variation Designer [from 2017] has been very
successful with its focus on variation analysis, and we have
over a 1,000 users -- at around 100 logos for it.
And what was very good about Solido is that very early [19 years
ago in 2005] on they had real AI waaaaay before everybody started
to talk about AI or ML. And they found a very smart way of
accelerating with a very good accuracy for SPICE simulation.
We launched Solido Design Environment at this DAC'23 -- not to
compete on the layout of Virtuoso aspect -- this is more for
a SPICE simulation. And we are very happy to have this
announcement -- it's now going very well.
We're supporting, of course, Siemens EDA's own BDA SPICE simulators
[AFS, AFS-XT, and Eldo]. We're also supporting Cadence Spectre
and Spectre-X -- as well as Synopsys FineSim, HSPICE --- pretty
much all the simulators.
This is really a next generation SPICE simulation design environment.
Cooley: You're saying you actually run Cadence Spectre-X and Spectre?
Jean-Marie: Yes. We've run pretty much all SPICE simulators out there in Solido
since the beginning.
Cooley: Who's using your product?
Jean-Marie: I think you have the names on the press release. SK Hynix and
all the names in the press release today.
[ EDITOR'S NOTE: Engineers who have the Solido Variation Designer product
automatically now get the new Solido Design Environment.
And here (above) are the 13 logos that I could find in the public who are now
hands-on using the new Solido Design Environment tool. - John ]
Jean-Marie: Many are our customers who are already doing SPICE, std cell,
memory, analog -- those type of customers.
Solido is positioned for the verification analysis, variation,
and SPICE.
That's one of the main reasons we launched Solido Design
Environment -- because we are better in capacity and handling
very large corner cases and the number of simulations.
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
[ EDITOR'S NOTE: Jean-Marie said Solido DE was one of Siemens' big AI products.
And since I haven't covered Solido ML/AI in a while, I decided to take a closer
look at what has changed. This new Solido DE is a "full nominal SPICE simulation
environment" plus Solido's variation-aware stuff. They added an environment.
That is, Solido today uses 4 main flavors of AI.
Amit Gupta's olde Solido has always had Adaptive AI even with his first tool
launch waaaay back in 2007 ...
1. Adaptive AI builds custom predictive AI models during simulation runs,
with the AI training/learning based on the customer's own circuits,
measurements, and PDK's. DeepChip first started getting user comments
on this technology back in 2011 -- but it started in 2007 when they
launched their first tool.
2011: An engineer's 65nm benchmark of Solido PVT+ and Monte Carlo+
Adaptive AI is how Solido gets its 10X to 1000X speedups versus traditional
Monte Carlo brute force -- with the same full SPICE accuracy. They use
AI sampling techniques as part of their algorithms, such as sampling the
tails for high-sigma (3 to 7 sigma) Monte Carlo accuracy.
and Amit's Solido announced Generative AI in 2017 ...
2. Generative AI helps designers create new IP designs or Liberty models based
on their existing data. Solido's Generative AI uses existing IP designs or
existing Liberty models, plus designer requirements to generate either
optimized new IP designs or new Liberty models. Solido started doing
generative AI in 2017 with their Solido Characterization Suite launch.
2017: Amit added 263 users on Library Characterization this year!
Solido ML, BDA Ravi, Tom Beckley SSS makes #2 for Best of 2017
User buzz on Siemens/Solido machine learning is #1 for Best of 2018
Siemens claims its generative AI now makes engineers 10X more efficient.
---- ---- ---- ---- ---- ---- ---- ---- ----
What's different is Amit's other two Solido AI capabilities that are newer
since the Mentor-Solido merger ...
3. Additive AI is a learning technology that automatically builds on existing
AI models and prior design data whenever possible -- meaning designers
don't have to start their SPICE simulation runs from ground zero all over
again to train new AI models. Siemens claims this gives engineers a
further 10X reduction in SPICE simulations.
4. Automatic verification of AI results. With Solido, its AI automatically
acquires more training data until it gets full-SPICE accuracy.
Step One: self-check the AI results for true SPICE accuracy. If the
results are non-convergent (as shown below) ...
Step Two: the Solido tool continues to train its AI model until it
converges -- which verifies it's hit true SPICE accuracy
(as shown below) ...
The model convergence graph is then displayed for designers to do their
own verification. Trusting your final AI results is a big deal because
designers must show the accuracy of their results in their design reviews.
---- ---- ---- ---- ---- ---- ---- ---- ----
Another thing I found out is that Solido DE claims to have pre-built workflows
for library IP verification, advanced node design, automotive, and 3D-IC design.
Future Note: I would like to find out from users if these pre-built flows
actually work and if they actually speed-up your circuit design process. - John]
---- ---- ---- ---- ---- ---- ---- ---- ----
Related Articles
Three MENT Solido vs CDNS ADE-XL user benchmarks is Best of 2019 #6a
Joe Sawicki on ML, Calibre, Solido, VC funding, and heuristics
Siemens Solido Variation Designer invisible ML is Best of 2020 #6a
User buzz on Siemens/Solido machine learning is #1 for Best of 2018
Solido ML, BDA Ravi, Tom Beckley SSS makes #2 for Best of 2017
Solido brainiac rebukes Cadence's doubts of Solido 6-sigma claims
Join
Index
Next->Item
|
