( ESNUG 600 Item 2 ) ------------------------------------------------------- [02/27/26]
Subject: Real Intent speeds up EDA algorithm while CDNS/SNPS/MENT formal goes AI
Subject: CDNS bigwig surprise endorses Real Intent tools into Cadence AI flow
The live DAC'25 Troublemakers Panel
Cooley: Prakash.
Prakash: Yes, John.
Cooley: Are you, Real Intent, in trouble as a company?
Because he [Cooley points to Paul Cunningham of Cadence] has a "prison"
AI system; and he [Cooley points to Ravi Subramanian of Synopsys] has a
"prison" AI system; and he [Cooley points to Amit Gupta of Siemens EDA],
sort of doesn't have a viable AI system yet -- he's just announced it.
So, your Real Intent tool is outside of that -- while CDNS/SNPS/MENT
is getting all sorts of benefits from AI -- and your Real Intent is
not included in the AI party ????
Paul: Aren't we [CDNS & Real Intent] collaborating?
[Paul looks at Prakash] We are, right?
Prakash: [Prakash looks at Paul] We can. We will.
Paul: But we're in prison together, I guess. [laughter]
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
Prakash: Well, first of all. We [Real Intent] are a smaller company, John.
We have fewer products.
I cannot aspire to build an AI prison -- perhaps a lockup cell?
[laughter]
Cooley: But seriously, Prakash, it's a real problem for your Real Intent...
Prakash: I'll build upon a comment that Paul had made -- that AI has to
necessarily fuse with the underlying EDA technology.
The opportunity for innovation is now at the workflow level using AI.
But Real Intent -- a verification company -- has a tremendous opportunity
for innovation in the core EDA technology space -- and that's where
our focus is at this DAC.
We are talking about minimally Boolean verification approaches.
What do I mean by that?
So, if you think about functional verification, the work horses have
been formal verification itself -- actually simulation is the main work
horse -- and along with formal verification. If you think about it,
they are all built upon an engine.
These engines are Boolean engines.
The simulation engine is a Boolean engine and the formal engine is
a Boolean engine. And then our EDA application is customizing the
engine to do whatever the EDA application wants the engine to do.
So, they are locked up in these Boolean shackles -- both EDA applications
are locked in those Boolean shackles.
Our new algorithms at Real Intent breaks those Boolean shackles.
Cooley: Isn't that's just a fancy way of saying CDNS Jasper is doing Boolean and
you at Real Intent are not?
Prakash: So, what we at Real Intent do is build customized EDA applications which
allows us to use abstract techniques which are minimally Boolean -- which
gives us a tremendous runtime and capacity advantage.
So, for example, Sentry, our hardware security tool can do 100 million
gates in a couple of hours.
---- ---- ---- ---- ---- ---- ---- ---- ----
[ EDITOR'S NOTE: To put this in context, formal analysis tools like CDNS Jasper,
SNPS VC Formal, Siemens OneSpin are all maximally Boolean which makes them
super accurate -- but at the price of being capacity limited to 2-to-3 million
gates and can take up to 12 months to formally "prove" those gates are error
free (with lots of engineering hand-holding by the formal tool operator)...
... while minimally Boolean Real Intent Sentry using static analysis can verify
100 million gates in 5 hours and is equally accurate, the problem is like formal
it's noisy with false positives, so you need to add in 3 weeks for an engineer
to do reviews and waviers.
Given 2-to-3 million gates in 12 months or 100 million gates in 3 weeks, it's
obvious why tiny Real Intent makes sales against CDNS/SNPS/MENT. - John ]
---- ---- ---- ---- ---- ---- ---- ---- ----
Cooley: Right.
Prakash: You don't have a prayer for analyzing something like 100 million gates in
a couple of hours with [CDNS/SNPS/MENT] formal tools.
And on the flip side these formal [CDNS/SNPS/MENT] EDA applications have
to be custom EDA applications. So, they [CDNS/SNPS/MENT] and Real Intent
have to keep adding specific EDA applications to lint, CDC, RDC ... We
have DFT, we have connectivity checking, we have ...
Cooley: Jasper has that, Jasper has that, Jasper has that, Jasper has that...
Prakash: Yeah, then the differentiating factor is what can be accomplished with
our minimally Boolean, minimally formal approach at Real Intent.
We have a very different EDA core, a completely different core technology
than what Cadence/Synopsys/Siemens has ....
Cooley: Right.
Prakash: It's maximally Boolean versus minimally Boolean. And that results in a
Real Intent runtime & capacity advantages as I mentioned -- like doing
HW security analysis on 100 million gate design in a couple of hours.
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ---- ---- ---- ---- ---- ----
Subject: CDNS bigwig surprise endorses Real Intent tools into Cadence AI flow
Cooley: I get it.
So ... Paul [Cunningham]... does your CDNS Jasper manager
talk to you at all? ... Does he know that you're doing stuff with
Real Intent AI wise?
Paul: Yes.
Cooley: How come he's not chewing you out? He's the Jasper manager!!!?
Paul: Per the point, our Cadence AI is not actually with a prison mindset.
In the end it's about making our EDA customers successful. So, I
have a Jasper CDC solution. I have a Jasper lint solution, right?
Cooley: Right, right.
Paul: But also, you know, several of my most important and respected
customers see real value in what Real Intent is offering -- and I'll
happily partner with Prakash to make sure that his technology can
work with our CDNS simulator and our formal Jasper tool for those
customers.
Cooley: Wow.
Paul: Competition is a good thing. I'm not going to stop investing in
really trying to differentiate my own Jasper CDC and Jasper lint
solution... but I see nobody wins by my CDNS not collaborating
also with everything that Prakash has... he has enough headwinds
of his own. I've done my own startup too, okay.
I think it's phenomenal what Prakash is doing at Real Intent.
Cooley: So, Paul,... you have a "liberal" AI prison?
Paul: I guess ... well ... or maybe our CNDS AI is just not a prison!
Cooley: Right. Right.
Prakash: And, John, very specifically to say what I mean is what Paul
elaborated. I think we just recently released the capability
for linking our Real Intent Meridian CDC to directly simulate
in the Cadence Incisive simulator.
Cooley: But if you're allowed in something in the Cadence AI flow; that's
a very powerful thing.
Prakash: We work with Cadence very closely to make that happen.
Paul: Yes.
Cooley: Ravi, is your SNPS AI prison liberal? Or conservative?
Is Real Intent allowed into your SNPS AI flow or not?
Ravi: And so, the answer is yes if it's for a SNPS AI customer.
I'll back it if a SNPS AI customer is actually asking us
to do that.
Cooley: Do you have a SNPS AI customer actually asking you to bring
Real Intent in?
Ravi: No comment.
But at the end of the day, I think the model of "prison" is one thing;
the model of country club is another.
Ultimately anybody [3rd party EDA vendor] can become a member in our
SNPS AI flow -- if there's an EDA customer who is driving a high-value
problem that says "I need these EDA tools to work together so I can
achieve this solution for this high value problem."
You can call it whatever you want -- prison, igloo, country club,
that's the crux of it.
Cooley: Igloo? I've not heard of that one. Okay all right, igloo.
Ravi: You need to go up to Saskatchewan, okay.
Amit: We know igloos.
Ravi: I've been to igloos
Amit: Ravi came up to visit us in Saskatchewan.
Cooley: [Amit] so how about Real Intent in your AI flow, yes or no?
Amit: Again, our Siemens AI flows are completely open. I don't know
specifically if Real Intent's in our flow?
I don't know specifically.
Prakash: We [Real Intent and Siemens AI] have a good working relationship.
Amit: Okay, good working relationship... we're open.
Cooley: OK, all right. So, "liberal" (Cadence), "no comment" (Synopsys),
and "we're nice to each other" (Siemens).
Amit: We're Canadian.
[Cooley laughs]
---- ---- ---- ---- ---- ---- ---- ---- ----
Related Articles
Real Intent Sentry HW security sign-off is 12x faster vs. Jasper formal
Real Intent low noise/multimode Meridian RDC gets the Best of 2020 #3a
Users choosing Meridian CDC over Spyglass CDC gets the Best of 2020 #3b
Real Intent Verix CDC true multimode analysis gets the Best of 2020 #3c
Ascent Lint "designer intent" and pre-submit gets the Best of EDA #3d
Real Intent smacks Synopsys CDC & RDC signoff as #3 "Best of 2018"
Real Intent trounces Synopsys Atrenta as the #6 "Best of" for 2017
Real Intent caught launching a "true" CDC linter under old name
Join
Index
Next->Item
|
|
