I upgraded from 32GB to 64GB RAM and compared the speed with a benchmark tool and with Dynamo.
Benchmark tool says RAM is now 10% slower.
Dynamo runtime is now 10% longer (tested on a script that runs 30 minutes).
So I´m a little confused because I was hoping for faster runtimes. I had in mind that it´s pretty much a linear correlation between runtime and RAM but i don´t know if i have this wisdom from own experience from years ago or just read it somewhere on the internet…
Our IT guy say this was expected because more RAM does not make anything faster, you can just run more programs at the same time. On The Benchmark Test I´m 0 from 100, that´s far from OK.
No my tests were not done with dynamo player, I can try that also.
I upgraded because RAM was running on 99% during dynamo runs, thats why I expected improvement.
Edit:
Oh and now with 64GB RAM is running constant on 50%
Can someone with 64gb test if he uses more then 50% when running dynamo?
You need to make sure it is exactly the same benchmarking otherwise the output is flawed. This means same versions/build of revit and dynamo, same amount of packages and exactly the same files utilised within the test process.
For example using dynamo 2.10 on the 32gb and then using 2.0.4 on the 64gb is not a good way to test because later versions have performance enhancements.
The type of ram from channel and timings can effect things speed.
Interesting, while i don’t know 100% how Revit and or Dynamo use memory and if i’m honest i haven’t found the silver bullet for the direction of what the big hitters are for hardware build perfection, you may want to check the following basics and reporting these to the forum.
-Speed of previous ram & New ram (mhz)?
-CAS Latency of previous & New ram (CL)?
-Motherboard Model
Some things to consider about your build before considering this a Dynamo Problem
*Check that your motherboard can handle the new rams mhz else it will be bottlenecking and just running at the slower speed.
*Open the task manager and check that the speed of the Ram matches what you are expecting, if it is not then you may need to enable XMP on the motherboard.
Not just the same versions, graph, and file. But the same background processes running (meaning quit everything else you can). Even the. You are certain to not get the same value(s), so you also need to compare the values of multiple runs (comparing the min, max, median, and mean over the course of a dozen or so tests) at multiple points in time.
Basically benchmarking requires either automation to do the runs (ie: journal playback), or some form of tool to recreate exact conditions at the same time…
This post can help with the former. For the later…
OK, so first the problem were the used slots for RAM. 1+3 did´t work properly, 1+2 works now as expected, I´m now getting the double GB/s on multicore processing so everything is fine.
So with high expectations back to dynamo testing. I´m for sure testing the same graph with same Revit and Dynamo versions. The results are pretty much equal at milliseconds, so the test setting is not a problem.
Result: Still 10 % slower with 64GB RAM.
Looking in the taskmanager I can see only 10GB are used by revit, this number increases by about 500mb/minute but without effecting the performance.
Only difference between the old and new RAM is that the old one was dual rank and the new ones are single rank. The machine is a HP z book, there are no bios settings regarding RAM, like XMP.
So I would say that revit/dynamo just can´t use more memory or doesn´t need more memory, so more memory has no effect.
Not quite - the new RAM is clocked at different speeds even in the content which you’re showing. Likely other aspects will differe, but we’ll digress pretty quickly.
Revit has a system cap in terms of the amount of memory it can use, both by it’s installation and by Windows. As you are in a Dynamo for Revit environment, Revt will control the resources available 100% of the time. And so what you want to test is Revit, ignore the Dynamo aspect for a bit. The Revit Forum would be where I recommend you start exploring such testing. Alternatively you could test Dynamo Core, as that would be a bit more consistent in terms of performance (lighter application, no worksharing operations, less internet connectivity to slow things down, won’t default to utilizing a the HDD for scratch disc/element expansion, etc.).
Did tests with 32GB and 64GB, at this benchmark the 64GB setup performs better!
So Revit gets faster but Dynamo gets slower with more RAM…
32GB
RVT 2023 - Full_Standard set - 2023.11.09 @ 21.43.19 on AT-5CG3130K2N.txt
RFO Benchmark v3.3 (build 09.05.2023)
RevitForum.org
All times are in seconds, lower is better.
Run on Revit 2023.1.1
_________________________________________________________________
Update
8.46 update previous version file
_________________________________________________________________
Model creation benchmark
5.64 opening and loading the custom template
11.42 creating the floors levels and grids
12.98 creating a group of walls and doors
19.49 modifying the group by adding a curtain wall
8.41 creating the exterior curtain wall
6.23 creating the sections
2.61 changing the curtain wall panel type
10.80 creating area plans
2.19 creating and applying view template
79.77 TOTAL
_________________________________________________________________
Export benchmark
40.52 export all views as PNGs at 300 dpi
38.45 export all views as DWFs
74.10 export all views as DWGs
35.49 print all views as vector *
62.15 print all views as raster *
Notes:
* Print Views tests require the 'Microsoft XPS Document Writer' printer be installed.
_________________________________________________________________
Render benchmark
44.33 render
_________________________________________________________________
Graphics - Standard View
5.93 activate View Styles view
0.32 change view to Wireframe
0.50 change view to Hidden
0.59 change view to Shaded
0.60 change view to Consistent Colors
3.35 change view to Realistic
0.54 refresh Wireframe Line view x10
1.86 refresh Hidden Line view x10
1.92 refresh Shaded view x10
1.87 refresh Consistent Colors view x10
2.03 refresh Realistic view x10
0.25 activate View Cube view
1.67 rotate view x1
21.43 TOTAL
Notes:
Graphics benchmark measures the entire graphics stack, which includes CPU and memory.
To meaningfully compare graphics cards, test all cards in the same machine
and use the 'Graphics - expanded' benchmark.
TEST CONDITIONS:
_______________________________________________________
Mfr: HP
Model: HP ZBook Fury 16 G9 Mobile Workstation PC
OS: Microsoft Windows 10 Enterprise 64-bit (build 2009)
_______________________________________________________
CPU0: 12th Gen Intel(R) Core(TM) i7-12850HX
Max Clock Speed: 2.1Ghz
Physical Processors: 16
Logical Processors: 24
Total Physical Memory: 32GB
Controller0ChannelADimm0: 32GB @ 4800Mhz
_______________________________________________________
Graphics Card: Intel(R) UHD Graphics
Graphics RAM: 1GB
Driver version: 30.0.101.1631
Screen Resolution: 1920 x 1200 x bit @ 60Hz (max 60Hz)
DPI Scaling: 125%
_______________________________________________________
Drive Type: Local Disk (NTFS)
Drive Model: MTFDKBA1T0TFH-1BC1AABHA
RAW DATA:
_________________________________________________________________
8.984 7.930 8.477 update previous version file
5.672 5.609 5.648 opening and loading the custom template
11.188 11.336 11.750 creating the floors levels and grids
12.609 12.750 13.586 creating a group of walls and doors
18.672 19.836 19.977 modifying the group by adding a curtain wall
8.312 8.336 8.594 creating the exterior curtain wall
6.094 6.297 6.297 creating the sections
2.711 2.453 2.680 changing the curtain wall panel type
10.570 10.836 11.008 creating area plans
1.633 1.984 2.945 creating and applying view template
40.383 40.625 40.539 export all views as PNGs at 300 dpi
39.914 37.719 37.727 export all views as DWFs
73.250 72.867 76.172 export all views as DWGs
35.492 35.375 35.609 print all views as vector
61.914 62.203 62.320 print all views as raster
53.641 39.797 39.562 render
5.969 5.992 5.828 activate View Styles view
0.320 0.305 0.320 change view to Wireframe
0.516 0.500 0.484 change view to Hidden
0.609 0.586 0.586 change view to Shaded
0.602 0.586 0.617 change view to Consistent Colors
3.305 3.422 3.320 change view to Realistic
0.555 0.523 0.531 refresh Wireframe Line view x10
1.875 1.844 1.859 refresh Hidden Line view x10
1.938 1.914 1.914 refresh Shaded view x10
1.867 1.898 1.852 refresh Consistent Colors view x10
2.219 1.930 1.953 refresh Realistic view x10
0.250 0.242 0.250 activate View Cube view
1.688 1.672 1.664 rotate view x1
64GB
RVT 2023 - Full_Standard set - 2023.11.09 @ 19.12.52 on AT-5CG3130K2N.txt
RFO Benchmark v3.3 (build 09.05.2023)
RevitForum.org
All times are in seconds, lower is better.
Run on Revit 2023.1.1
_________________________________________________________________
Update
8.28 update previous version file
_________________________________________________________________
Model creation benchmark
5.57 opening and loading the custom template
10.28 creating the floors levels and grids
12.00 creating a group of walls and doors
18.68 modifying the group by adding a curtain wall
7.20 creating the exterior curtain wall
5.93 creating the sections
2.35 changing the curtain wall panel type
9.58 creating area plans
1.38 creating and applying view template
72.97 TOTAL
_________________________________________________________________
Export benchmark
40.32 export all views as PNGs at 300 dpi
37.11 export all views as DWFs
72.70 export all views as DWGs
35.32 print all views as vector *
61.58 print all views as raster *
Notes:
* Print Views tests require the 'Microsoft XPS Document Writer' printer be installed.
_________________________________________________________________
Render benchmark
41.14 render
_________________________________________________________________
Graphics - Standard View
5.74 activate View Styles view
0.32 change view to Wireframe
0.49 change view to Hidden
0.59 change view to Shaded
0.59 change view to Consistent Colors
3.27 change view to Realistic
0.52 refresh Wireframe Line view x10
1.83 refresh Hidden Line view x10
1.88 refresh Shaded view x10
1.83 refresh Consistent Colors view x10
2.18 refresh Realistic view x10
0.24 activate View Cube view
1.63 rotate view x1
21.11 TOTAL
Notes:
Graphics benchmark measures the entire graphics stack, which includes CPU and memory.
To meaningfully compare graphics cards, test all cards in the same machine
and use the 'Graphics - expanded' benchmark.
TEST CONDITIONS:
_______________________________________________________
Mfr: HP
Model: HP ZBook Fury 16 G9 Mobile Workstation PC
OS: Microsoft Windows 10 Enterprise 64-bit (build 2009)
_______________________________________________________
CPU0: 12th Gen Intel(R) Core(TM) i7-12850HX
Max Clock Speed: 2.1Ghz
Physical Processors: 16
Logical Processors: 24
Total Physical Memory: 64GB
Controller0ChannelADimm0: 32GB @ 4800Mhz
Controller1ChannelADimm1: 32GB @ 4800Mhz
_______________________________________________________
Graphics Card: Intel(R) UHD Graphics
Graphics RAM: 1GB
Driver version: 30.0.101.1631
Screen Resolution: 1920 x 1200 x bit @ 60Hz (max 60Hz)
DPI Scaling: 125%
_______________________________________________________
Drive Type: Local Disk (NTFS)
Drive Model: MTFDKBA1T0TFH-1BC1AABHA
RAW DATA:
_________________________________________________________________
9.008 8.008 7.828 update previous version file
5.617 5.594 5.484 opening and loading the custom template
10.406 10.273 10.164 creating the floors levels and grids
12.000 12.078 11.922 creating a group of walls and doors
18.656 18.617 18.773 modifying the group by adding a curtain wall
7.273 7.164 7.148 creating the exterior curtain wall
6.086 5.859 5.836 creating the sections
2.328 2.344 2.367 changing the curtain wall panel type
9.633 9.664 9.430 creating area plans
1.430 1.344 1.367 creating and applying view template
40.086 40.414 40.453 export all views as PNGs at 300 dpi
37.086 37.297 36.945 export all views as DWFs
72.641 73.000 72.445 export all views as DWGs
35.266 35.477 35.203 print all views as vector
61.703 61.211 61.828 print all views as raster
39.633 39.469 44.312 render
5.680 5.891 5.648 activate View Styles view
0.328 0.320 0.312 change view to Wireframe
0.492 0.492 0.500 change view to Hidden
0.594 0.602 0.578 change view to Shaded
0.586 0.594 0.602 change view to Consistent Colors
3.320 3.234 3.266 change view to Realistic
0.531 0.508 0.523 refresh Wireframe Line view x10
1.828 1.836 1.828 refresh Hidden Line view x10
1.875 1.883 1.883 refresh Shaded view x10
1.828 1.828 1.828 refresh Consistent Colors view x10
2.141 2.156 2.250 refresh Realistic view x10
0.227 0.258 0.250 activate View Cube view
1.633 1.617 1.641 rotate view x1
