Teeworlds - general

Just a curiosity: I compared resources consuming of Teeworlds and DDnet. I noticed surprisly that DDnet use 3x more ram memory (around 750 mb, while Teeworlds only 250 mb) and 10 times more CPU usage in percentage on an intel i3-1125g4 chip (approximately 20%, while teeworlds only 2%) I guess in the evolution of its development DDnet became heavier along adding new features and/or adopting different libraries What do you know about?

that²s more of a question for ddnet

carmine

Just a curiosity: I compared resources consuming of Teeworlds and DDnet. I noticed surprisly that DDnet use 3x more ram memory (around 750 mb, while Teeworlds only 250 mb) and 10 times more CPU usage in percentage on an intel i3-1125g4 chip (approximately 20%, while teeworlds only 2%) I guess in the evolution of its development DDnet became heavier along adding new features and/or adopting different libraries What do you know about?

If u compare cpu usage u also have to compare fps, frame time, input polling rate

10:08

But yeah it uses more ram

10:09

Generally tho using vk backend ddnet can get more performance out of your hardware

10:09

Because it's much more efficient cpu wise

Test made with vsync on, 60 fps

Didn't matter. Ddnet has cl refresh rate

So, if I set an optimal value to that variable I get better results? (edited)

cl_refresh_rate 60; gfx_refresh_rate 0; gfx_asyncrender_old 0; gfx_backend vulkan Restart

10:46

Also note that in any way u cannot compare ddnet anyway. It fixes vanilla stuff such as using mipmaps for tilesets and used a higher quality sampling approach

The two codebases have functionally diverged a little too much for this to be a meaningful comparison to be honest. We do a lot more prediction, we handle text differently, we render differently

carmine

So, if I set an optimal value to that variable I get better results? (edited)

What did the results say?^^

still an interesting comparison IMO

RAM usage is mostly due to rendering method i think

It should at max only use double the ram per map So from 250 to 750 mb its likely to be other stuff too. Or the memory is kept hot internally and not truely deallocated

14:06

It also depends heavily on the map and menu background map

14:08

On pure igpus it might be the missing vram

Jupeyy_Keks

What did the results say?^^

Benchmark improved, cpu usage went down, now very similar to Teeworlds, ram didn't changed...

16:16

But I was offline so I played locally, I don't know if much stuffs run if find an available internet connection (edited)

carmine

Benchmark improved, cpu usage went down, now very similar to Teeworlds, ram didn't changed...

i can only test with 240 fps cap bcs thats my vsync value but my results show this:

teeworlds between 25-30%
ddnet between 15-20%

16:48

(i used steam versions for both btw)

interesting... however is better to leave cl_refresh_rate to 0

just set gfx_asyncrender_old 0 to the job to improve performance, even with openGL

on the Tutorial map ddnet with openGL 1.4 use 350 megabyte Ram and 3% cpu, now is fine!

22:55

20% gpu, just a simple integrated, not a dedicated card

22:56

on my PC openGL seems better supported than Vulkan that use almost double amount of ram and the same cpu/gpu (edited)