Trickster Brainscan
02/12/2006, 00:18
eccoti alcuni test sulla modalità multiplayer che ho letto in giro ;)
Interessante test
In my PW all of our external areas are 32x32. The load time is maybe 5-10 seconds longer compared to a 16x16 area, but other than that they are just fine with no complaints about lag. Since the non-walkable area around the outside is always the same width (4), the 32x32 areas are more efficient when considering the percentage of usable area.
32x32 = 64.00% walkable area
24x24 = 56.25% walkable area (13% more RAM for same coverage)
16x16 = 44.44% walkable area (44% more RAM for same coverage)
12x12 = 36.00% walkable area (77% more RAM for same coverage)
08x08 = 25.00% walkable area (156% more RAM for same coverage)
04x04 = 11.11% walkable area (476% more RAM for same coverage)
Using this information, a PW that can be covered with 40 external areas 32x32 each (40,960 grids), would take 160 areas if they were 16x16 and those areas would rquire 44% more RAM due to the inefficiency of that size area.
Combine this with the fact that the players will have to make far less transitions in the external areas and you have a hard time arguing against the 32x32 size areas. Personally, I hope they increase the external size limit to 64x64 or even 96x96 in the future as more computers approach the 4-8 GB of RAM mark.
modalita di test
I am running NWNX4 connected to a MySQL database (on a different server, but on the same network), modified HCR, Knat's PWFX Spawn Engine, and a modified version of Forsetti's Quest Builder. I have load tested with 11 folks in the same area with 50 critters/npcs and combat. There was no noticable performance difference over a 16x16 area.
I couldn't believe it either when we first performed the tests so we did them 3 additional times and compared notes. Again - there were no noticable change in performance over the 16x16 area. I believe tht server dynamics are just vastly different with NWN2 so the we are having to relearn what affects what all over again.
la memoria occupata nel server è circa la meta di quella del file generato dal toolset la buona notizia è il 3/4% dell'uso della cpu
I am seeing that now. I loaded a 1GB (disk space) module into NWServer and my Available RAM only dropped by 430MB.
The server ran fine and we (4 players) noticed no noticable lag as compared to a 100MB (disk space) module. The CPU was operating at 3-6% with 4 players online. But that was not in combat with lots of spells going and so forth.
This is better than I had hoped. If the ratio holds, I should be able to take the module size in disk space up to 3 or 4 GB and still run the module in the 2GB of Available RAM.
I will keep testing and reporting.
test finale
I duplicated a 32x32 area several times and tested the module. Each time, I loaded the module in NWServer and my son, his roommates and I connected through GameSpy (4 total players). The results disproved an assumption I was making that Disk Storage was the same as RAM Usage. The results turned out better than I anticipated. Here are my findings:
1 32x32 areas = .093GB Module size (93MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.38GB OnLoad (60MB RAM for 93MB Storage = 65%)
CPU usage at 3-6% with 4 players online.
5 32x32 areas = .444GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.23GB OnLoad (210MB RAM for 444MB Storage = 47%)
CPU usage at 3-6% with 4 players online.
6 32x32 areas = .532GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.20GB OnLoad (240MB RAM for 532MB Storage = 45%)
CPU usage at 3-6% with 4 players online.
10 32x32 areas = .887GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.09GB OnLoad (350MB RAM for 887MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
12 32x32 areas = 1.064GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.02GB OnLoad (420MB RAM for 1064MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
15 32x32 areas = 1.330GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to .921GB OnLoad (519MB RAM for 1330MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
18 32x32 areas = 1.590GB Module size (89MB/ area) -
Available RAM 1.44GB to .820GB OnLoad (620MB RAM for 1590MB Storage = 40%)
CPU usage at 3-6% with 4 players online.
EXTRAPOLATED MAXIMUMS WITH 2GB RAM
33 32x32 areas = 3.000GB Module size (89MB/ area) -
Available RAM 1.44GB to 0.24GB OnLoad (1200MB RAM for 3000MB Storage = 40%)
CPU usage at 3-6% with 4 players online.
faro anch'io un paio di prove campione per vedere se i conti tornano. Le seguenti conclusioni sono puramente tecniche e non entro nel merito dell'estetica di una mappa 32x32 o della sua vivibilità.
conclusioni, le aree 32x32 sono piu efficienti perchè contenggono una maggior area utilizzabile ed utilizzano meno memoria, che se ne dica non ci sono notizie di maggior lag rispetto le 16x16 con 11 player nella stessa area e 50 npg che ccombattono, di contro hanno che 33 aree riempiono il server. Purtroppo da questo test risulta che le aree sono troppo grandi in termini di memoria richiesta e raggiungono subito il limite dei 2 giga che mette a disposizione winzoz 32 bit. Quindi sarà necessario passare a windows xp 64 che supporta maggiore memoria. .
Interessante test
In my PW all of our external areas are 32x32. The load time is maybe 5-10 seconds longer compared to a 16x16 area, but other than that they are just fine with no complaints about lag. Since the non-walkable area around the outside is always the same width (4), the 32x32 areas are more efficient when considering the percentage of usable area.
32x32 = 64.00% walkable area
24x24 = 56.25% walkable area (13% more RAM for same coverage)
16x16 = 44.44% walkable area (44% more RAM for same coverage)
12x12 = 36.00% walkable area (77% more RAM for same coverage)
08x08 = 25.00% walkable area (156% more RAM for same coverage)
04x04 = 11.11% walkable area (476% more RAM for same coverage)
Using this information, a PW that can be covered with 40 external areas 32x32 each (40,960 grids), would take 160 areas if they were 16x16 and those areas would rquire 44% more RAM due to the inefficiency of that size area.
Combine this with the fact that the players will have to make far less transitions in the external areas and you have a hard time arguing against the 32x32 size areas. Personally, I hope they increase the external size limit to 64x64 or even 96x96 in the future as more computers approach the 4-8 GB of RAM mark.
modalita di test
I am running NWNX4 connected to a MySQL database (on a different server, but on the same network), modified HCR, Knat's PWFX Spawn Engine, and a modified version of Forsetti's Quest Builder. I have load tested with 11 folks in the same area with 50 critters/npcs and combat. There was no noticable performance difference over a 16x16 area.
I couldn't believe it either when we first performed the tests so we did them 3 additional times and compared notes. Again - there were no noticable change in performance over the 16x16 area. I believe tht server dynamics are just vastly different with NWN2 so the we are having to relearn what affects what all over again.
la memoria occupata nel server è circa la meta di quella del file generato dal toolset la buona notizia è il 3/4% dell'uso della cpu
I am seeing that now. I loaded a 1GB (disk space) module into NWServer and my Available RAM only dropped by 430MB.
The server ran fine and we (4 players) noticed no noticable lag as compared to a 100MB (disk space) module. The CPU was operating at 3-6% with 4 players online. But that was not in combat with lots of spells going and so forth.
This is better than I had hoped. If the ratio holds, I should be able to take the module size in disk space up to 3 or 4 GB and still run the module in the 2GB of Available RAM.
I will keep testing and reporting.
test finale
I duplicated a 32x32 area several times and tested the module. Each time, I loaded the module in NWServer and my son, his roommates and I connected through GameSpy (4 total players). The results disproved an assumption I was making that Disk Storage was the same as RAM Usage. The results turned out better than I anticipated. Here are my findings:
1 32x32 areas = .093GB Module size (93MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.38GB OnLoad (60MB RAM for 93MB Storage = 65%)
CPU usage at 3-6% with 4 players online.
5 32x32 areas = .444GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.23GB OnLoad (210MB RAM for 444MB Storage = 47%)
CPU usage at 3-6% with 4 players online.
6 32x32 areas = .532GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.20GB OnLoad (240MB RAM for 532MB Storage = 45%)
CPU usage at 3-6% with 4 players online.
10 32x32 areas = .887GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.09GB OnLoad (350MB RAM for 887MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
12 32x32 areas = 1.064GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to 1.02GB OnLoad (420MB RAM for 1064MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
15 32x32 areas = 1.330GB Module size (89MB/ area) - Loads and Runs fine.
Available RAM 1.44GB to .921GB OnLoad (519MB RAM for 1330MB Storage = 39%)
CPU usage at 3-6% with 4 players online.
18 32x32 areas = 1.590GB Module size (89MB/ area) -
Available RAM 1.44GB to .820GB OnLoad (620MB RAM for 1590MB Storage = 40%)
CPU usage at 3-6% with 4 players online.
EXTRAPOLATED MAXIMUMS WITH 2GB RAM
33 32x32 areas = 3.000GB Module size (89MB/ area) -
Available RAM 1.44GB to 0.24GB OnLoad (1200MB RAM for 3000MB Storage = 40%)
CPU usage at 3-6% with 4 players online.
faro anch'io un paio di prove campione per vedere se i conti tornano. Le seguenti conclusioni sono puramente tecniche e non entro nel merito dell'estetica di una mappa 32x32 o della sua vivibilità.
conclusioni, le aree 32x32 sono piu efficienti perchè contenggono una maggior area utilizzabile ed utilizzano meno memoria, che se ne dica non ci sono notizie di maggior lag rispetto le 16x16 con 11 player nella stessa area e 50 npg che ccombattono, di contro hanno che 33 aree riempiono il server. Purtroppo da questo test risulta che le aree sono troppo grandi in termini di memoria richiesta e raggiungono subito il limite dei 2 giga che mette a disposizione winzoz 32 bit. Quindi sarà necessario passare a windows xp 64 che supporta maggiore memoria. .