Page 1 of 46 1234511 ... LastLast
Results 1 to 10 of 459
Like Tree16Likes

Thread: RFOBenchmark

  1. #1
    Administrator Gordon Price's Avatar
    Join Date
    December 7, 2010
    Location
    Berlin!
    Posts
    2,208
    Current Local Time
    10:33 PM

    RFOBenchmark

    RFOBenchmark (2015) is here!
    This is an extension of the AUBench tool, with some added tests, verified function for all verticals, and modularized to make adding tests and porting to new versions of Revit much easier.

    Please post your results in
    RFOBenchmark 2011
    RFOBenchmark 2012
    RFOBenchmark 2013
    RFOBenchmark 2014
    RFOBenchmark 2015

    Some things to be aware of...
    • The 2015 (and revised 2014) Benchmark will launch and close a non full screen session of Revit at the beginning of execution, to ensure that the follow on tests can be run fuel screen and with Coordination discipline default and to preload any required DLLs and ensure a more accurate test..
    • The 2015 (and revised 2014) Benchmark automatically sets the default view Discipline to Coordination, to address the RST crash, and make the resulting test more of an equal comparison between verticals.
    • Be aware that this DOES have a material impact on performance. Coordination views are rather faster across the board, so if you want to compare 2014 to 2015, make sure you get the revised version of the 2014 test. Both are build 5.7.2014.
    • The 2014 benchmark automatically runs for the appropriate vertical of revit installed on your machine. If you have multiple verticals it will run for all of them automatically.
    • If you are testing RVT (i.e. OneBox) AND you have ONLY the Structural UI enabled, answer no to the popup. This will run the OneBox test as if it was RST 2014.
    • The benchmark runs full screen automatically now.
    • The results file is automatically named with a date/time stamp, name of machine and vertical being benchmarked.
    • The RFOBenchmark folder doesn't need to be renamed. Just copy the folder from the ZIP to your disk and run the _RFOBenchmark.vbs file. You will then be asked what vertical you want to test, progress will be shown in a popup between tests, and a final popup will let you know the benchamrk is done and let you open the results file. 2013: With 2013, you have more vertical options. RVT is OneBox, and RVT-S is OneBox with ONLY Structural UI enabled.
    • 2013: The new test also asks if you want to run full screen or not. PLEASE run full screen. The impact is not trivial. if we all do full screen tests, we can then look at the screen size information reported in the hardware section and make much more meaningful comparisons.
    • The test can ONLY be run via the VBS file. Running the journals manually will not produce the results you want. Simply double click the _RFOBenchmark_2011.vbs or _RFOBenchmark_2012.vbs file or _RFOBenchmark_2013.vbs file.
    • The test is longer than before. Up to 15 minutes. I think the expanded data is worth it, and since most people won’t be running it regularly, hopefully this won’t be a problem.
    • Read the readme. Especially the part about using _Kill Test.vbs
    • The test attempts to quantify GPU performance. However, the results you see are in no way a complete picture. Over time we will see what the data tells us. It may be that the results are in fact misleading, and we decide to remove that part of the test. In the mean time, use the benchmark as a datapoint for decision making, but remember that nothing is as meaningful as a butt-dyno comparison of two different cards, on your machines, on your projects.2013: The new test expands a bit on the GPU test, which when combined with running full screen is getting close to being a meaningful metric. Not there yet, but closer!
    • On the GPU test, you will notice that Consistent Colors is very fast on the non hardware accelerated test in 2011, but on a par with the other views in 2012. This is because CC doesn’t do shadows in 2011, so the major hardware acceleration benefit is lost. In 2012, where shadows are available in Consistent Colors, the results match expectations.
    • 2013: Lastly, I regret to say that they new test will not run in 32 bit Windows XP. If I can find a way to make this happen I will update ASAP, but for the moment it looks like that is not going to happen. That said, I hope that Win XP x32 has become an edge case and this won't affect too many people.

    Please keep all discussions about the tools in this thread, and use the Result thread ONLY for results. One last note. Some folks have had issues with the downloads being corrupted. If you have this issue, I recommend you try the download with a browser that is NOT Internet Explorer. Maybe IE9 is better, but even IE8 sometimes results in corrupted download. Chrome and Safari for Windows have both worked well for me, and others have reported good results with Firefox.Thanks all, I hope you find RFOBenchmark useful.Gordon
    Attached Files Attached Files
    Last edited by Gordon Price; May 9th, 2014 at 11:30 AM. Reason: Added Revit 2015
    mawlsam, rjmarquez, Dvi and 2 others like this.

  2. #2
    Administrator Munkholm's Avatar
    Join Date
    December 7, 2010
    Location
    Kingdom of Denmark
    Posts
    4,046
    Current Local Time
    10:33 PM
    Great job Gordon! Rep. for you

    I┤ll post my results shortly

  3. #3
    Moderator
    "OMG I killed Revit"
    dzatto's Avatar
    Join Date
    December 9, 2010
    Location
    Dallas, TX
    Posts
    5,019
    Current Local Time
    03:33 PM
    It still throws my work machine into interactive mode at the exact same spot the other one did.

    Would that be a corrupt download, or is that some other issue? All I have is IE8 at work. IT won't let me use chrome or firefox.

  4. #4
    Administrator Munkholm's Avatar
    Join Date
    December 7, 2010
    Location
    Kingdom of Denmark
    Posts
    4,046
    Current Local Time
    10:33 PM
    Dan, I┤ve just sent the Benchmark by mail... maybe that┤ll work...

  5. #5
    Administrator Gordon Price's Avatar
    Join Date
    December 7, 2010
    Location
    Berlin!
    Posts
    2,208
    Current Local Time
    10:33 PM
    Gah, I was really hoping for a nice clean working benchmark for everyone. Dan, can you PM me with a copy of the journal file that was made as the benchmark ran? I'll see if that sheds any light.

    Gordon

  6. #6
    Forum Co-Founder iru69's Avatar
    Join Date
    December 7, 2010
    Location
    United States
    Posts
    1,303
    Current Local Time
    01:33 PM
    Very slick! Fun stuff.

    Quote Originally Posted by Gordon Price View Post
    • The test attempts to quantify GPU performance. However, the results you see are in no way a complete picture. Over time we will see what the data tells us. It may be that the results are in fact misleading, and we decide to remove that part of the test. In the mean time, use the benchmark as a datapoint for decision making, but remember that nothing is as meaningful as a butt-dyno comparison of two different cards, on your machines, on your projects.
    Unfortunately, I don't think this part of the test is saying anything about the GPU (video card). The test that uses the HWA is so minimal that every video card is going to be about the same. The test that doesn't use the HWA is pretty much just benchmarking the CPU (as far as I would guess).

    p.s. - sorry I didn't think to mention it when you first had me take a gander at it (I wasn't really paying any attention to what was actually being benchmarked, only that it didn't crash).
    Last edited by iru69; March 24th, 2011 at 08:22 PM.

  7. #7
    Administrator Gordon Price's Avatar
    Join Date
    December 7, 2010
    Location
    Berlin!
    Posts
    2,208
    Current Local Time
    10:33 PM
    Iru69,
    I am sure you are correct. The numbers are useful, in that they tell you that shadows and HA go hand in hand. But beyond that I think we will want to have a discussion here about what the numbers really mean.

    BUT, I find it VERY interesting that Consistent Colors AND Realistic View are faster than Hidden line, with and without shadows, with and without hardware acceleration (I have run those variations of the test seperately, in 2011 and 2012, and will do a writeup of the results soon). My guess is that the "new" view types where done with optimized code to start with, and the Hidden Line view still uses older (original?) unoptimized code. I know people who specifically avoid Consistent Colors thinking hidden line must be faster, so just for that data point I am happy with the GPU test. But it isn't saying anything about the performance of a particular card, which I had hoped for. And thus my extended caveat on the GPU test specifically.

    Gordon

  8. #8
    Moderator
    "OMG I killed Revit"
    dzatto's Avatar
    Join Date
    December 9, 2010
    Location
    Dallas, TX
    Posts
    5,019
    Current Local Time
    03:33 PM
    Sure thing. Let me try the one Munk sent.

    Well, it hasn't shown up yet. So.......uh...........how do I attach to a PM or a RFO email for that matter? lol
    Last edited by dzatto; March 24th, 2011 at 08:44 PM.

  9. #9
    Administrator Gordon Price's Avatar
    Join Date
    December 7, 2010
    Location
    Berlin!
    Posts
    2,208
    Current Local Time
    10:33 PM
    Well, I thought you could do it in a PM as well. But here in the thread, click the Go Advanced button, then you will get a slick little Attach Files section lower down the screen. Click Manage Attachments, Add Files, Select Files. Choose your file the click Upload File.

    Thanks,
    Gordon

  10. #10
    Moderator
    "OMG I killed Revit"
    dzatto's Avatar
    Join Date
    December 9, 2010
    Location
    Dallas, TX
    Posts
    5,019
    Current Local Time
    03:33 PM
    Yeah, I knew I could do it in a thread, but you said PM. LOL
    I wasn't sure if it was because you didn't want everyone to see it for some reason.

    Here it is. Have fun! lol
    Attached Files Attached Files

Page 1 of 46 1234511 ... LastLast

LinkBacks (?)


Similar Threads

  1. RFOBenchmark 2012 results
    By Gordon Price in forum Hardware and Infrastructure
    Replies: 113
    Last Post: April 11th, 2013, 04:22 PM
  2. RFOBenchmark 2011 results
    By Gordon Price in forum Hardware and Infrastructure
    Replies: 52
    Last Post: December 26th, 2012, 08:40 PM

Posting Permissions

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