( ESNUG 583 Item 6 ) -------------------------------------------- [06/06/18]
Subject: Sawicki and Anirudh on Veloce2, Crystal 3, Palladium Z1, Zebu4
DAC'17 Troublemakers Panel in Austin, TX
Cooley: I want to jump on emulation, because there are two guys here
fighting in the emulation space. Cadence Palladium had a really
good 2016. But, I think it was February or March 2017, Veloce
came out with the Crystal 3 chip and the Quattro box.
Usually, whoever has the new chip has a really fantastic next
year. Those two rivals seesaw back and forth. So, Anirudh,
has your product, has Palladium zenithed this year -- has it
peaked?
Anirudh: First of all, I'm not sure seesaw with the new chip so...
Cooley: It's been doing that for 20 years.
Anirudh: Well I think Cadence has been in a leadership position for a long
time. This is what I see when I look at the data. So, through
several generations of Palladium and emulation platform, we have
always been in the leadership position.
Cooley: Well I'm going back to Pi and Quickturn. It's whoever had the
hot new chip always did well.
Anirudh: I'm coming to the chip, John, I will come to the chip. But I'm
just saying in the last several years, as long I've been
involved, I think Palladium has always had a very strong
position, in the market, and that goes back several years.
And it's not just because of the chip, because in the end we are
selling a system.
This is a true system we are selling, which includes hardware, it
includes software, application layer, compilers, firmware, and I
believe Palladium has the best overall system architecture and
chip architecture to be successful. That's why if you look at
the market, I think we openly say that "9 of the top 10 companies
are using Palladium" and "all the complicated, big designs are
run on Palladium," and we are fortunate to have that position in
the market.
Now coming to the chip. You say Palladium was introduced like
18-24 months ago. We started beta 24 months ago, and officially
introduced 18 months ago. And you say, well there's a later
chip, so it should be better, because of Moore's Law.
But if you really look at the [Chrystal 3] chip, both Palladium
and the new system you mentioned [Veloce2], now have their chip
at the same [28nm] process node.
So, one way to say it is: "You have a new Veloce2 chip."
But maybe another way to say it is: "That Veloce2 chip is
late, and not necessarily new." [audience laughter]
So, it's just how you want to say it.
Cooley: Talk about a backhanded insult.
Anirudh: Well, I'm just trying to state the work is... So [Veloce2] has
a chip at the same [28nm] process node, we believe [Palladium]
has a better overall architecture and we have more use models
and a scalable platform, and we have a huge investment in R&D,
and probably the most talented team in emulation.
Overall, I feel pretty good about our CDNS position.
Cooley: Well, wait a minute. You flipped that around. So you're saying
the old Veloce had an older node [65nm], you have an advanced
node [28nm] and you [Palladium] are barely keeping up with
them is what you're saying. [audience laughter]
Anirudh: No John, what I'm saying is.. how did you get that from what I
said?
Cooley: 'Cuz they just had a new chip -- and you said "they just caught
up to you."
Anirudh: But it's at the same [28nm] process node as ..
Cooley: Now. Now. Joe, jump in.
Sawicki: Sure, there are a lot of things that get conflated in the
emulation market.
When our Veloce2 came out, there were a number of things that
occurred. First off, yeah, that was a new [Crystal 3] chip, for
us going up against what was Palladium 1 at the time.
Probably also as important was also that it was an architecture
that was meant to support the "virtualization test". We could
have protocols, and run on the system.
What that meant is we could put Veloce2 in the customer data
centers, it was air cooled, you didn't need liquid, so Veloce2
really changed the model. We really cleaned the street up in
terms of going into networking companies who were looking
to get significantly higher verification cycles.
We were starting to make some progress in some other companies
when the Palladium II came out, oops that's "Palladium Z1",
and it stopped momentum for us [Mentor] in those other markets.
So, we continued to do well in networking, but lost a lot of
momentum. Yes, we have a new [28nm Crystal 3] chip coming out.
I don't know if really comparing process nodes between the Z1
architecture versus our architectures is particularly relevant.
What's been particularly cool about Veloce2 is it's completely
blowing away our expectations in terms of uptime and reliability
at the customers where we have it.
Emulation has always been one of those things where you worry
about how long your system stays up? What's my availability?
This Veloce2 thing just seems like it's the Eveready bunny.
My expectation is that this is going to put us [Mentor] back
in the emulation game, more momentum in this space. Did we
end up backing up enough market shares to claim victory in
this market this year? That might be a little bit tough with
our delivery schedules. But our machine looks really, really
good.
Cooley: Anirudh?
Anirudh: First of all, let's all remember that emulation is a growing
market. Sometimes we say in the EDA industry there are not
enough growing markets. Emulation is definitely a growing
market -- more and more customers want to use emulation. This is
fabulous for everybody playing in emulation.
Sometimes we just beat each other up, but I think this is one
cyclical trend that is going to help all the players in the
emulation space. But with respect to Palladium Z1, like I
said, we are very confident. We are not seeing any slowdown.
We see further acceleration of Z1 and the other thing is, in
verification, going back to my previous comment, just having one
engine is no longer sufficient.
Just having emulation is no longer sufficient. Sim XL is a big
user of emulation, so you must have ties to a logic simulator,
you must have ties to FPGA prototyping, and to Formal. These
are the 4 key engines. Like what happened in implementation,
one engine itself is not going to cut it.
The job is so complex that you need four state of the art
engines. And for Emulation, we will definitely be state of the
art. But we've invested a lot of energy in making sure all the
other three engines are state of the art. That's where the
ballgame is going to move to: emulation with prototyping,
emulation with simulation and Formal.
Cooley: So, I have a question for both of you. Why is Aart stalling on
launching his Zebu4?
Anirudh: I think Joe wants to take that question. [laughter]
Sawicki: My generous friend. [more laughter]
It's interesting, Zebu is such an odd emulator.
Cooley: Well, it's FPGA based.
Sawicki: Well not ... well, in some ways so are we [Veloce2]. We've
got an architecture that's FPGA based. But with theirs [Zebu],
it's once you get it compiled, and as long as you don't have a
lot of bugs that you're trying to get through, it runs really
fast and it's kind of interesting for when you're solid. But
that's just not what people use emulators for, for the most part.
So, they're [Zebu] in an odd position in terms of what capability
they have that could peak in the market. So, whether that's a
new tool or not a new tool they've [Synopsys] got some handicaps
for the overall play.
Anirudh: The way I would answer the [Zebu4] question is... I looked at
this market carefully... you know I'm new to verification. So,
I look at this market carefully. I looked at emulation and
prototyping. There is room for FPGA platforms. But it's not
in emulation.
So, there isn't room for FPGA platforms to do software
development and quick bring up at a lower cost in some cases.
So, the way to solve the problem, and what we came to realize is
that we need a platform play that includes emulator which is
Palladium and the FPGA product which is Protium, and that's the
way to solve the problem.
Cooley: Alright, commercial time.
Anirudh: Well, you want to hear the truth!
Cooley: Yeah, great. Love them Cadence products yum, yum.
---- ---- ---- ---- ---- ---- ----
Related Articles
Jean-Marie warns Synopsys Zebu needs 9X more "warm bodies" to run
A surly Jean-Marie sasses Cadence Palladium and Synopsys EVE Zebu
SCOOP -- will the new MENT Veloce Crystal 3 chip crush Palladium?
The 14 metrics - plus their gotchas - used to select an emulator
Join
Index
Next->Item
|
|
