Autodesk Flame – SpeedCheck

2022 – updating for flame 2023…

It has been a while since I updated this page. I checked on WordPress when I initially created it. This was on the 30th of December 2015.

A lot of time has passed since then and lots of cool new flame versions were released. I was not working so much on flame in the last years, that’s why I also lost a bit of track of the benchmark results that were hosted on the Logik Facebook group. And I also forgot to update the Flame-Archive with the benchmark and check it with new releases of flame. The Flame Logik Facebook group is history now.

So I was also late to join Logik.tv and the user forum. I finally joined in November 2021 as a Patreon, because I was so thankful to find the “Logik Academy” videos about the “Connected Conform” by Jeff Kyle.

This year I will work more on flame again and I am lucky to have one of the new shiny M1-Max MBPs which will run Flame at some point. I plan to host an updated archive on this site again and maybe some benchmark results as well. The most recent benchmark results you can find in the forums at forum.logik.tv.


History – Why does this benchmark exists

Last time updated in 2018

Some years ago still working on Linux flames only there was coming up the 2nd gen. Smoke on Mac.
We were in the middle of a giant project and were looking for a simple test to see how quick was the Smoke vs. Flame.

The clip I used was from the project and therefore not a good clip I could share or re-use easily.
So I created a test setup based on a 1000 frames of HD noise.
The test result looked ugly and the archive was still very big because of the base clip.

When we finally changed from flame 2013 to 2015 and skip 2014 entirely I came up with a much smaller setup based on a single black frame and three tracks of timeline and BFX setups. This still was created for a 1000 frames timeline, but I added so many things in each BFX that the render times were getting too high. So I trimmed down the Sequence to 250 frames.

There are some issues with the test depending how the tracks are selected in the sequence and how the preferences are set in flame. Maybe I can come up with a failsafe way how everyone always gets the same results.

The discussions can be found on Facebook in the FlameUsers – Logik Group.
There is also a link to a google docs sheet with lots of results from different sytems.

Flame2016_frame_based_render

The actual version of the file archive is written with flame 2018.
Please set the Prefs/Timeline/Rendering to “Frame Based Render”.

Flame2016_Select All

In the active Sequence – Select All

Flame2016_render

Then start a stopwatch and hit render. As you can see on the screenshot the time estimation is completely off. The same as the top frame count. When the lower one reaches 250 of 250 frames the render is done. On this MACPRO 12Core D700 64 GB (Yosemite 10.10.5) the test takes about 3’35” to finish in FRAME BASED render mode.

If you follow the instructions and set the timeline render mode to TRACK BASED the render time will be twice as long (6’46”). If you flush again the render and deselect everything in the timeline and hit “RENDER” instead of “RENDER Selected” my render time goes down again to half the time 3’14”, but this time only the topmost layer V1.3 is rendered. I don’t get why this is, therefore I prefer the FRAME BASED mode to the TRACK BASED mode.

Here is the zip file for download.

I don’t recommend to download it though. It is outdated. I will post an updated link soon.

SpeedCheck_2018

Speed Check 2018 – rendered on a M1-MAX MBP. Yeah!
This cup I got on one of the last Friday parties at IBC in Amsterdam. How much I miss Amsterdam and the IBC! Maybe 2022?