Advertisement
Snorflake

buying a csgo server vs hosting one, the ram impact

Jun 30th, 2016
109
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 1.52 KB | None | 0 0
  1. CUserCmd send diff updates
  2.  
  3. you move, look
  4.  
  5. 2x viewangle change - 8 bytes
  6. 2x mousedx/y change - 4 bytes (they are shorts)
  7. 2x move change - 8 bytes
  8. button change - 4 bytes (regardless, its a bitflag)
  9. tickcount change - 4 bytes
  10.  
  11. total - 32 bytes/tick
  12. 4096 bytes/second (4kb)
  13.  
  14. 12 people = 49152bytes/second input (49kb)
  15.  
  16.  
  17. a total CUserCmd on CS:GO is 100bytes.
  18.  
  19. 100*12 = 1200 (aka 1.2kb)
  20. this must be stored in ram, so tracking EVERY player is now only takes 1.2kb of ram.
  21. lets say demos. they write the cusercmd to a file every tick so another 1200bytes every tick used (then freed, so total only uses 1200bytes)
  22. hitreg and other things are also handled serverside, lets chuck in 100mb to be safe.
  23.  
  24. the whole server, on 128tick, is now using 1.2+1.2+100000+49152 kb of ram or, 149mb of ram.
  25.  
  26. let's add in warmod shall we? let's be EXTREMELY generous and add another 100mb. that's still only 249mb of ram MAX it will use.
  27.  
  28. Why would 512mb be any different to 1024?
  29. The argument is it "feels smoother" yet there's no difference in usage. If anything the CPU matters because it can process more, but increasing ram won't change how fast your CPU works.
  30.  
  31.  
  32.  
  33. Case study:
  34. core server using ESL's competitive config
  35.  
  36. 0 players my server is using 138mb
  37. 10 player on and it's using 314mb (this is including bot ai processing) the server feels choppy.
  38. After bumping it up to 1024mb, the same results happened and it still felt choppy. This is because game performance depends on CPU and not on ram.
  39.  
  40. Result: 512mb ram is sufficient.
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement