Entity tier list for UPS

[varundevan]

Hi,

this might sound impractical or silly,
Is there a list , that any one have come across , which has the list of entities in increasing order of compute power requirement.

Thanks for your answers.

[mrvn]

Hi,

this might sound impractical or silly,
Is there a list , that any one have come across , which has the list of entities in increasing order of compute power requirement.

Thanks for your answers.

Entity is not enough. Recipe matters too. Note that many entities are furnaces or assembler. So you better look at the entity type.

Best are solar cells because the game adds up all solar cells in an electrical network once and each tick it uses that count to add the right amount of electricity without having to do any work per cell. Pipes (anything fluid( are afaik the worst. Belts change depending on how many gaps there are. Some stuff is really complex to quantify.

[varundevan]

Hi,

this might sound impractical or silly,
Is there a list , that any one have come across , which has the list of entities in increasing order of compute power requirement.

Thanks for your answers.

Entity is not enough. Recipe matters too. Note that many entities are furnaces or assembler. So you better look at the entity type.

Best are solar cells because the game adds up all solar cells in an electrical network once and each tick it uses that count to add the right amount of electricity without having to do any work per cell. Pipes (anything fluid( are afaik the worst. Belts change depending on how many gaps there are. Some stuff is really complex to quantify.

Thanks for the answer,

is there a way to identify weather the ups has improved or not by when i reduce complexity of pipes , nuclear setup etc

i have a tilable design which i repeat again and again in my base,
when i improve the design for better UPS , i could not guess it is effective or not .
the ups / fps changes rapidly , that it is hard to identify or notice any improvement. Are you aware of any workaround ?

[Rseding91]

Thanks for the answer,

is there a way to identify weather the ups has improved or not by when i reduce complexity of pipes , nuclear setup etc

i have a tilable design which i repeat again and again in my base,
when i improve the design for better UPS , i could not guess it is effective or not .
the ups / fps changes rapidly , that it is hard to identify or notice any improvement. Are you aware of any workaround ?

Are you having UPS problems? If not; it doesn't really matter until you do. You have to build a lot of something before you can even measure the differences since each thing takes such an incredibly tiny amount. Even then fluids are all multi-threaded so it's hard to tell without a large-scale build how much time it will even take.

[varundevan]

Thanks for the answer,

is there a way to identify weather the ups has improved or not by when i reduce complexity of pipes , nuclear setup etc

i have a tilable design which i repeat again and again in my base,
when i improve the design for better UPS , i could not guess it is effective or not .
the ups / fps changes rapidly , that it is hard to identify or notice any improvement. Are you aware of any workaround ?

Are you having UPS problems? If not; it doesn't really matter until you do. You have to build a lot of something before you can even measure the differences since each thing takes such an incredibly tiny amount. Even then fluids are all multi-threaded so it's hard to tell without a large-scale build how much time it will even take.

i am happy that you showed up here coincidentally,
i do have UPS problems,
i did a post here , and you took a look already ..
viewtopic.php?p=505013#p505013

[orzelek]

Save game is pretty impressive
I can see few things that might help (no guarantee tho):
1. You are using power switches to control oil processing - it might be tiny since there are not many of them but controlling pumps will be better. (If I recall correctly game doesn't like frequent power switching and out of power entities are not sleeping - out of input resources are).
2. You don't have that much use of productivity modules. Making especially high end goods with 4 of those and speed only in beacons would help you reduce amount of active machines - potential big gain for UPS. It will also reduce incoming resource needs (using productivity on varius stages of production reduces input resources significantly). It might be a lot of work to prepare new designs so thats a tradeoff.
3. Grid based design for trains is not recommended for pathfinder - it leads to more pathfinding. It doesn't seem to hav emuch impact in this save since trains don't take that much compared to belts/entities.

[varundevan]

Save game is pretty impressive
I can see few things that might help (no guarantee tho):
1. You are using power switches to control oil processing - it might be tiny since there are not many of them but controlling pumps will be better. (If I recall correctly game doesn't like frequent power switching and out of power entities are not sleeping - out of input resources are).
2. You don't have that much use of productivity modules. Making especially high end goods with 4 of those and speed only in beacons would help you reduce amount of active machines - potential big gain for UPS. It will also reduce incoming resource needs (using productivity on varius stages of production reduces input resources significantly). It might be a lot of work to prepare new designs so thats a tradeoff.
3. Grid based design for trains is not recommended for pathfinder - it leads to more pathfinding. It doesn't seem to hav emuch impact in this save since trains don't take that much compared to belts/entities.

Thanks
point 1 , i am using circuit to control chemical plants to maintain balance of fluids, i don't know how to do that with pumps . A little insight will be helpful.
points 2 is good and has to be implemented. and yes , i did put in lot of effort in designing the grid and loading stations. i think it can be made more compact with beacons .. but . yes it is big challenge .. i will see what i can do in it.. .
point 3 is valid, i looking for ways to cut the number of trains, but grid is better for compact design, and expansion is better this way .. and i TBH .. i put in lot of effort in this design .. so ...
and may i know how much fps you got in my game file ..

[orzelek]

Ad. 1
You can enable/disable pump in same way as power switch. Disabled one won't transfer fluid.
You can put pump on inputs of oil cracking and then use it in same way as power switch right now.

[varundevan]

Ad. 1
You can enable/disable pump in same way as power switch. Disabled one won't transfer fluid.
You can put pump on inputs of oil cracking and then use it in same way as power switch right now.

you are right , i didn't think of it earlier, and i do not know that out of power entities are not sleeping .. are you sure fo it ?
any may i know how much fps you got in your computer and what processor do you use ?

[orzelek]

Ad. 1
You can enable/disable pump in same way as power switch. Disabled one won't transfer fluid.
You can put pump on inputs of oil cracking and then use it in same way as power switch right now.

you are right , i didn't think of it earlier, and i do not know that out of power entities are not sleeping .. are you sure fo it ?
any may i know how much fps you got in your computer and what processor do you use ?

I get about 42-45 fps on Ryzen 3700X with DDR-3200 memory.
And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

[Koub]

And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

Indeed. out of power entities consume compute time because they could be reactivated any time power comes back, whereas entities gone dormant because output is saturated are just put in sleep mode and start to consume CPU only once there is room for the output.

[varundevan]

I get about 42-45 fps on Ryzen 3700X with DDR-3200 memory.
And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

Thanks for confirmation, and
45 ups seems to be low for a 3700X processor.. i think, it must be because of cache latency that.. which i read about few days ago ..
thanks for sharing

[varundevan]

And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

Indeed. out of power entities consume compute time because they could be reactivated any time power comes back, whereas entities gone dormant because output is saturated are just put in sleep mode and start to consume CPU only once there is room for the output.

thanks koub..

[mrvn]

I get about 42-45 fps on Ryzen 3700X with DDR-3200 memory.
And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

Thanks for confirmation, and
45 ups seems to be low for a 3700X processor.. i think, it must be because of cache latency that.. which i read about few days ago ..
thanks for sharing

DDR-3200 memory is kind of meaningless. There are many different kinds of DDR-3200 memory with different clock speeds that make as much as 50% speed difference.

[orzelek]

I get about 42-45 fps on Ryzen 3700X with DDR-3200 memory.
And out of power entities not sleeping I'm pretty sure of - it was discussed on forums few times.

Thanks for confirmation, and
45 ups seems to be low for a 3700X processor.. i think, it must be because of cache latency that.. which i read about few days ago ..
thanks for sharing

DDR-3200 memory is kind of meaningless. There are many different kinds of DDR-3200 memory with different clock speeds that make as much as 50% speed difference.

I wen't into CPU-Z to check and I think I see why I have quite poor ups on my PC.
XMP profile would suggest memory should run at 3200 with timings 16-18-18 but it seems for some reason actual timings are 22-22-22. So I guess my setup is not working well together.

[Koub]

The performance difference between CL 18 and CL 22 will be infinitesimal, and an order of magnitude smaller than the one you get by getting from, say, DDR4 3000 to DDR4 3200.
Just a difference in CL won't explain a significant difference in UPS.

Also you originally wrote DDR 3200. You missed the most important information : the DDR generation. DDR4 3200 will perform way better DDR3 3200 (not twice as fast, but close to).

[mrvn]

The performance difference between CL 18 and CL 22 will be infinitesimal, and an order of magnitude smaller than the one you get by getting from, say, DDR4 3000 to DDR4 3200.
Just a difference in CL won't explain a significant difference in UPS.

Also you originally wrote DDR 3200. You missed the most important information : the DDR generation. DDR4 3200 will perform way better DDR3 3200 (not twice as fast, but close to).

If my understanding is correct the first number (16 vs. 22) of the CLs is the time between near reads in memory setting the address and getting the data. 22 cycles is a 37.5% slowdown vs. 16 cycles. How is that not causing a significant difference?

[Koub]

CL is just a very small part of the whole cycle of operations for performing one "read".
On the contrary, frequency applies to literally everything. So 5% higher frequency globally means the whole read operation will last close to 5% less (I'm rounding here, it's a bit less in reality), while a 37.5% improvement on CL just applies to maybe 1/15th of the whole operation, so you'll be into the 2.5-ish % improvement.

It's like making your hand 5% shorter will result in a shorter hand than only shrinking your fingernail by 37%, if the analogy fits you.

I do not guarantee the numbers are correct (I'm just an enthusiast who's read a lot, not an expert), but the overall idea is.

[orzelek]

CL is just a very small part of the whole cycle of operations for performing one "read".
On the contrary, frequency applies to literally everything. So 5% higher frequency globally means the whole read operation will last close to 5% less (I'm rounding here, it's a bit less in reality), while a 37.5% improvement on CL just applies to maybe 1/15th of the whole operation, so you'll be into the 2.5-ish % improvement.

It's like making your hand 5% shorter will result in a shorter hand than only shrinking your fingernail by 37%, if the analogy fits you.

I do not guarantee the numbers are correct (I'm just an enthusiast who's read a lot, not an expert), but the overall idea is.

I think it might not be that small. You are missing the fact that this is a whole number sequence for all the memory operations and they all bascially scale of that.
SPD numbers for my DDR4-3200 are: 16/18/18/36/74 while actual ones are 22/22/22/52/74 so there are quite few differences. And that 36 vs 52 is a time between subsequent memory reads so difference seems significant.
I can't try to overclock the memory in BIOS to enter timings manually so I can't test what the actual difference would be.

I think (also only from experience) that you might be overestimating frequency a bit. For example:
DDR4-3000 CL 16 - thats 0.66(6)ns with 16 cycles = about 10.6ns wait
DDR4-3200 CL 16 - thats 0.625ns with 16 cycles = 10ns wait
DDR4-3200 CL 22 - thats 0.625ns with 22 cycles = 13.75ns wait
Frequency difference might actually mean less then expected if all the times are lets say 30% longer with worse cycle timings.
How to validate that any of that I have no idea

[mrvn]

CL is just a very small part of the whole cycle of operations for performing one "read".
On the contrary, frequency applies to literally everything. So 5% higher frequency globally means the whole read operation will last close to 5% less (I'm rounding here, it's a bit less in reality), while a 37.5% improvement on CL just applies to maybe 1/15th of the whole operation, so you'll be into the 2.5-ish % improvement.

It's like making your hand 5% shorter will result in a shorter hand than only shrinking your fingernail by 37%, if the analogy fits you.

I do not guarantee the numbers are correct (I'm just an enthusiast who's read a lot, not an expert), but the overall idea is.

You are certainly right for random memory access as you have to load the right frame, set row and column addresses before you get read any bytes. But from the timing diagrams I've looked at the first CL is the major part for sequential reads. It's just increment column address, wait CL, read data, repeat.

But whatever. Just wanted to point out that "DDR 3200" is far from precise for the memory performance. The CLs are one factor and DDR3 vs DDR4 an even bigger one affecting the actual speed besides the frequency of 3200. Another factor not mentioned yet is the frequency the CPU and chipset on the board actually use. It might have DDR3 3200 ram but actually only use them at 3000.

If you are looking into updating your system check closely what the cpu and board support. It's easy to go overboard on one component and not get any gain.