URI:
   DIR Return Create A Forum - Home
       ---------------------------------------------------------
       Micro-Cap EDA Users
  HTML https://mc12.createaforum.com
       ---------------------------------------------------------
       *****************************************************
   DIR Return to: Micro-Cap 12 Discussion
       *****************************************************
       #Post#: 88--------------------------------------------------
       MCNC test benchmarks: MC12 Vs Qspice.
   DIR By: Rohit
       Date: February 23, 2025, 1:15 am
       ---------------------------------------------------------
       MCNC test benchmarks: MC12 Vs Qspice.
       There is a new free spice simulator called Qspice by the author
       of LTspice. I was browsing through Qspice forum and came to this
       post:
  HTML https://forum.qorvo.com/t/qspice-simulation-speed/14327
       .
       Here Qspice was compared against two other spice simulators in
       MCNC benchamrk suits, it beats them in speed and simulation
       completion. The MCNC benchmark is a set of circuits used to
       evaluate and compare the performance of electronic design
       automation (EDA) tools, particularly those involved in analog
       and mixed-signal circuit simulation.
       The post doesn't mentions what are spice#1 and spice#2 and alos
       doesn't provides specifications of the machine on which the test
       was carried out.
       I select a few circuits which spice#1 and spice#2 failed to
       simulate, ran these tests for Qspice and MC12, measured time.
       And the results are:
       Here are some points to consider:
       1. The test were run on a PC which specs: OS: Windows 10, CPU:
       intel core i7-3770 @ 3.40GHz, RAM: 12GB DDR3.
       2. This is the MCNC benchmark suite was downloaded from the
       Intusoft page (
  HTML http://www.intusoft.com/benchmarks.htm
       ). The
       one used by Qspice
       (
  HTML https://www.marcusaureliussoftware.com/decks.zip)
       have a
       different syntax.
       3. Qspice was set to use "Fast (less accurate) math". MC was set
       to use "Standard Default" in the Global setting options. All
       other options were left to default for both programs.
       4. MC, unlike Qspice, doesn't provides time to complete the
       simulation. it was measured using a stopwatch. For small
       circuits, simulation time was insignificant (< 1s). It has been
       shown as "1s" in the test results.
       #Post#: 89--------------------------------------------------
       Re: MCNC test benchmarks: MC12 Vs Qspice.
   DIR By: MC12
       Date: February 23, 2025, 2:51 am
       ---------------------------------------------------------
       Impressive results for MicroCap, considering it is no longer
       being developed.  Perhaps it had a head start by being
       commercial.
       A few questions come to mind:
       Why are your tests a subset of the benchmarks you linked to, and
       how/why did you choose which subset to run?
       The "pc_frame" test (which MC failed to complete) has the
       largest number of MOSFETs to simulate in your subset (14,265),
       and the second largest in the whole benchmark set (the largest
       being 18,816).  Do you think it exceeded MC's capacity?
       #Post#: 90--------------------------------------------------
       Re: MCNC test benchmarks: MC12 Vs Qspice.
   DIR By: Rohit
       Date: February 23, 2025, 3:57 am
       ---------------------------------------------------------
       --- Quote from: MC12 link ---
       >
       >
       > Why are your tests a subset of the benchmarks you linked to,
       and how/why did you choose which subset to run?
       >
       > The "pc_frame" test (which MC failed to complete) has the
       largest number of MOSFETs to simulate in your subset (14,265),
       and the second largest in the whole benchmark set (the largest
       being 18,816).  Do you think it exceeded MC's capacity?
       >
       --- End Quote ---
       As you can see, simulation take a lot of time, I chose only
       circuits which other spice software (spice#1 & spice#2) failed
       to simulate. In future I may test all files, but this subset is
       enough to prove MC's capabilities.
       "pc_frame" probably uses some non-standard feature/syntax that
       other spices fail to parse. Ngspice and LTspice also fail to
       simulate this see:
  HTML https://sourceforge.net/p/ngspice/discussion/127605/thread/c5eb7e7f/<br
       />. MC may be able to simulate after some syntax correction.
       "smult20" is the biggest file (bigger than pc_frame ) I tested
       it and simulation started (but I didn't have time to complete
       it) so pc_frame is well within MC's capability.
       I encourage you to run some tests on your PC and post the
       result. Download link for the MCNC test suite:
  HTML http://www.intusoft.com/models/MCNC.zip
       #Post#: 91--------------------------------------------------
       Re: MCNC test benchmarks: MC12 Vs Qspice.
   DIR By: MC12
       Date: February 23, 2025, 4:11 am
       ---------------------------------------------------------
       --- Quote from: Rohit link ---
       > I chose only circuits which other spice software (spice#1 &
       spice#2) failed to simulate.
       --- End Quote ---
       Good answer!
       --- Quote from: Rohit link ---
       > I encourage you to run some tests on your PC and post the
       result.
       --- End Quote ---
       I am dealing with significant challenges for the foreseeable
       future, and this might be an interesting exercise but a
       distraction too far.  Besides, my main (AMD Ryzen) machine is
       Linux and MC12 doesn't run well in WINE.  My Win7 machine is
       pretty weedy in comparison with yours.
       *****************************************************
       Page 1 of 1