( SNUG 05 Item 19 ) ---------------------------------------------- [12/20/05]
Subject: What Users Would Change About Synopsys
THE OPEN ENDED QUESTION (Part I) -- This is one of the parts I like best
about this survey because you get to see what's bothering users and what's
the top thing on their mind.
"A voodoo priestess formerly from New Orleans swaps your soul with
Aart de Geus, the CEO of Synopsys. Even though you were a Synopsys
customer before, now that *you* are the new CEO of Synopsys:
What 2 or 3 things about Synopsys would you DEFINITELY change?"
have less bugs in releases: ######################### 25%
have R&D visit customers more: ###################### 22%
Milkyway/.db database trouble: ###################### 22%
prices too high: #################### 20%
improve support or hotline: ################## 18%
too much general arrogance: ############## 14%
sales & sales arrogance: ############# 13%
fix internal tool flows: ############ 12%
licensing too complex: ########### 11%
drop Magma lawsuit: ########## 10%
fix Astro: ####### 7%
fix DC: ####### 7%
use ICC or PhysOpt or Astro?: ####### 7%
fix Formality: ##### 5%
use Scheme/Tcl/Tcsh/dc_shell?: ##### 5%
fix SolvNET: #### 4%
back compatibility troubles: #### 4%
fix DW: #### 4%
fix non-Synopsys tool flows: ## 2%
fix PrimeTime: ## 2%
No surprises here (at first). It doesn't take Eistein to figure out that if
you asked a bunch of EDA customers what they'd want from *any* EDA company,
the customers would want "less bugs" (#1), more "R&D visits" (#2), and that
"prices too high" (#4). And pretty much everything else is some specific
variant on "less bugs" -- with the exception of 3 complaints -- "general
arrogance" (#6), "sales arrogance" (#7), and "drop Magma lawsuit" (#10).
---- ---- ---- ---- ---- ---- ----
Improve SolvNET
Stop the lawsuit against Magma
- [ An Anon Engineer ]
Concentrate on their customers and not taking rivals to court.
- [ An Anon Engineer ]
Stop suing my competitors to get them out of business instead of competing
with them on technical aspects/ service offerings/ solution orientation.
- [ An Anon Engineer ]
Dump the lawsuit with Magma, make up with Rajeev, and apologize to my
customers, and tell them I'm going to win their accounts based on
technical merits, and not FUD.
- [ An Anon Engineer ]
Improve Astro/PhysOpt. Match the frontend where DC is best in class.
Compete on products -- not by evangelization, acquisition and litigation.
Stop relying on pre-emptive long term pricing chokeholds.
- [ An Anon Engineer ]
The litigation circus -- it is embarrassing.
Lower prices -- make up for it in volume.
Scheme interface to Astro -- change it all to TCL and slurp in
def/pdef/db directly from PhysOpt.
- [ An Anon Engineer ]
Try to change the arrogance of the sales force. That's always bugged me
as a Synopsys customer for more than 15 years.
- [ An Anon Engineer ]
Arrogance of some individuals must go!
- David Black, consultant
Get rid of the smug attitude and put together real flows.
- [ An Anon Engineer ]
Become less arrogant
Improve quality of tool releases - run regressions (novel idea)
- [ An Anon Engineer ]
1. Attitude. Synopsys is arrogant
2. Sales practice. Cadence is MUCH easier to do business with than
Synopsys. Synopsys seems to come to the table with assumption that
you are going to pick their tools no matter what they do or don't
do (see #1).
They negotiate from that position. It takes a Declaration from
Congress to get an eval license, heaven forbid a loaner license in
a time of need!
- [ An Anon Engineer ]
Not rape you by absurdly raising prices when they have you by the balls.
- [ An Anon Engineer ]
Change the spirit from "we are best" to "how can we make ourself the best"
- [ An Anon Engineer ]
Make Synopsys more customer friendly
- Dinesh Venkatachalam of Legend Silicon
They should listen to more of their customer voices for tool improvements.
Some of their new tools actually had worse log file format and user
interface.
- Wanhao Li of Zoran
Show some heart to the guys in the trenches instead of exploiting their
time of need (e.g. peak use license flexibility). Synopsys is notorious
for this.
- [ An Anon Engineer ]
Tool qualification and release procedure. I've yet to come a dc_shell
release that does not contain show stopper bugs.
Unify sales groups. Every sales person should be able to fully support
the customer. It makes absolutely no sense that we have to negotiate
twice, once with the tools sales person, and a second time with the IP
sales person. It's a waste of time, and you don't get the proper support.
- [ An Anon Engineer ]
1. Free license Fridays
2. English lessons for support techs
- Tom Mannos of Sandia National Laboratories
Pricing, customer support
- Richard of Reindl of L-3 Communications
I would fix the mess of IC Compiler vs. PhysOpt/Astro. Apparently
there are TWO horses running inside of Synopsys: Avanti team and
PhysOpt team. Avanti is winning. Was it Aart's intention?
Fix that DC/PhysOpt runs out of steam.
Quick consolidation of *sim, VCS and HSPICE.
- [ An Anon Engineer ]
Re-write Astro so it will work with a large flat netlist without
breaking or myriad scripts to get around problems. Ditto on Star-RCXT.
- [ An Anon Engineer ]
Merge the best of DC with the best of the Avanti flow. I will consider
it merged when both tools use the same timing engine.
- Kevin Broe of Britestream Networks
The AEs need better training on an overall flow. They seem to be good at
one tool and have no clue about what happens downstream on the next tool.
Typical example would be DC/PhysOpt and Astro. I have spent 5-6 months
getting XG flow with Milkyway working and most of it was debugging.
- [ An Anon Engineer ]
Good standalone tools. Please don't start packaging them.
- [ An Anon Engineer ]
Understand that the easiest sell is to a returning customer.
- [ An Anon Engineer ]
The company has a culture & business focus that is myopic and inward
focused. They need to change to pay attention to their customers and to
have strategic vision. DC is a cash cow for only so long... It's very
interesting to me that we don't have any customers pushing us to work
with Synopsys backend flow. That's very telling given the large number
of high-profile semi companies we're working with.
- [ An Anon Engineer ]
The company has gotten too big and lethargic. It's become an amalgamation
like Cadence, making it difficult to stay energized. They need a major
restructuring, getting fewer layers between R&D and customers. Get R&D
out to talk to customers directly, find out what the product focus should
be. Don't rest on your laurels (even PrimeTime) because companies like
Magma are closing in fast.
- [ An Anon Engineer ]
dump Scheme and convert EVERYTHING to tcl.
- [ An Anon Engineer ]
Add a &**(((**&^^!!!! tcsh interface to dc_shell.
- [ An Anon Engineer ]
Maybe talk to some of the users now and then.
Switch to C based API's. Acknowledge that Tcl is NOT faster than C.
Be more willing to admit when something is wrong or broken.
- [ An Anon Engineer ]
Replace the dumb forms fill out style of Astro and put in a real language.
- [ An Anon Engineer ]
Make tools more user-friendly and simple. Scripts get too complicated to
work right. Make tools & scripts more backward compatible.
- [ An Anon Engineer ]
Why in hell our scripts blow up everytime somebody patches the tool?
Moreover, why time after time it is claimed to be a feature?
- [ An Anon Engineer ]
Improve the VCS compile time and integrate it better with Vera.
- [ An Anon Engineer ]
Do whatever it took to get TSMC to start to use Hercules. Put all
available resources on IC Compiler - to get it to market and to get
it debugged.
- [ An Anon Engineer ]
Align the documentation -- man pages, Sold and online help all are
different, and often contradict each other.
Multiple licenses per tool, it always seems the feature you need
requires the special super/ultra/max license you don't have.
- Brendan Barry of S3 Ltd.
Licensing complexity and buddle complexity. Feel like buying a Toyota
Corolla, where everything is an option.
- [ An Anon Engineer ]
Not force customers into the subscripton licensing model.
- [ An Anon Engineer ]
Give the local sales reps a little more power to cut temp keys and
negotiate deals and get his work done rather than go through all the
red tape involved. Lower prices.
- [ An Anon Engineer ]
Change the strange licence politics.
- [ An Anon Engineer ]
Improve the quality of SW releases. Anticipate problems, not fix them.
- [ An Anon Engineer ]
Support is the next most important thing behind new and actually WORKING
new features. The support guys are the ones actually in day-to-day
contact w/ customers (engineers). These are the important guys to
keep around.
R&D needs to REALLY shake out a tool/revision before handing it off;
which to me means even MORE support guys so that my customers
(engineers) don't get bitten by a new change to the tools that's
an undocumented FEATURE!!
Understand that my actual end customer is the engineer using the tool NOT
the VP buying said tool/tool suite. Maybe some actual user face time on
site might be helpful, and not just a speech, but an actual discussion.
- Gzim Derti of Agere
I would like project based tooling cost. I define a project, and for a
fixed cost, get access to all the tools to complete the job.
- [ An Anon Engineer ]
Address tool quality. It slipped for DC and Formality last year.
Add more "reality checks" between R&D and users to understand how
tools are used in the field and how various tools (even those from
Synopsys) need to better work together.
- [ An Anon Engineer ]
Stop with the patches and generate tools from scratch with all latest
customer requirements - see Magma as an example.
- [ An Anon Engineer ]
Stop being forced to pay money for minor new features that should be
part of a tool upgrade.
Get a real Silicon Virtual Prototyping tool -- Jupiter does not cut it.
- [ An Anon Engineer ]
Kill off IC Compiler. Start from scratch, need a new start on the DB.
Develop a REAL SW QA team.
Stop doing IP and things that compete with customers
- [ An Anon Engineer ]
IC Compiler needs timers to match with one common unified DB!
- [ An Anon Engineer ]
Drop Formality as it's "abandonware" and the resources could be better
used elsewhere. Make the PrimeTime group more responsive to customers.
- Russell Petersen of Scientific Atlanta
Make an easier and better equivalent checking tool (Formality), especially
when it comes to debug.
- Robert Tricerri of iStor Networks
I'd work on getting Formality to recognise DC/PhysOpt's optimisations more
reliably, and improving the analysis facilities so that the debugging
process is not such a black hole of despair.
I'd improve the coreBuilder/Consultant/Assembler tools so that they were
a lot more flexible in what they could accept and generate.
- David Smith of STmicroelectronics
Take the 3M approach to PrimeTime... multithread, multicorner and
multimode... and I'm not talking that hack they have in now. In about
a year, if they don't, they will lose PrimeTime's market to a
competitor. At 65 nm and below my current business model means MANY
simultaneous runs, and runs that take days to run. Right now, tapeout
is a race between PrimeTime and Calibre as to who gates the tapeout.
When Mojave comes to town, PrimeTime sticks out like a sore thumb.
- [ An Anon Engineer ]
PrimeTime, it's behind in static timing technology, variation/
statistical timing, etc.
For God's sake, abstract away collections. There's no reason for a
separate foreach_in_collection command. A simple foreach should do
and the tools should figure it out.
- [ An Anon Engineer ]
Catch up with tool development for requirements of 65 nm processes node.
Low power in general (65 nm and 90 nm) design requirements are not
adequatly met with today's tools.
OCV issues are woefully missing tools that truly address the problems.
Physical understanding must be brought into tools like PT to comprehend
variations due to Photo/Etch variations (OPC/RET, micro-loading ...)
and what they will do to your timing.
- [ An Anon Engineer ]
Fire engineers who close tickets before the problem is solved.
- John Schritz of Tektronix
More focus on releasing tools that are less buggy, don't crash, etc.
Have some form of anon customer eval of AE support. They have some AE's
who should not have their job.
More open customer relationships regarding bugs. Creating testcases
can take considerable time, and often the answer back is "we already
know about that and it will be fixed in the next release" is the answer.
SolvNET articles need reviewed and checked before being published.
They often include incorrect commands, typos, and don't work as expected
"out of the box". It is nice that one can provide direct feedback to the
publisher of each article.
- [ An Anon Engineer ]
Be more clear about the future roadmap of products especially in P&R.
Hire more Application Consultants since they always seem to be in
short supply.
- Sunil Malkani of Broadcom
If the tool has big change in technology/database please keep tools
backward compatible and fix the bugs. Since last year we have chased
Synopsys to adopt XG, Milkyway, new DFT technology, etc. Had lots of
issues along the way. Sometime we are told the bugs can't be fixed
unless we adopt new form. This prevents us to lock on one version and
force us keep changing. This situation is very hard when you have
tapeout work to do.
Establish some transparent evaluation system for Synopsys AEs, so
customers have a chance to feedback the support they get from SYNOPSYS
engineers.
As tool has major change, keep providing consistent and honest info to
customers. We got confused when Synopsys keeps changing recommendations
on PhysOpt library, from 2.5D->tluPlus->Milkyway FRAM view in a short
period. Inconsistent and lacking of details information makes customer
hard to be convinced and follow.
- [ An Anon Engineer ]
Acquire a company that provides a user-friendly tool that allows RTL
designers to do front-end Design Planning and very early floorplanning
that generates DEFS for PhysOpt. It should also be able to read DEF.
Synopsys is way behind in providing this capability (in a usable form)
and we're stuck without a solution until Jupiter completely migrates out
of Scheme and into TCL and provides a usable black-box flow for frontend
design teams. Without this capability, a frontend team can't get any
useful data out of PhysOpt.
- Lauren Carlson of StarGen Inc.
Real support for industry standard databases and links to competitor
tools.
- [ An Anon Engineer ]
Make the tools more open standards (OpenAcess) and interoperable with
other vendors.
- [ An Anon Engineer ]
Embrace open standards. Turn away from using proprietary formats just
to lock the customers in.
- Are Arseth of Atmel
Low cost batch simulator with SystemC & PSL support for regressions.
Open tool framework to support alternative tool integration.
- [ An Anon Engineer ]
Stop excessive trend towards tool integration to get captive users.
Allow advanced users more control over DC (switches, runtime scripts)
More open API's
- [ An Anon Engineer ]
I would continue to lock customers into my databases: layout dbs are HUGE
and getting bigger -- Moore's law of dbs.
- [ An Anon Engineer ]
I would set in place better mechanism to receive, engage and act on
suggested tool improvements.
I would get out of the IP business (spin it off and maybe keep a
partnership in place).
- [ An Anon Engineer ]
Pricing is high for smaller companies. Lite version is very limiting.
- John Gedde of Aeroflex Laboratories
Restructure the pricing to change the dynamic. Look at acquiring Magma.
- [ An Anon Engineer ]
Lose Library compiler. Asking money for it is akin to extortion.
Tools should be able to read plain-text formats without extra licenses.
You should not need to have Star just to use TLU+ models, again tying
tools together this way is pretty dodgy.
- [ An Anon Engineer ]
Get rid of the .lib to .db part in Library Compiler and make sure
all tools can read .lib properly.
Stop treating VHDL as a second tier language and support it properly.
Get a clear startegy on which tool is supposed to do placement in the
Synopsys tool flow.
- [ An Anon Engineer ]
As a US citizen (born and raised here), I am very leary of dealing
with China. That is where my HSPICE bugs/testcases go.
- [ An Anon Engineer ]
There's no voodoo priestess powerful enough to turn me into a CEO.
- [ An Anon Engineer ]
Aart is doing a good job. I would not make any major changes to Synopsys.
- Jonah Probell, consultant
Index
Next->Item
|
|
