Moderator: Forum Guards
(active networking objects?)
INJECT USERFLAG 1
When connected to a server displays statistics once per tick.
i=20 cli=3 act=4.3 (640) net=2.3 pv/c=80 re p/c=0 rp c/c=0
i=Server tickrate, Cli= Number of Clients, act= Time spent on processing Actors (ms), (640) number of actors, net= Time spent on network processing (ms). Assuming a tickrate of 20, if act + net = 50 the server is running at 100% CPU utilization ( 1000ms / 20 = 50 ). If cli is greater than the number of players in the game the server most likely suffers from the win2k creeping ping bug. Note: INJECT USERFLAG 0 turns off output.
Dejavu wrote:http://wiki.beyondunreal.com/Legacy:UT_ConsoleINJECT USERFLAG 1
When connected to a server displays statistics once per tick.
i=20 cli=3 act=4.3 (640) net=2.3 pv/c=80 re p/c=0 rp c/c=0
i=Server tickrate, Cli= Number of Clients, act= Time spent on processing Actors (ms), (640) number of actors, net= Time spent on network processing (ms). Assuming a tickrate of 20, if act + net = 50 the server is running at 100% CPU utilization ( 1000ms / 20 = 50 ). If cli is greater than the number of players in the game the server most likely suffers from the win2k creeping ping bug. Note: INJECT USERFLAG 0 turns off output.
do test on your own and you will understand what i mean.
Dejavu wrote:i mean, test in game and you will see that server with less actors uses less processing time.
put on server on another box, and join DXMP_Cmd map. remember the values for act and net.
then modify DXMP_Cmd, remove all actors and join on modified map, and compare act and net values to previous ones
then you will understand what i mean.
Longc[A]t wrote:I still think Dae is a russian spambot.
~[A]Daedalus~ wrote:There will be a day when my patience goes away and you, along with all who rant with you, will get banned.
ô¿ô¥[GODZ]¥NOCHANC wrote:I can ban any one I want ANY time I want. You have no rights here.
~[A]Nobody~ wrote:Make a cube and 2 spawn points and you got your optimized map.
Dejavu wrote:sorry about my terminology alex(active networking objects?)
that what i meant. of course not all objects in MP are being updated constantly. brushes and shit like that gets initialized on start only. then they are not touched anymore. i hope you understand now what wanted to say.
E8300 is not so special just as it is. 2,83ghz = 333 FSB x 8,5 multi, with 6 meg L2 cache. however, it still reaches around 16,5 sec with super pi 1 meg benchmarking (ive reached around 20 sec with OCed E8300, and my CPU ran at same FSB, the only important difference was that my CPU had only 2 meg L2 - here you can see how important L2 is).
E8300 @ 4ghz (500 FSB x 8 multi) reaches 11,5 sec in super pi 1 meg. show me 1 cpu that is capable of getting such super pi result with stock settings. no xeon cpu is capable of that. you can spend 1000 € on xeon cpu and yet, you still wont get out such performance.
just for reference, pentium 4, 1,8ghz reaches 1min 30 sec super pi 1 meg, pentium 4, 2,66ghz reaches 1min, amd xp 3200+ around 50 sec. of course, super pi runs on single core only since its single threaded benchmarking test. you can get super pi here: http://www.xtremesystems.com/pi/
and test 1 meg on your cpus and you will see what im talking about. this tool is for pure cpu speed testing.
what has OS version to do with dxmp hosting? i think the most important is to use as little as possible services and to use win32 nt based OS for max reliability (win2k and up).
4 or 6 or whatever number of dxmp servers, it will still beat every other dxmp server. and considering there are almost never more than 30 players playing dxmp at the same time, he could eventually put unlimited number of dxmp servers on (as long as there are enough RAM resources avaliable).
Alex wrote:Destroyerzero wrote:Thanks
I'll apply it next server boot. I still want to test under the settings I have
Remove DXMTL160.
And as for the hardware, I * think * each gameserver runs on it's own core. And I'm afraid you can't compare pings from other games with Deus Ex, as there's a big difference between 'ping' and how it 'feels'.
Destroyerzero wrote:Alex wrote:Destroyerzero wrote:Thanks
I'll apply it next server boot. I still want to test under the settings I have
Remove DXMTL160.
And as for the hardware, I * think * each gameserver runs on it's own core. And I'm afraid you can't compare pings from other games with Deus Ex, as there's a big difference between 'ping' and how it 'feels'.
Pardon me for double posting, so I will keep this short...since last post was a wall of text.
When you run a program on Windows XP, the processor affinity marks all cores by default. If a single core program assigned to 4 cores, it copies scripts and wastes memory. DX is a single core program.
Launch a Server, then go to start/run/taskmgr, to the processes tab, click on DeusEx -----> Processor Affinity and you will see what I mean...Uncheck all but one box.
INJECT USERFLAG 1
When connected to a server displays statistics once per tick.
i=20 cli=3 act=4.3 (640) net=2.3 pv/c=80 re p/c=0 rp c/c=0
i=Server tickrate, Cli= Number of Clients, act= Time spent on processing Actors (ms), (640) number of actors, net= Time spent on network processing (ms). Assuming a tickrate of 20, if act + net = 50 the server is running at 100% CPU utilization ( 1000ms / 20 = 50 ). If cli is greater than the number of players in the game the server most likely suffers from the win2k creeping ping bug. Note: INJECT USERFLAG 0 turns off output.
Destroyerzero wrote:All my tests on Windows XP 64-bit and Ubuntu 64-bit have given me 12 - 30% more performance when running the same 32-bit programs on a 64-bit Environment.
Secondly, there is only one real place inside of a computer that has a millisecond response time in information....that is a hard drive itself. You can eliminate the millisecond and convert it to nanoseconds by throwing the whole server into a Ramdisk.
Time spent on processing Actors (ms), net= Time spent on network processing (ms).
Very bad example, since Super Pi itself is a program that generates a Synthetic result and not true real world data. The program is integer based and doesn't use processor registers to calculate its data...so its just a RAW SPEED TEST on how fast it can fetch from System Memory....So its more of a memory test than a processor test, but the morons try telling you otherwise.
DejaVu wrote:try to avoid putting in useless shit like crates and actors that have no use in multiplayer.
Alex wrote:I once counted ALL objects in Smuggler. (or Area51Bunker, cant remember which of the two). You know how many I counted?
Up to 54.000.
Dejavu wrote:Destroyerzero wrote:All my tests on Windows XP 64-bit and Ubuntu 64-bit have given me 12 - 30% more performance when running the same 32-bit programs on a 64-bit Environment.
are u claiming that your dxmp server runs 12-30 % server faster on 64bit OS?
if its so, then you are again full of shit.
tell me, how can an old game such as dx, which has been compiled for 32bit cpus benefit running on 64bit. i would understand that, if the dxmp itself has a lot of memory management need to do where 64bit addressing of it gives you more space.
but naturaly, 64bit registers offer no better performance to program, which is using only 32 bit registers. you would have to recode whole dx to work with 64bit registers.
and since, the 32 bit program is running on 64 bit environment, lets not forget that kernel mode is in 64bit. so each time dx calls any WIN API that has its roots in kernel mode, the system has to switch between 32 bit mode and 64bit mode, then after finished work in kernel mode, switch back to 32 bit mode. this effect can even slow down 32 bit program on 64 bit OS!!!Secondly, there is only one real place inside of a computer that has a millisecond response time in information....that is a hard drive itself. You can eliminate the millisecond and convert it to nanoseconds by throwing the whole server into a Ramdisk.
again im wondering if you ever even knew "inject userflag" command.
the quote itself says:Time spent on processing Actors (ms), net= Time spent on network processing (ms).
where is any talk about response times? its time needed for processing various stuff.
eg.:
variable start saves current time
do work
...
read current time and substract start variable and show the result as act or net
this has nothing related to diskdrive, its access time etc. ramdisk is useful only for WINAPI calls like CreateFile, ReadFile, WriteFile etc.
if you have a lot of phisical ram memory, then its best to disable paging file and you force win to load all his stuff directly into memory + programs load all needed stuff into memory (not using HD).
ramdisk is therefore useful only on server start and at each map load, it wont reduce net and act time at all.Very bad example, since Super Pi itself is a program that generates a Synthetic result and not true real world data. The program is integer based and doesn't use processor registers to calculate its data...so its just a RAW SPEED TEST on how fast it can fetch from System Memory....So its more of a memory test than a processor test, but the morons try telling you otherwise.
im gonna debug it and see if thats fully true. ive not seen any program in my life thats not using cpu registers.
claiming that program uses no registers is very stupid, as you know that registers like EIP, ESP,... are always used/changed no matter what.