( ESNUG 589 Item 01 ) --------------------------------------------- [04/07/21]
Subject: Anirudh on Innovus crushing Aart's ICC2 & Fusion Compiler at 5/3/2nm
Virtual DAC'20 Troublemakers Panel
Cooley: Anirudh, you've been lucky with Cadence. Basically, Aart's
ICC Compiler II PnR fell down -- and your Innovus PnR got to
jump into its space.
But in my report this year I saw Synopsys users starting to chime
in about Fusion Compiler. Fusion Compiler is not falling down
the same way. (See DAC'19 #9)
Is Cadence still going to be able to compete?
Anirudh: John, I think what I see in the marketplace is that we are doing
pretty well with PnR at advanced nodes. It started at 16nm, it
accelerated at 10nm and 7nm, and it's going further at 5nm and
3nm. Most of our PnR R&D work now is at 3nm and 2nm.
We are pretty confident in our PnR position. If you look at these
advanced geometries, Cadence is providing the best PPA. So, there
are several things happening in that market...
Cooley: So you're saying point blank -- Innovus is beating Fusion Compiler
at 5nm and 3nm?
Anirudh: Yeah, still most of the top customers' advanced nodes are using
Innovus as their place and route solution.
What happens though -- if you look at one particular company (and
let's say it's an advanced node customer doing 7nm or 5nm design
... Innovus -- it depends on the customer -- maybe let's say
is 70-80% of their PnR usage, and maybe the other PnR tool [SNPS]
is 20-30% of their usage.
So, for those designs, or that usage, of course the customers
will upgrade to the latest version of the other Synopsys tool
[Fusion Compiler] -- but that doesn't mean that the usage of
Innovus or the Cadence solution is any different from before.
Whereas what I do see in the marketplace is that -- and this is
well known -- we started with Innovus. And I don't consider ...
I'm happy to take luck ... but we were prepared, and we were a
better PnR tool. So that's why Innovus won.
And what's happening now is that we started with place and route
but, we built the solution several years ago to be fully
integrated (because at lower nodes I always knew that synthesis
place and route and sign-off will get integrated) and that's what
we see now.
So, what I see now is that with most of our advanced node
customers, there's rapid deployment of Genus, which is our
synthesis tool, and Tempus, which is our sign-off tool.
To give you some numbers, out of the top 20 companies,
Innovus PnR is used in about 18 of the top 20. And Genus RTL
synthesis is now used in 15 of the top 20 companies.
Cooley: Genus' RTL synthesis is in 15 of the top 20?
Anirudh: Yes, 15 of the top 20. Tempus sign-off is used in 11 of the
top 20. And Tempus ECO, which is like a timing convergence tool
that works with Innovus, is used in 16 out of the top 20.
What I see in the marketplace is not only is Innovus maintaining
its lead, we are extending our usage to synthesis and timing
sign-off -- which is what I always intended and which is what we
knew at advanced nodes the market would go to for a full solution.
We [at Cadence] are pretty happy with that.
So that's one activity that I see using more of a full flow of
Cadence synthesis, place & route, and timing sign-off. And we
continue to work with our customers in that respect.
Cooley: How does that differ from what Aart is offering [in Synopsys]
with Fusion Compiler? They have tools in all those spaces.
Anirudh: Absolutely. Synopsys has done well historically in that market
anyway. If you go back 20 years, they started with SNPS Design
Compiler synthesis.
But I'm saying that advanced nodes, we at Cadence are doing
pretty well, and sometimes, even at advanced nodes, the
customer will have two sets of tools [both SNPS & CDNS]. So
that doesn't mean there is no usage of the other tool.
Cooley: Yeah, you're right. There's a lot of overlap on that.
Anirudh: Yeah, it's just that we at Cadence have a better PnR solution
and more usage than Synopsys at these advanced 5/3/2nm nodes.
So that's the way it is. And I'm confident about that position
[of CNDS leading] going forward because it is also extending to
RTL synthesis and timing sign-off.
---- ---- ---- ---- ---- ---- ----
Related Articles:
SNPS DC/ICC2/Fusion vs. CDNS Genus/Innovus war is Best of 2019 #9
2nd Fusion Compiler vs. CDNS 19.1 benchmark plus 3 CDNS 19.1 bugs
User benchmarks DC-ICC2 vs Fusion Compiler vs Genus-Innovus flows
12 good and 4 bad switches in new Genus/Innovus/Tempus 19.1 flow
Genus RTL synthesis gaining traction vs. DC is #4 of Best of 2017
Costello on SNPS PnR "still in catch up mode" in 2 years from now
Synopsys layoffs means ICC2 rewrite is unknown for 3 to 4 years out
Engineering comments point to SNPS vs. CDNS PNR shakeout at Apple
Join
Index
Next->Item
|
|
