Factorio Forums

For my 4.2.2 8t benchmark, I removed the mods and got rid of flames' test blueprint. I found having mods active actually eat up a little performance and can throw results off. Can you rerun Flame?

guitars11 wrote: ↑

Sat Apr 15, 2023 1:07 am

For my 4.2.2 8t benchmark, I removed the mods and got rid of flames' test blueprint. I found having mods active actually eat up a little performance and can throw results off. Can you rerun Flame?

I use a python script for the benchmark, which forcibly disables all mods

maybe 270k spm is not enough for the reds
0.8 ms is not enough, perhaps everything is placed in the processor cache and the results are distorted

Did a v4.2.2 vs 4.2.2-8t w/ 3 x 10.000 tickt test: average is 582 vs 561 UPS, +3.8% on my computer. Factorio box test was 563 vs 561 (same computer and maps).

I tested some variations for 8 cycle timing (based of v4.2.2). This version was the fastest:
Factoriobox: Red Science UPS Editor - guitars11 v4.2.2 8t-7side test by Belter: 579 UPS
But still, margin of error, no big gain. Still thinking about a perfect or close to perfect timing. NowI have a consistent tile offset.

flame_Sla wrote: ↑

Sat Apr 15, 2023 2:37 am

guitars11 wrote: ↑

Sat Apr 15, 2023 1:07 am

[...] I found having mods active actually eat up a little performance and can throw results off[...]

I use a python script for the benchmark, which forcibly disables all mods

FactorioBox also disables all mods (checked the windows script - it sets a temp file as the mod directory).

hmm. I'm getting 955 ups for the 8t-7side which is a little slower than just the regular 4.2.2 which I got 1029. I'm not sure what would be making yours be a little faster though on your computer.

guitars11 wrote: ↑

Sun Apr 16, 2023 3:53 pm

hmm. I'm getting 955 ups for the 8t-7side which is a little slower than just the regular 4.2.2 which I got 1029. I'm not sure what would be making yours be a little faster though on your computer.

4.2.2 still the winnner for now I keep that 1st, dont worry. Once I got my timing I will do a final measurement. We need multiple measurements and might end up saying these N setup as about the same. We're not there yet

Re-done the measurement w/Factoriobox, 5% slower yes:

v4.2.2
Performed 1000 updates in 1639.598 ms
Performed 1000 updates in 1648.822 ms
Performed 1000 updates in 1674.029 ms
Performed 1000 updates in 1674.616 ms
Performed 1000 updates in 1670.110 ms
Map benchmarked at 610 UPS

vs

v4.2.2 8t-7side
Performed 1000 updates in 1726.519 ms
Performed 1000 updates in 1741.636 ms
Performed 1000 updates in 1741.516 ms
Performed 1000 updates in 1743.960 ms
Performed 1000 updates in 1720.938 ms
Map benchmarked at 581 UPS

I think the discrepancy is that to get more consitent results: I quit Chrome, stop Google Drive and disable AV. I'll update all measurements. Yes, I have Win11.

Edit: redone all numbers, see the first post viewtopic.php?f=193&t=105846

Yep, both was me 4.2.2:
https://factoriobox.1au.us/results?map= ... 055c1b5f06

imho for the benchmark it is better to use PS scripts that set a high priority for the factorio process

My first submission

https://factoriobox.1au.us/map/info/016 ... 43c248b3ee
Offsetting saves a lot of beacons that reduces electric update significantly.

Nice work steve! Good find on how to make it more tileable

Stevetrov wrote: ↑

Mon Apr 17, 2023 1:16 pm

My first submission

Welcome! 590 UPS, added.

flame_Sla wrote: ↑

Sun Apr 16, 2023 8:31 pm

imho for the benchmark it is better to use PS scripts that set a high priority for the factorio process

I'm happy to test anything running on Win11 and giving me more consistent UPS (or timing) numbers. Currently for example I have 590/583/593 UPS for Stevetrov v1 - really annoying, this is not within 1% (even quit Spotify...).

Like FactorioBox as
- the script used is pretty good (no mods, 5 x 1000 ticks)
- easy to repeat
- gives public results
- Linux support for hc users (have not tried)

I don't like that
- I download a script then execute it w/o review
- don't see which savegame it is testing, it should update the window title and also show it in the output

Adjusted Steve's design to use more stubs.

https://factoriobox.1au.us/map/info/79a ... 7d6c7f4211

hm, seems to have a little worse performance than 4.2.2.
4.2.2 has an extra beacon on some of assemblers/furnaces so maybe that is the deciding factor.

On your computer it's faster flame? interesting...

Edit : there are a couple that aren't working fully... I'll have to fix that when I get back

Added v4.2.7. TY for the measurements, here is the discrepancy:

And the times visalized from flame_Sla's CSV file - sorry its not milisecs but microsecs:

updated 4.5 and is now 4.5.1. I found 2 furnaces that were not working as intended, somehow the ore got messed up. This one should be good now and hopefully still around the same performance since it was only 2.

https://factoriobox.1au.us/map/info/90d ... 61c4c4c171

4.2.2 is still just a tad better for me.