WIP Beta released BeamNG BananaBenchmark - The visual approach

Disclaimer: I have started this project before I joined the BeamNG-Team and therefore the tool is my private work and has no official ties to BeamNG.


Hello everyone,
today I am presenting you the work of the last few weeks, a tool meant to take a little different approach to the well known Banana Benchmark of BeamNG.

Let's start with some eye candy:

But it doesn't stop here, benchmarks aren't all that much fun without a nice way of analyzing and comparing them:

Now that we have seen that we can access benchmark results online, let's get to the main reason why this is actually the case:
The Online Scoreboard.

Last up are two less interesting screenshots, one is about a little help section with some useful hints and tricks
and the other one shows you the settings page where you can adjust a few benchmark related things.

Development status: Open Beta
Source Code: https://github.com/Diamondback2010/BeamNGBananaBenchmark
License: MIT

Please let me know what your initial impressions are and what you think could be improved just from looking at it.

Downloads:

Installation Instructions:
Extract the files somewhere (like your desktop) and run the BananaBench.exe file.
If you have used a previous version of the app, make sure to remove any trace of it from the BeamNG install directory (ie. the BananaBenchmark folder your copied there)

 

Awesome!

 

I am not entirely sure myself but here's my theory:
The beam/s go up somewhat in a linear fashion, until you run into performance issues. Have a look here:
http://beamng.software-enkler.com/v1/displaybench.php?token=87d1cca445c258

See the two distinct slumps at 3 and 7 vehicles? That's part of the issue from what I can see.
As we all know BNG basically uses one thread per vehicle, so in a perfect scenario, a quadcore processor doesn't care if you are running 1, 2, 3 or 4 cars at once
as each of them has their own thread and each of those threads is handled by one core.
In practice this doesn't work quite out as you never have a truely idle processor, there's always some stuff going on in the background that needs CPU time.

So perfomance wise it's more like we have a tri-core for BNG. Three vehicles are no problem, as we have 3 more or less idle cores that can be used,
but the 4th core is already working on other things, so the 4th car doesn't get quite as much CPU time as cars 1 to 3. That's where the first drop off happens.
Now if we are using more cars than we have cores, each core has to handle multiple cars at once. Some high performance CPUs don't need all their power for a single vehicle per core,
so there's some power left unused. However with more cars than cores, we can use this excess power again so we are seeing a slight increase in Beam/s. Until we reach 8 vehicles,
where one of our cores needs to handle two vehicles + the background stuff, which is usually too much.

So all in all, it's not linear because one vehicle does not use all the power of one core, but two vehicles need too much power for a linear increase.
I guess one could find a CPU / car combination that would translate to a much more linear increase (up until core count -1) but for comparability reasons we need to stick to proven cars.

 

I have just updated the OP with a link to the source code of the project (MIT license) and a download link for the first public beta.

Let me know how it goes and if something goes wrong, make sure to explain exactly what happened when and what pc specs you have.

 

Haha damnit I forgot the installation instructions. Wait a sec...

- - - Updated - - -

Done, please look at the instructions in the OP and make sure to redownload the zip file.

 

Eh what? Can you please express whatever you want to say a little bit clearer?

Are you trying to say that you got 38 MB/s as your result with an i5 (maybe an i5 4690T @2.5Ghz?) which equals about 3 pickups and in reality you can run 5 to 7 T75 at once?
If so, sorry, but I highly doubt this is the case. Even the i7 5930K (heavily overlocked) can run 5 T75 maximum at once. There is no way an i5 @ 2.5Ghz can come even close to that.

But I'm sure I am missing something here, so please explain again what you talking about.

 

Can you provide a link to your benchmark? The number of possible vehicles might need some tuning to fit everyone.

- - - Updated - - -

I see a static image of you having 20fps. Can you show us a video with all those vehicles driving at once?
By all means, in the screenshot you could have deactivated the physics...

Oh and the benchmark doesn't report in %...

 

Not talking about pictures here, I need a link to the benchmark details you can get it if click on the details button below the score.
Concerning the score, maybe energy saver mode?