Stats

  • Locked Locked
  • Replies 5
  • Subscribers 126
  • Views 17709
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Hardware for Best Simulation Performance?

SteveMarshall

 We are upgrading our workstations and are wondering if Cadence software will gain a performance advantage if processors with Hyperthreading are used. We have to choose between the i7 and i5 processors. Most work done in the lab is corners and monte-carlo analysis of analog circuits, in the Analog Design Environment. The simulator is spectre.

 I am not too hopeful that hyperthreading will give a performance benefit, even to highly parallelized tasks such as Corners analysis, but if it does it would be a huge benefit. If the simulation engine can take advantage of hyperthreading, please let me know.

 

Thanks!

  • Spectre doesn't support multithreading. Do you have APS (Accelerated Parallel Simulator)? It is one of MMSIM7.1 or MMSIM7.2 simulators.

  • Actually spectre does support limited multithreading. Spectre turbo uses more still.

    However, it's APS which is really taking advantage of multi-core machines.

    I have seen some benefit with using hyperthreading with APS because the threads are not always fully utilized. However with running multiple parallel simulations (such as over corners) you'd generally be better off with real multi-cores rather than hyperthreading because they're much more likely to be able to run flat out at the same time.

    However, I don't personally have much experience of hyper-threaded machines, so don't take my reply as an official recommendation from Cadence - above is just a personal viewpoint.

    Regards,

    Andrew

  • Don't use core and it's hyperthread for the same simulation task. It will slow down your simulation time (but CPU usage will show higher percentage). But if you will run two simulation job at the same time you can assign one job to different processors core and another simulation job to their hyperthreads. In this case you will get the best performance.

  •  Thank you for the replies.

    I ran a very rudimentary test: The same simulation on two machines, with multithreading both enabled and disabled.

    On a dual-core Opteron processor, the simulation took longer with multithreading ENABLED

    On a quad-core i5 processor, the simulation took longer with multithreading DISABLED

    I did not specify a number of threads when enabling multithreading. The test was a transient simulation of an analog circuit performed in the Analog Design Environment. There were no complex calculator functions involved; I only measured certain node voltages and performed no calculations on them.

  • I also would like to check another thing.

    I found out that CPU frequency is changed during simulation because of CPU frequency scaling feature. Currently my setting for frequency scaling governor is "userspace" which changes CPU frequency on demand. But there is another scaling frequency governor "performance" with max CPU frequence regardless CPU loading. I am going to verify if there any benefit in simulation time if I change governor from "userspace" to "perfofmance".

    Does anybody tested that already? 

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.