Not the problem.

The research is consumed. The research points are awarded. The production screen reports the number of flasks consumed incorrectly.

The labs are designed to be able to eat something like 12000 of each pack per minute to clear backlogs of science.
I don't see any evidence of that.

36 labs * 350% research speed * 70% module speed / 60 research time = 1.47 packs per second = 88.2 packs per minute.

That is a lot lower than 12000 ...

Initial post updated. Further clarified and simplified and a savegame included.

Easy to check!

20180520031641_1.jpg

The game's consumption screen reports 111/m

Measuring from the feed belts shows 150/m


Where I said 12000/min before was a typo btw. My numpad 0 key likes to stick sometimes and I just didn't notice it. I meant 120/min.


That being said, neither the game's ...

Your science consumption graph shows sharp drops every 12 minutes or so. Your pic shows the situation during the "normal" period, but what happens during the drops?

It's also suspicious that frequency of those drops matches the frequency of white science production. Could your labs experience ...

An additional test -

As shown earlier, a circuit network setup reads that the labs are getting handed 109.2/min for each pack, but the consumption graph disagrees. So which is correct? Perhaps they are both correct and science packs are getting lost at the labs somehow?



My understanding is that ...

As pointed out, the expected bottleneck is the red/black production at 109.2/min

The labs eat science a bit faster than I can make it. That way, if there's a backlog of built up science packs, it will eventually be cleared. For example, if I disable research for a while, or forget to press T to ...

Note the difference between consumption measured in the game's production/consumption window and that which was measured directly by the circuit network. Probably a bug. Will wait for others to confirm before reporting.

The math I have done on the research points getting generated strongly suggests ...

I found the tutorial problem you mention to be poor specifically because it does not give you enough chain signals to properly signal the whole junction.

You can solve the tutorial problem with a bit of simplification though. Just treat the whole mess past the straight rails as a single block like ...