( ESNUG 463 Item 1 ) -------------------------------------------- [03/16/07]
Subject: Answers from Rajeev Madhavan, CEO of Magma
> Why doesn't Magma have any high-speed customers like Qualcomm, Philips,
> Intel? If required speeds are over 600 MHz, you don't show up.
All 3 customer you have listed are Magma customers - one is in the top 20,
and we expect to add more penetration into others - Customers like TI,
Broadcom, Cray are doing some of the highest performance designs and designs
at 1.5 GHz has been done using Magma.
As one of our customers, Jeff Echtenkamp said on Deepchip:
"P&R guys are all religious about our tools... So I'm sure the
statements that Magma getting the low hanging fruit of the P&R
world is from a guy who goes to church at Synopsys :) Anyways,
I would disagree strongly with that statement. Honestly, all
backend tools have issues, and it's a matter of the designer's
preference for working around things. If you look in our industry,
there are a lot of different mentalities. P&R started out as
glorified layout... a very graphical thing done by technicians.
Today, it's a very complex problem often run by guys with Masters
in EE. You also have ASIC houses who get netlists blind from a
customer. You get companies who do large bottom-up SOCs. We all
have our styles for a reason." (ESNUG 454 #1)
---- ---- ---- ---- ---- ---- ----
> Raj: how is work going on the Bonn router and when (or will) customers
> be getting builds with it in?
Bonn was a technology exchange between Magma and IBM. As our users know the
latest Fusion and Talus code has significant new routing technology, many of
them are ideas born from the Bonn-IBM relationships - you will see more of
these incorporated into our mid year release.
---- ---- ---- ---- ---- ---- ----
> Multi-CPU machines have been available for several years now. When will
> there be a multi-CPU version of Fusion (placer, timer, extraction, router,
> optimization) available?
Our Talus fully automated chip creation flow can parallelize most time
consuming RTL2GDS tasks including RTL synthesis, placement, routing and
extraction. In fact, it is because of this distributed automation that
we claim Magma can iterate the largest chips from RTL to final routing in
less than 2 days.
---- ---- ---- ---- ---- ---- ----
> Rajeev, with such "superior" tools over Synopsys, Cadence, and Mentor, why
> is your growth in marketshare stalled?
Market leaders will take up to 50% of the market share and Magma is on its
way to just do that. Magma has more than 1/3rd of the market share today
but that share significantly increases in 90 nm and 65 nm designs. We
believe that the complexity and challenges with 45 nm designs will play to
Magma strengths and take our market share well over 50%.
---- ---- ---- ---- ---- ---- ----
> Rajeev Madhavan: Every time anyone talks to you about your products, you
> tell them you are the best and there is nothing like it in the market. So,
> why don't you own 100% of the market?
See Above. For as long as people design less complex, 130 nm or above nodes
(still many designs) the pain point is not significant to consider change.
Synopsys and CDNS were there first and customers will need to experience the
need to change. We believe the rate of conversion will be accelerated with
the introduction of 65 nm production designs.
---- ---- ---- ---- ---- ---- ----
> Rajeev - You said in interviews that "DFM" is "Design for Marketing".
> Is that Magma's stance, too?
There is no question that DFM has been hyped with half a dozen smaller
companies evident with your panel invitees providing post-processing
solution to bridge the gap between design and manufacturing.
If you are talking about accounting for manufacturing during design, we
believe in that and we are investing in tools that improve yield during
design. We do not believe in post-design corrections as they introduce
timing, power and other issues.
---- ---- ---- ---- ---- ---- ----
> Is Magma developing a System Verilog simulator? Why or why not?
We support System Verilog in our RTL implementation flow and we see the
usage of it taking off.
We do not comment on areas we have not launched products in, but suffices
to say that should Magma enter logic verification through acquisition or
internally developed we have to provide significant differntiatable
capability. A "me too" is not in our DNA and will not be undertaken.
---- ---- ---- ---- ---- ---- ----
> What's your strategy for making money in the FPGA EDA business? Other
> ASIC focused companies, including Cadence and Synopsys turned tail and
> ran. Are you making a similar mistake?
Our strategy with FPGA has been simple; we are a technology provider for
FPGA vendors and have been working with large and smaller FPGA vendors.
Our FPGA technology is geared to some partners who distribute to their
end customers and to mutual ASIC customers, the ones that use FPGAs for
early prototypes.
---- ---- ---- ---- ---- ---- ----
> Blast FPGA, a "coming soon" demo on deepchip, what else?
Blast FPGA has been delivered to partner FPGA companies. John, I can
arrange for you to see a demo.
---- ---- ---- ---- ---- ---- ----
> 1. Why doesn't Magma improve their documentations, and why is their
> flow not yet stable, after remaining in the industry for so many
> years, why does it still has issues? Every time we try to do
> something new, try to develop some new flow, we get hit by Magma
> tool limitations which hits back painfully.
Advanced flow and new capabilities often get ahead of documentation and we
make them available to our customers to finish their designs. This level
of openness makes standardization on a limited set of scripts harder.
Having said that we are introducing special flows (Low power, speed, SI
closure, DFM) in Talus to our customers to address their needs in their
specific areas.
> 2. Why doesn't Magma try to integrate products in its portfolio, like
> cross domain clock crossing flow, Auto identification of MCP and
> False paths?
We have several requests from customers to look into exception verification
flow. Please stay tuned.
---- ---- ---- ---- ---- ---- ----
> To Magma: When are you going to get serious about test and DFT? Follow-up
> depending on response: Why don't you buy LogicVision before Mentor does.
We do not publicly discuss any acquisitions, etc.
LogicVision & Mentor are both partners of Magma and we have mutual customers
working on a joined flow. Providing standard DFT solution to end customers
is more important than owning the solution.
---- ---- ---- ---- ---- ---- ----
> Since LAVA is late in the game for TEST, how is LAVA going to compete with
> MENT & SNPS who already have well established TEST solution. MENT Fastscan
> and TestKompress is doing very well and SNPS TetraMAX and DFTMax seems like
> catching up, but not quite yet. CDNS had acquired the TEST solution from
> IBM, but not much progress was noted in the past few years.
>
> Rajeev, don't you think partnering with MENT may be a better solution than
> developing from scratch?
Yes, we continue to partner with Mentor for a common good solution. Magma
strategy is to provide a working solution than owning a product. Remember
we enter a market only If we can offer a significant benefit to our end
customers and until such point, we are not ready to announce any tools.
---- ---- ---- ---- ---- ---- ----
> Rajeev: What is Magma's plan to improve the link between Volcano data base
> and Open Access?
We have said this a million times. Volcano is a data model & different than
OA, a data base. We have nothing against Open Access and in fact worked with
a few smaller EDA companies to provide a mutual solution to our customers.
We are a customer driven company and if asked and shown mutual benefit of
working with OA we will do that. Ultimately, this has clearly never been an
issue for our customer, once they realize the strength of our data model.
It's clear that complex 65/45 nm designs require a unified single data model
for concurrent optimization and analysis design flow and not a binary
exchange of files. We stated that in past and our competition started to
imitate us in that making Open Access less relevant.
---- ---- ---- ---- ---- ---- ----
> Your Quartz physical verification tool is brand new. The command syntax
> is quite different from Calibre. If you don't provide any easy conversion
> programs or assists, how do you think you can get quick adaption, not to
> mention the customer QA cycles?
Conversion assists are good, so stay tuned. But note that for advanced
technology nodes, traditional runset languages are flat and have
15,000+ lines -- this is not scalable. In contrast, Magma's runset
language is fully procedural, and offers much better productivity and
compactness (e.g. 3x less effort). IBM presented an interesting MUSIC
paper about their experience with Magma's runset language.
At advanced nodes, customers want native Quartz DRC/LVS runsets. Here,
Magma has worked with leading customers to deploy runsets for 130, 90,
65, 45 and 32nm -- e.g. the runsets that customers download from TSMC
and UMC for Quartz DRC/LVS.
---- ---- ---- ---- ---- ---- ----
> What is happening with Quartz DRC & LVS? Is it ready for production? Who
> can testify about the tool performance vs. Calibre? There was yet another
> demo and the feature list at DAC 2006 -- but no happy customers who are
> willing to talk?
Actually we have lots of happy customers. Quartz DRC has been in production
since June 2005, and LVS since January 2006. IBM, TSMC, Broadcom and Nvidia
were our key beta partners and each has discussed their Quartz DRC/LVS use
in press releases or in papers presented at MUSIC, Magma's users conference.
LSI, IBM and TI spoke at MUSIC about their adoption of Quartz DRC/LVS.
We have over 25 customers using Quartz DRC/LVS and offer production support
for designs at 90, 65, 45 and 32 nm. We expect to continue gaining market
share: Quartz DRC is sign-off certified at TSMC and UMC, and certification
at other foundries is imminent. Our estimate is that we will get 10% market
share for this year.
---- ---- ---- ---- ---- ---- ----
> Why did Hamid Savoj leave Magma? He was a co-founder of Magma in 1997.
> He started as a principal engineer ended as Senior Vice President,
> Product Development. He recently joined Envision Technology, which is
> a kind of interesting stealth-mode company, in Q4 (I think) 2006 as
> VP Engineering.
Hamid is a great friend and I enjoyed working with him - in fact, he, myself
and Karen are attempting to organize a 10 year Magma anniversary party.
Hamid has contributed to Magma's growth for 9 years and he felt it was the
time to change from the 24x7 job that he had at Magma. I along with all
the Magma people wish him the best.
---- ---- ---- ---- ---- ---- ----
> Are there plans for something like a wrapper for the Magma tools that
> would enable major stakeholders in Synopsys to use their current Synopsys
> scripts and changeover a full environment to Magma with little risk or
> loss of time? Our biggest problem with changing tools is infrastructure.
Magma supports all the standard constraints and advanced library modeling,
but we have not attempted to provide support for proprietary historic
scripting formats. In fact, if you were to ask all Magma users to why
they use Magma over our competition you would find that open data model
and faster TAT comes out a lot (an inherent benefit of integrated unified
flow). We prefer that all EDA vendors compete on their tools and not
proprietary formats.
---- ---- ---- ---- ---- ---- ----
> What's your reaction about the Cadence Power Forward Initiative (PFI)?
> Magma is also trying to do the same stuff but why? Cadence has already
> open sourced the common power format in Si2 and Magma is a part of Si2.
Standard formats need to be open and accessible and debatable by all.
Cadence power format came short of that. It was neither open nor debatable.
After talking to some of our biggest customers concerned with power, it was
apparent that we needed to initiate a new OPEN power format with our
customers and EDA vendors to join. Synopsys, Mentor, and many large
customers joined to create this new initiative (UPF, Unified Power Format)
and Cadence has participated in this effort.
Design formats are not political tools to control by 1-2 bodies. They need
to be donated to standard bodies such as Accellera or IEEE which is what
UPF is.
Magma is part of Si2 and we support many of their initiatives but we are
not obligated to support all of them. A better question is put to CDNS as
to why not join the other 3 major EDA vendors and 5 major users on power
format?
---- ---- ---- ---- ---- ---- ----
> How does Magma plan to tackle the advanced processes having RC dominated
> paths, with respect to the Fixed Timing Methodology (FTM) of super cells?
> FTM highly depends on lib cell granularity, linearity & characterization.
Talus has already eliminated super cells. It uses a new delay model that
takes into account load and buffering as well as library granularity.
Across the board, we have better results and more usability wrt preparing
libraries in Talus. I invite you to the Magma users group to see
a tutorial on this.
---- ---- ---- ---- ---- ---- ----
> Why has Molten had minimum success, while Synopsys SolveNet has been very
> successful? It lets us rely less on Synopsys AEs to answer questions?
Molten, which uses commercial tools like SalesForce.com, allows us to
implement beyond proprietary tools. To date, we have used this to let our
customers report issues and track them. That phase of Molten has been
completely implemented and we have now moved to second phase where we
will provide common solutions, best practices, workarounds to our users.
The Magma AE support burden has significantly been reduced which means the
usage of Molten is picking up. What we need to do is add more solutions;
you will see this being added at a rapid pace. Having said that,
communication with an experienced AE knowledgeable of a full flow should
never be replaced with an automated response tool and we believe in making
your tapeouts successful and will ensure that premium support is provided.
Index
Next->Item
|
|
