Results 1 to 10 of 10

Thread: Big campus project crushing our PCs - need benchmarking tools

  1. #1
    Member Duncan's Avatar
    Join Date
    April 28, 2012
    Location
    Denmark
    Posts
    95
    Current Local Time
    10:17 AM

    Not allowed! Not allowed!

    Big campus project crushing our PCs - need benchmarking tools

    Hi there

    I'm involved in a large (600.000 m2) campus hospital project. Some of our users PCs are collapsing under the strain of our models. I have some ideas as to what we could do but so far all my suggestions are met with skepticism and concern over the effort involved. We are not even using worksets to control opening/closing the 20+ linked models ...

    What I need is some tools I can use to benchmark the system strain of running the same model set up in different ways. Then I can show them "If we unload most links and just show a simple representation with model lines the model file will perform like this...", "If our links were on worksets and could be closed the model file would perform like this..."

    So what sort of tools are suitable for this and what do I need to know to judge the accuracy of the results? Feel free to just point me to other threads if this has been discussed.

  2. #2
    Member need4mospd's Avatar
    Join Date
    March 2, 2011
    Location
    Houston
    Posts
    481
    Current Local Time
    03:17 AM

    Not allowed! Not allowed!
    There's no need for a benchmark. All links go on a separate Workset, and links are loaded/unloaded by opening and closing Worksets. My experience with a 4,000,000 s.f. campus says that is the ONLY way to do it. I go a step further and change each file to open directly to the Specify Worksets dialogue by default. Do this by doing a "Save As" in each central file, selecting options, and changing the Open Workset Default to Specify.

    If you want to benchmark something, record the time to load a file with only 3 links loaded, and a file with all 20 links. Multiply that times your number of users, and a typical 2 loads a day(restarting Revit is a must with a large project).
    Example: One of our larger files with multiple links would take 20 minutes to open completely(on a good day), while selecting just the links you needed would take 2-5 minutes depending on the links. 15 minutes * 15 users * 2 loads per day = 7.5 HOURS PER DAY! So one person would be working just to catch up that missing load time!

    On top of that, each computer should be running 16Gb MINIMUM RAM and they should be the fastest(Ghz) i7 processor you can afford.

  3. #3
    Moderator cellophane's Avatar
    Join Date
    August 9, 2011
    Location
    Louisville, KY
    Posts
    6,234
    Current Local Time
    04:17 AM

    Not allowed! Not allowed!
    I second what mospd said but if you want to benchmark have a look here: http://www.revitforum.org/hardware-i...benchmark.html

  4. #4
    Forum Co-Founder iru69's Avatar
    Join Date
    December 7, 2010
    Location
    United States
    Posts
    1,649
    Current Local Time
    01:17 AM

    Not allowed! Not allowed!
    Quote Originally Posted by cellophane View Post
    I second what mospd said but if you want to benchmark have a look here: http://www.revitforum.org/hardware-i...benchmark.html
    I really don't think that's the kind of benchmark that Duncan is going to find helpful. If he wanted to really get crazy-sophisticated about it, he could take that benchmark and use some of the programming strategies to develop a "benchmark" program around their specific project... but that's really unnecessary as need4mospd points out.

  5. #5
    Administrator Gordon Price's Avatar
    Join Date
    December 7, 2010
    Location
    Rotterdam, South Holland
    Posts
    3,070
    Current Local Time
    10:17 AM

    Not allowed! Not allowed!
    Duncan, I am going to pile on with the "Benchmarks won't help here" comments. The RFO Benchmark certainly won't help, as it is testing a VERY small model, and does nothing at all with regards to worksets and the like.

    Probably the tool that will get you the most milage is actually Task Manager. On the memory tag you should be able to monitor how much RAM is used vs free, with the different approaches to model management. It is worth noting that for the most part RAM has no impact on performance. If you have enough RAM, more RAM will not make the machine faster. And if you don't have enough RAM, then speed isn't the issue, crashing is. But maybe, just maybe, you are right in that transition, where lack of ram triggers thrashing of the swap file, but does not yet cause crashing, or at least consistent crashing. Task Manager can help you monitor that.

    Also, I should mention an experience I had some years ago, which may shed some light, and will also expose just how risky decisions based on benchmarks can be. Anyway, about 5 years ago I was using Phil Read's original AU Benchmark to test a bunch of machines, and what I found was that an entry level graphics card had exactly the same performance as a higher end card that cost about 5 times as much. And we all knew that Revit was using all of 5% of a graphics card's actual capabilities anyway (still "mostly" true today). So it made some sense that 5% of a high end card or 25% of a low end card, what difference does it make, most of the card is just being wasted. And based on that, we spec'ed the low end cards. And the results where as expected, no major problems. UNTIL... a really massive project reached late phases, and the model was just graphically super heavy. "Over modeled" by most arguments, but a "rightly modeled" project is just a project that has been gimped by low end hardware and Revit performance issues. Anyway, what was discovered is that at the very extremes of model complexity, a cheap @$$ card just couldn't cut it. Upgrade to the higher end card and suddenly things got MUCH better. Still nothing like what you would expect given the card's raw performance, but still proof that you need a good card. As in, a GTX-5xx or 6xx series card, not some junker NVS card, but also not $3000 worth of Quadro. And yet, the benchmark is going to tell you that all three perform basically the same. thus the big fat caveat in the benchmark with regards to GPU numbers, and my attempts every year to make the GPU test actually useful rather than just cute. This year is a little better than last, but still not numbers that you want to take to your CFO to validate a purchase decision.

    What does that mean for you? Well, first I would track the RAM, on a machine that is handling things fine, and one that isn't. And if the RAM doesn't pan out as an issue, then I would switch graphics card between machines and see if that helps. And if the problem machines are so old they can't handle the newer graphics cards, then that might be the problem. because all the model management effort just means that adequate hardware goes from taxed unnecessarily to over burdened. But no amount of model management is going to make inadequate hardware get the job done. Of course adequacy is a fluid thing, so it might help if you posted your machine specs for both the viable machines and the problem children.

    And one last thing to check. Any chance there are DWGs in there? A little bit of DWG will shag Revit performance much more than a massive linked model. Revit actually does a pretty good job of handling linked RVT files efficiently. But a DWG is a problem waiting to happen. A single workset JUST for all the DWGs may be the quick answer to better performance.

    Best of luck,
    Gordon

  6. #6
    Member Duncan's Avatar
    Join Date
    April 28, 2012
    Location
    Denmark
    Posts
    95
    Current Local Time
    10:17 AM

    Not allowed! Not allowed!
    (Edit: written before Gordons post showed up)

    I think you're all on the right track. I have been using RFOBenchmark for testing, but it's only good for comparing how well different machines handle the RFOBenchmark Revit Project. The projects I'm working on are considerably bigger - especially in the number and size of linked projects.

    What I find problematic with need4mospds suggestion is that I will just be told by the more tech savvy users 'yeah, but did you have the same programs running when you started the test as when you finished. And anyway how did you time this, did you sit still and not use your PC for 15 minutes with a stopwatch?'

    I have the impression that something in between is needed - I've tried creating journals based on my project files, but had no success so far. Perhaps if I read the load times directly from the journal file? That should be possible right?

    At the moment thinking of suggesting two things, neither of which are new ideas

    1. All links are placed in three worksets, typically closed and not visable. And we create specific 3d views to check workset disciplin.
    1. Workset for all building files in this project
    2. Workset for all building files in neighboring projects
    3. Workset for all 2d links (DWG), could be combined with one of the others
    2. The rough geometry of neighboring buildings is drawn in as needed with one model line type for plan views and one for elevations, each filtered invisible in the other view

    Opinions?
    Last edited by Duncan; October 24th, 2012 at 04:26 AM. Reason: (Edit: written before Gordons post showed up)

  7. #7
    Member need4mospd's Avatar
    Join Date
    March 2, 2011
    Location
    Houston
    Posts
    481
    Current Local Time
    03:17 AM

    Not allowed! Not allowed!
    Quote Originally Posted by Duncan View Post
    What I find problematic with need4mospds suggestion is that I will just be told by the more tech savvy users 'yeah, but did you have the same programs running when you started the test as when you finished. And anyway how did you time this, did you sit still and not use your PC for 15 minutes with a stopwatch?'
    Start the testing right after bootup, before you start any other programs. And yes, you basically sit there with a stop watch. Show up an hour early and test it if you don't want to be seen "wasting time". Do the math when you get some results, tell them it will save X number of hours for approximately Y number of hours invested in fixing it. When you show the results in dollars and man-hours saved, people tend to take notice.

    Pick out some quotes from these Autodesk documents that mention it to help your case.
    http://images.autodesk.com/adsk/file..._tech_note.pdf
    http://images.autodesk.com/apac_grtr...ory_en_v20.pdf

    I'd like to know what some of the concerns are from the others. This is really an industry standard procedure we're talking about here.

  8. #8
    Member Duncan's Avatar
    Join Date
    April 28, 2012
    Location
    Denmark
    Posts
    95
    Current Local Time
    10:17 AM

    Not allowed! Not allowed!
    Quote Originally Posted by need4mospd View Post
    I'd like to know what some of the concerns are from the others. This is really an industry standard procedure we're talking about here.
    Thanks for that and those links, I'd forgotten about that second document.

    What the 'BIM Manager' is concerned about is people putting things on the wrong workset. So I'll really need to show him how easy it is to make some 3d views for cleaning that up. But in reality it will seldom be a problem as the worksets we're talking about will never be set as active and therefore never be inadvertently used to put objects on. I can even add a Filter to our VTs which makes use of the link worksets very visable... hadn't thought of that before.

  9. #9
    Member need4mospd's Avatar
    Join Date
    March 2, 2011
    Location
    Houston
    Posts
    481
    Current Local Time
    03:17 AM

    Not allowed! Not allowed!
    Quote Originally Posted by Duncan View Post
    But in reality it will seldom be a problem as the worksets we're talking about will never be set as active and therefore never be inadvertently used to put objects on.
    Bingo.
    Not to mention the Worksharing Display in 2012+ can color objects by their Workset with the click of a button.

    Your BIM Manager is basically saying "My users don't get this concept, and rather than train them properly, I'll be lazy and continue along a path that doesn't really work."

  10. #10
    Forum Co-Founder Twiceroadsfool's Avatar
    Join Date
    December 7, 2010
    Location
    Dallas, TX
    Posts
    10,533
    Current Local Time
    03:17 AM

    Not allowed! Not allowed!
    Quote Originally Posted by Duncan View Post
    (Edit: written before Gordons post showed up)

    1. All links are placed in three worksets, typically closed and not visable. And we create specific 3d views to check workset disciplin.
    I wouldnt do it exactly like this. I personally would give every single Link its own workset. And i make every single workset always visible by default. I know "a lot of people" like to make them not visible by default, but imvho its just an opportunity to accidentally hide things that shouldnt be hidden. Since you can turn off Links in a VT under VG:RVT Links, doing so in a Workset is just a slightly faster, but hugely more error prone way to achieve a task that already had another way to do it. Yeah, you can have "check views." But if you can discipline everyone enough to check before every plot, maybe you can discipline them enough to not mess it up in the first place. Good luck with that. So for us: Links on individual worksets for demand loading and unloading, and all worksets always visible (by default, and in VG).

    1. Workset for all building files in this project
    2. Workset for all building files in neighboring projects
    Im not sure what you mean by "Building Files," unless you mean Modeled elements. But seperate worksets for seperate buildings that are in the same file, makes sense.

    3. Workset for all 2d links (DWG), could be combined with one of the others
    I would still put them each on their own workset, for demand loading and unloading. Its often the only way to troubleshoot when a cad file gets jacked and breaks your model.

    2. The rough geometry of neighboring buildings is drawn in as needed with one model line type for plan views and one for elevations, each filtered invisible in the other view

    Opinions?
    I wouldnt bother with this at all. Where you need to see the neighboring buildings, youll see them. View range, and View extents, take care of that.

Similar Threads

  1. Model management for campus site - splitting up a model.
    By macroen in forum Architecture and General Revit Questions
    Replies: 3
    Last Post: June 6th, 2012, 12:48 PM
  2. Revit Express Tools - Revit Family Tools, and lots more.
    By RevitExpressTools in forum Commercial & Free Add-Ins/Extensions
    Replies: 0
    Last Post: September 28th, 2011, 04:33 PM

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •