Testbench for 3/4way intersections

[hansjoachim]

In v5.0.6

Set 2 RHT 3-way is different from set 2 LHT 3-way. East coming trains prefer to go west for RHT instead of south.

Thanks, I'll fix that. Basically means all results with 3 way on 5.0.6 is invalid.

[hansjoachim]

In v5.0.6

Set 2 RHT 3-way is different from set 2 LHT 3-way. East coming trains prefer to go west for RHT instead of south.

It's now fixed with 5.0.7.
I have a new update underway, where I redo all the schedules, but I want to have a stable working version as a backup.

[hansjoachim]

Major update 5.1.1
New map with 0.0 closer to the center of the bench.
Editor extension instead of creative mod
Now 3-way can be placed in any direction and tested correctly

[AristomachosCZ]

Hi,
why does the testbench map allows only a specific maximum game speed (probably 10x)? If I have a several dozens work-versions of intersection and do a three test per one, it can take a lot of time. The editor extensions mod, that is required (if I am right), can boost the game up to 64x game speed. It would be nice, if the test map would support this speed.
Thanks.

[Avona]

why does the testbench map allows only a specific maximum game speed (probably 10x)?

Hello!
I am able to go up to 64x speed in the test bench (and have always been able to I believe).

[hansjoachim]

Hi,
why does the testbench map allows only a specific maximum game speed (probably 10x)? If I have a several dozens work-versions of intersection and do a three test per one, it can take a lot of time. The editor extensions mod, that is required (if I am right), can boost the game up to 64x game speed. It would be nice, if the test map would support this speed.
Thanks.

The testbench doesn't affect console commands at all. You can set game.speed = 100000 if you want. The limiting factor for how fast the tests runs is your PC. The editor mod isn't essential, but it's convenient.

[hansjoachim]

v.5.2.2
Added decimals. Makes the lamps go crazy=)

[AristomachosCZ]

Well, I did some testing and I can say the following:

a) I don't have to care about the test or game performance on my gaming computer. Always 60 FPS (vsync on) in this scenario, regardless of the game speed.

b) I ran the auto test – that should have according to the display about 56 minutes in real time with basic game speed (1x) – several times and I got these results (with my stopwatch).

game speed 8x - test lasted 7 minutes 8 seconds
game speed 10x - test lasted 5 minutes and 54 seconds
game speed 64x - test lasted 4 minutes and 55 seconds (performed twice for sure)

Game speeds 8x and 64x were set by Editor extensions mod, 10x in console.

c) I don't know what is wrong, but I have expected those times will be in mutual accordance. Because look at this:

game speed 1x = 56:00 (about 56 minutes)
game speed 8x = 7:08 – expected 56/8 = 7 (7 min) => corresponds exactly
game speed 10x = 5:54 – expected 56/10 = 5,6 (5 min 36 s) => it's near, but ok
game speed 64x = 4:55 - expected 56/64 = 0,875 (52,5 s) => wrong

d) In any case, this isn't meant as a criticism, I like your work, I like your idea. I only try to point at there is something that – probably – works incorrectly. At least for me.

[Avona]

game speed 64x = 4:55 - expected 56/64 = 0,875 (52,5 s) => wrong

Your computer can only run the simulation so fast. Factorio can only be run as fast as your computer can compute the algorithms. If you set the speed higher, your computer *attempts* to run the game faster, but can fail to do so as you found out. To get higher speeds than you currently do, you would need to upgrade your computer.

[AristomachosCZ]

Sorry, but did you read what I wrote under a) ? I have a gaming computer. During these tests I had max. 50% CPU usage, only 6 or 7 GB from 32 GB RAM and max. 30% GPU usage. I also have a fast SSD (PCIe NVMe SSD). What upgrade do you think I need, when Factorio doesn't even use a full potential of my computer?

game speed 64x = 4:55 - expected 56/64 = 0,875 (52,5 s) => wrong

Your computer can only run the simulation so fast. Factorio can only be run as fast as your computer can compute the algorithms. If you set the speed higher, your computer *attempts* to run the game faster, but can fail to do so as you found out. To get higher speeds than you currently do, you would need to upgrade your computer.

[Avona]

What upgrade do you think I need, when Factorio doesn't even use a full potential of my computer?

Your point a) was about fps, not ups, therefore it isn't applicable (graphics are not the limiting factor for Factorio; it's a 2d game).

If the computer is not running the game at the stated ups, it is not the fault of the test map, editor extensions mod, or the console command. The game attempts to run at the stated ups and if it fails, it runs the game at the highest ups that it can. A large calculation intensive map will run at a lower maximum ups. The intersection test map is very efficient, especially the latest versions, since extra entities were removed and the map as a whole was made smaller. I believe the editor extensions mod's ability to change game speed is just a user friendly version of the console command.

I don't have the technical knowledge to determine the limiting factor for you. Check out the Technical Factorio Discord server. I have seen them talking about the computer limitations before and their answers are not obvious to me (meaning that it's not so simple as "cpu fully utilized", etc). It's very complicated. My understanding is that the full potential of some part of your computer is being used and that's what is causing the limitation.

It could also be that your computer cannot be upgraded further and that you are seeing the maximum ups possible on the test bench. Also, programs running in the background will negatively impact performance.

[hansjoachim]

Sorry, but did you read what I wrote under a) ? I have a gaming computer. During these tests I had max. 50% CPU usage, only 6 or 7 GB from 32 GB RAM and max. 30% GPU usage. I also have a fast SSD (PCIe NVMe SSD). What upgrade do you think I need, when Factorio doesn't even use a full potential of my computer?

You have to check the load on individual threads some of them might max out. If they do your CPU is the bottleneck. If not it's probably your RAM speed/(bandwidth) or latency. RAM speed and latency has been mention by the devs as a bottleneck. You can read this post, by a dev, about it.
https://www.factorio.com/blog/post/fff-204

[AristomachosCZ]

I know about the UPS, but I didn't mention that, because the changed game speed impacted the number a lot (from 60 to several hundreds). So I had not idea, what is correct, or what is caused by the mod.
About the performance, yes, maybe ram frequency could be higher and as it is usual with the games, a single thread performance of cpu.
Anyways, thank you guys for your willing and quick reactions.

[hansjoachim]

Version 5.3
Fixed TPM, Now it's way more accurate and works for every train length, also it's lane dependent preparing it for having different TPM per lane. Now trains are the same colour from all directions in every set before there was some unintended mixing. Also some minor weirdness with set 3 blinking when you are testing with low TPM.

[nOObe]

5.3.1
set to 7
The mod Automatic Train Deployment (0.1.11) caused a non-recoverable error.
Please report this error to the mod author.

Error while running event AutomaticTrainDeployment_Updated::on_train_changed_state (ID 25)
__AutomaticTrainDeployment_Updated__/control.lua:108: attempt to index field 'schedule' (a nil value)
stack traceback:
__AutomaticTrainDeployment_Updated__/control.lua:108: in function <__AutomaticTrainDeployment_Updated__/control.lua:32>

[Avona]

The mod Automatic Train Deployment (0.1.11) caused a non-recoverable error.
Please report this error to the mod author.

Kano may see it here, however you should probably post mod problems to the mod page to make sure:
https://mods.factorio.com/mod/Automatic ... nt_Updated

[hansjoachim]

5.3.2
Major chage.
Now with manual mode, you can set TPM per lane and have separate trains per lane with different schedules, inventory and length. This will allow you to test any intersection with the load you want and you can replicate your own base. Now its truly an every intersection tester.

[hansjoachim]

Here is the version with your 8-way correctly on and running.

[hansjoachim]

v5.3.3
Fixed an issue causing the bench to reset to set 1 when using set 5. Caused by the 3/4 way checker.

[hooray]

Will the train U-turn be included in the test?