XFlow: meshless CFD is here



  1. Particle based CFD is promising and it is very encouraging to see MSC bet on it. I wonder if you have any application demos for coolant flow in electric motors.

    1. Hi Joe, I don’t think we specifically have an example of that, but it would certainly be great for it. Please contact me offline if this is an active project you’d like to look at: jeff@jeffwaters.com

  2. Particle based CFD has been developed for quite a few years now, basically in the fluid motion rendering arena with a remarkable visual appeal both for free surface motion and smoke/fog movement.
    It is not surprising indeed that the xflow developers are the same as the Maxwell Render, one of the best rendering packages for fluid motion applications.
    That said, it comes natural to ask for verification and validation studies in order to assess the suitability of the method for engineering applications, which demand reasonable accuracy rather than visual appealing.
    Are these studies available?

  3. Hi Vmax, thanks for the comment. Yes, validation documentation is available. Please contact me offline to discuss your application and relevant validation data.

  4. Hi Jeff,
    I saw the high lift benchmark on the website. It’s very nice to see such a high profile benchmark where others compare themselfes on real world examples.
    XFlow did a nice job there but I have two questions on the benchmark and XFlow in general.
    As far as I know Xflow is not distributed memory capable (so for Clusters or networked computers), does it support GPU since usually LBM is pretty good for GPU.
    Also there is the test computer mentioned in the paper on the XFlow benchmark. From the paper one can read that the RAM necessary was around 16GB as the resolution had to be switched, but I wonder how long the computation took for this case.
    I know LBM has a good linear scalability and it is also clear that the time for the calculation is different between the different angles of attack but can you tell what the min/max time was for the calculation to get a feeling for other codes I used?


    1. Hi Anton,
      I don’t know how long that took to run, but I believe we’re presenting more completely on this in the upcoming AIAA meeting (around middle of the year I think?)

      We are putting the finishing touches on the cluster version of XFlow now. Release is very soon. This will allow scalability on Linux or Windows clusters.

      XFlow does not currently take advantage of GPU, but this is definitely something we’re looking at for the future.