Stats

  • Locked Locked
  • Replies 4
  • Subscribers 125
  • Views 15514
  • 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

APS: better Hyperthreading ON or OFF on the machine?

dontpanic

Hi! We recently got some new servers, and IT asked if they should enable or disable Hyperthreading on them (which effectively doubles the number of cores available for simulation). In the past I've tried APS in servers both and without APS, and it seemed to me that the gain in number of cores available made it totally worth using Hyperthreading, in the context of large sweeps and/or Montecarlo simulations (which we do a lot). However I am unsure how much this penalizes the speed of single-run simulations (using APS).

I found an 8-year old post about this, and I am wondering if since then a clear position has developed on the advantages/disadvantages of using Hyperthreading for APS simulations (in particular, of ADCs). Does anybody have some opinion/experience/information on this?

Thanks and regards,

Jorge.

  • dontpanic
    in reply to Nasser Arif

    Thanks Nasser, that helps to clarify the issue. I guess we'll leave hyperthreading OFF as it seems to be the official recommendation, although a 10% loss of performance for a 100% increase in number of parallel simulations you can run seems a good deal to me! Or am I missing something here?

  • Andrew Beckett
    in reply to dontpanic

    Hi Jorge,

    That's not what the article says. Note that floating-point centric applications such as circuit simulators cannot really take advantage of hyper threading because there is only a single floating point unit. 

    So even if it's turned on, you cannot expect to run double the number of parallel simulations - it doesn't work that way.

    Also, with SPECTRE16.1 and 17.1 we no longer recommend  you turn it off - it's handled within the simulator - this is stated at the bottom of the article. However, that still doesn't mean that you can expect to throw twice as much simulation work at the machine - it just means that at least the hyper threading shouldn't get in the way.

    Regards,

    Andrew.

  • dontpanic
    in reply to Andrew Beckett

    Oh, I see. I was running lots of parallel APS simulations in the server with hyperthreading ON and it seemed to me that each point was running as fast (if not faster!) than with hyperthreading OFF! And definitely I could get a big advantage in the number of parallel simulations I could run before red-lining the server! Is there any official Cadence appnote / whitepaper / release note / etc that I can show our IT guys to convince them that hyperthreading is no longer discouraged?

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.