Getting higher than 600 user throughput Gamazon

[FlamingTomato]

So I have a design for Gamazon with one server splitting into two sorting blocks, and it can support about 600 users it seems (that's how many 'yellow' users there are). But there are another 20000 users in red, so a massive overload of the server. I've tried parallelizing with three servers going into 4 sorting blocks, which seems faster but doesn't improve the number of yellow users at all. And beyond that I don't know what else I could do to improve.

Also I feel like the feedback when building for startups is very unclear - I see absolutely no overloading going on in my design when testing it. I would love to be able to test the system under pressure and see what's causing so much overload.

[Pakl]

I have no idea either how this is supposed to work. With the given node limit my best was to fulfill around 1100 users (at a huge loss, mind you) but as the users grow the throughput decreases for me to around 400.

I suspect there are quite a few bugs left in that system. I updated my expensive highly parallized gamazon (because it was bleeding money) with a simpler version to save on cost and the next day I got -450 yellow users and 11K reds.

I don't see any information in the game how many users my design would serve or why some users are not served.

[Delth]

Agreed - parallelizing seems to simply not work for that one. I managed to get to 700 users with the help of DLL trick (use Custom Nodes wherever possible, copy your design into DLL, use that DLL in another DLL, and use the final DLL in the actual design), but that's it, doubling the amount of servers the most correct way possible (one "solution" vs parallelizing node splitting into two identical copies of "solution") doesn't change a thing.

[Dangerous-D]

Here's what I came up with a while ago, I tried to cut the cost but that cuts the users.

I'd like to think that this could be improved but I don't know yet.

https://steamcommunity.com/sharedfiles/filedetails/?id=1629721210
https://steamcommunity.com/sharedfiles/filedetails/?id=1629721069

[kaibioinfo]

You could replace the red C/S/T sift by a single sift checking for triangles.

PS: I did the same scheme without custom nodes (cause I didn't knew about this custom node ♥♥♥♥♥♥♥♥). It's unbelievable how many more users your scheme can support and how much more money you receive.

[spityou]

here is something that is not only for this one startup but for a lot of stuff if you can fit in the power it needs. it will sort red blue and green with zero mess ups and with a big load and with a little load it will do it fast and with not that much back up (depends if you have a few hundred) and you can config it so you can combine two colors or 3 if you want here is a photo https://imgur.com/a/uhPSoPR

[Yes-Man]

Recent comments from other users (in other threads) lead me to believe this issue is still not fixed. It has been over 1.5 years ... Can't be that hard :/

[rybikson]

I will just leave it here :)
Theres an easy way to reduce the balancer count in the whole project by tuning the usual 3-output "Quick C/S/T Sifter" to "Quick C/S/S/T/T" sifter and avoid putting another balancer to have extra squares and triangles on the output. Same with Quick Sorters.

[Smock]

I've come up with a solution that gets 9500+ customers and around $4k+ for several weeks

https://www.youtube.com/watch?v=_ZpLXQzPVZM

I'm up to about 8-10 weeks with $100k+ ready to cash out and it's still going strong

https://steamcommunity.com/sharedfiles/filedetails/?id=2798067122