( ESNUG 584 Item 10 ) ------------------------------------------- [11/07/18]
Subject: SCOOP -- Nvidia leaks new MENT tool to take on CDNS AMS Designer
NVIDIA LEAKED: It was a sleepy rainy day in Boston. I had a tool run going.
Because it had buzzwords like "Big Data" and "AI" in it, I thought I'd kill
30 minutes watching a Mentor U2U talk by an Nvidia guy on GPUs and PHYs.
(click on pic to enlarge image)
|
Why? GPUs are digital. PHYs are analog. I figured it'd be interesting
because Analog Mixed Signal is a market that MENT isn't in. The very rough
*estimated* 2018 AMS marketshare data:
Cadence AMS Designer: :################################## $138.7M (73%)
Synopsys VCS AMS: :######### $34.2M (18%)
Mentor Questa ADMS: :## $12.3M (7%)
Mentor ICanalyst: :# $1.8M (1%)
Mentor Tanner AMS: :. $0.4M (1%)
This is based on the 2018 estimates from GSEDA. The absolute $ millions can
be way off -- but the relative marketshare ratios are accurate.
---- ---- ---- ---- ---- ---- ----
So the Nvidia guy (Ting Ku) was talking about using Verilog for his digital
blocks (as expected) plus a Verilog-A checker in his AMS design.
(click on pic to enlarge image)
|
Notice the Analog part is in yellow, and the smaller Digital control logic
in blue. And right when Ting Ku was supposed to telling us how he used
Mentor Questa ADMS to do the AMS environment -- this is a Mentor User Group
talk after all -- Ting Ku instead mentioned at time stamp 23:35 a totally
new and yet to be announced Mentor AMS environment called "Symphony"!!!
Got that? The Nvidia guy didn't say squat about the Mentor Questa ADMS
environment at all -- but instead he was chatting up this new Symphony!
By time stamp 24:20 the first slide with "Symphony" popped up!
(click on pic to enlarge image)
|
Look close to see BDA AFS for SPICE simulations and VCS/Incisive/Questa for
the Verilog simulations. (TECH NOTE: whenever I discuss EDA flows, I tend
to reference the leader in each category. For Verilog, it's VCS. For AMS,
it's CDNS AMS Designer. But the truth is both CDNS and SNPS only have
*closed* AMS environments. That means the SNPS VCS does not work with CDNS
AMS Designer; and CDNS Incisive does not work with SNPS VCS AMS.)
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah Symphony had to run within our existing
tool flow blah blah blah and had to be easy to set up blah blah blah and
it did both!
WHAT I HEARD: We're using CDNS AMS Designer plus a few thousand BDA AFS
SPICE and Synopsys VCS licenses. We'll swap out AMS Designer for Symphony
only if it's a direct swap plus it's a painfree set-up blah blah "AS IS"
tool insertion blah blah ... and it did both!
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah incumbent tool took 160 hours while the
new Symphony took 19 hours blah blah blah Symphony found the same error at
870 nsec same as the incumbent tool...
WHAT I HEARD: Blah, blah, blah Cadence AMS Designer took 160 hours
while the new Symphony took 19 hours -- 6 days vs. 1 day -- 8.4X faster!
blah Symphony found same error at 870 nsec as Cadence AMS Designer did...
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah digital testbench blah blah blah blah with
the analog DLL block was being tested inside the digital testbench...
WHAT I HEARD: Blah, blah, blah digital (VCS) testbench blah blah blah blah
with the analog DLL block (in BDA AFS SPICE) was being tested inside the
entirely digital (VCS) testbench...
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah here's the Symphony command line itself
blah blah blah see how easy it is? It uses the digital simulator options
AS IS! And BDA AFS follows with its own options...
WHAT I HEARD: Blah, blah, blah here's the Symphony command line itself
blah blah blah see how easy it is? It uses VCS with its options AS IS!
VCS inserts in painlessly. And BDA AFS follows with its own options.
Your SPICE guys don't need VCS skills. Your Verilog guys don't need
BDA AFS skills. Both simulators just cut-and-paste stitch into the
Symphony flow AS IS!
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah did the DDLL lock test successfully blah
blah blah Symphony took 9 hours, incumbent tool took 18 hours...
WHAT I HEARD: Blah, blah, blah did the DDLL lock test successfully blah
blah blah Symphony took 9 hours, Cadence AMS Designer took 18 hours...
---- ---- ---- ---- ---- ---- ----
(click on pic to enlarge image)
|
WHAT HE SAID: Blah, blah, blah our experiences comparing Symphony versus
the incumbent tool blah blah Symphony was 8X-2X-2X faster ....
WHAT I HEARD: Blah, blah, blah our hand-on benchmark of Symphony versus
Cadence AMS Designer blah blah Symphony was 8X-2X-2X faster ....
---- ---- ---- ---- ---- ---- ----
RAVI'S WAR: Let's see... Mentor... BDA AFS... SPICE... AMS... it doesn't
take Einstein to see Ravi Subramanian's fingerprints all over this yet-to-be
announced Siemens/Mentor Symphony AMS design environment.
Why? Because Symphony is a direct frontal assault on Tom Beckley's 74%
market share CDNS AMS Designer. You know, the very same Cadence that when
BDA AFS benchmarked 5X-10X faster than Cadence Spectre/APS (ESNUG 494 #9)
Cadence sued Ravi's Berkeley Design Automation in order to put BDA out of
business on bullshit charges? You know that lawsuit where Cadence had the
balls to even hint that it was going to also sue the BDA AFS users, too?
Unnamed 25 "John Doe" users as defendants in the lawsuit. (ESNUG 510 #9)
And then the EDA customers themselves went total bullshit angry on Cadence
for this Oasis/BDA/Cadence lawsuit? (ESNUG 523 #2) And then finally the
Federal judge in the U.S. District Court in Northern California dismissed
this CDNS vs. BDA lawsuit "with prejudice" -- meaning Ravi won, BDA won,
and the 25 unnamed "John Doe" BDA AFS users won, too? (ESNUG 535 #7)
It's *that* Cadence that Ravi's secret new Symphony AMS tool is attacking.
And I'm not surprised. :)
- John Cooley
DeepChip.com Holliston, MA
---- ---- ---- ---- ---- ---- ----
Related Articles:
SCOOP -- Judge dismisses CDNS vs. BDA lawsuit "with prejudice"
So far public opinion is strongly anti-Cadence on CDNS vs. BDA
Does Cadence-Berkeley lawsuit mean some ADE users will be sued?
Berkeley AFS benchmarks 5x-10x faster than Cadence Spectre/APS
Join
Index
Next->Item
|
