( ESNUG 499 Item 3 ) -------------------------------------------- [02/23/12]
Subject: Users voting to save Magma Quartz DRC/LVS
TAKING ON WALLY: The neat things users liked about Quartz is its already
MCMM, multi-CPU, multi-threaded -- and it was cheaper than Calibre!
"Assuming the Synopsys-Magma merger goes through, as a Magma user,
which specific Magma tools do you want to survive and why?"
---- ---- ---- ---- ---- ---- ----
I would like to keep Quartz DRC/LVS. Mentor had long been the only game
in town with the Calibre toolset, a great set of tools but a little
tough to get a decent price on a critical tool that is only used for a
small portion of your design schedule.
I initially welcomed Magma to the DRC/LVS space hoping to get some
pricing pressure on Mentor, in the end I preferred the tighter tool
integration and amazing runtimes allowed by Quartz in block and
full-chip runs.
In a small chip taped out last week (~2 M instances) we had Calibre
runtimes of ~2.8 hours and Quartz at ~17 min. Couple that with
phenomenal runset support from within the Magma organization you have
a tool you don't want to lose.
- Mike Lafferty of Availink
---- ---- ---- ---- ---- ---- ----
Over the past two years we have taped out four products at 65 nm and
28 nm using Magma tools. Quartz DRC/LVS are our sign-off tools.
It's very simple to use these tools and show good quality results:
- Runtime: Blocks would run in a matter of minutes, Top level
of ~40 M transistors is verified in about 1.5 hours.
- Accuracy: You cannot tape out with LVS or DRC violations.
Following this logic - the results were accurate.
Quartz DRC/LVS was a great alternative to Calibre; Calibre could be seen
as a very expensive "nice to have" tool for a small company.
- Roman Trogan of Adapteva
---- ---- ---- ---- ---- ---- ----
We would like to see Magma's Quartz DRC/LVS tools kept after the
Synopsys acquisition is completed.
We have used the latest Quartz releases as a supplement to our current
DRC/LVS sign off flow. It has performed very well, and contributed
tremendously to shortening our DRC/LVS cleanup loop.
On our two recent 65 nm tape-outs, we leveraged Quartz' distributed
multi-threading capability to shorten our DRC runtimes. Additionally we
have used LVS tool on the digital blocks to quickly find errors before
running the sign-off tools. This allowed us to turn around several DRC
iterations while churning timely top level runs.
Using 7 threads, a 4.7 GB GDSII database completed a full DRC run in 48
minutes using 3.4 GM of memory; the multi-thread setup completed 2.5X
than a single thread run of Quartz and with the same violation counts
and locations. We have seen similar memory and run time improvements in
Quartz' DUMMY fill runs.
Quartz' memory and runtime improvements along with its multi-threading
capabilities in addition to the cost of the tool, makes it a very good
candidate to be kept around.
- Aldwyn Sealey of Availink, Inc.
---- ---- ---- ---- ---- ---- ----
Our design was run on a network of machines to test Quartz scalability.
For this size "design", Magma advises 8 to 16 CPU cores for best results
(the cores can be on the same machine, or more typically on multiple
machines). What we observed:
# of CPU cores Quartz Runtime (hours)
-------------- --------------
8 5.90
16 2.97
32 1.95
48 1.48
Quartz scalability is great from 8 to 16, and pretty good at 32. Their
results match their claims. At 48 CPU cores there were diminishing
returns, but we believe that for larger designs the scalability should
continue from 48 to 64 CPU cores. Why? Because the design is larger.
More transistors and wire in the design requires more testing of the
data. This means that a task like checking metal spacing has more edges
to check -- which also means the tasks can be spread among more CPUs
efficiently.
Linear or not, being able to do our full chip verification in under
2 hours is what counts for us.
- Bob Patti of Tezzaron in ESNUG 483 #2
---- ---- ---- ---- ---- ---- ----
Juniper used Magma Quartz for block and full chip physical verification
work on our last set of chips. These are some of the largest chips that
were taped out in the 40 nm domain. Quartz was able to successfully meet
and exceed runtime and accuracy requirements that are key in a signoff
product.
Quartz has proven to be a strong performer for us in LVS/DRC, and the
excellent short-tracing capabilities make it stand out above competing
tools.
- Bharat Bisen of Juniper Networks
Join
Index
Next->Item
|
|
