Multi cpu / hyperthreaded flashpoint server s/w

[darkpeace 0]

Multi CPU / HyperThreaded Flashpoint server s/w

Will it ever happen ?

~~~~~~~~~~~~~

Why all these USA admins are running Dual 2.8 Xeons is beyond me, surely they check the CPU load on the server ?

Or is there a magic trick or did someone edit, then recompile the source code for the server to make it more optimized on the multi CPU / threaded side of things ?

Also with the recent updates I have seen a few servers cap themselves at 32 fps, is there anyway to make this 50fps or more ?

2 servers:

-Intel Pentium 4 2.4/800/HT @ 2.9/1066/HT

-'Borrowed' Intel Quad Xeon 2.66/533/HT @ stock speeds

-All CPUs have 512kb l2 cache, and no l3 cache

-The P4 has a higher FSB, and memory benchmarks are better in SiSoft SANDRA because of this

The Xeon makes a better file-server than the P4, but the P4 makes a better Flashpoint 1.96RC server. Surely you could squeeze more usage from all those idle CPUs ?

Well done on the memory usage though guys, that was quite impressive.

Someone ?

And does anyone know, (and is willing to answer), if VBS1 servers are more optimized thread wise for 8 logical CPUs under Linux ? - I have asked around and been given conficting reports, I suspect that VBS1 server is NOT any faster, or uses any more CPU than Resistance 1.96RC (unless they removed the 50fps or 32fps limiters. Suspect it is no faster on the Xeon above, than Resistance was)

Anyone with FACTS ?

Regards,

DarkPeace

[RN Malboeuf 12]

Simple

OFP Server is single threaded soft ware

running a dual 2.8 our self hurts us compared to our old 3.0@3.6 with 1 gig ram and 960 fsb

the p4 just had brute force power, i have listend to guys with linux and heard guys brag thier xeon server is better.

the're not. in fact i think they hurt OFP when they are linux based as well

reason is proof in the cpu usages, we have a dual with HT which shows 4 CPUs and you'll see it start share loading by switching the server thread from 1 CPU to the other, two of the cpus are HT based and are not really fully powered 2.8s

during that trade off i'll notice player skipping, its not lag since we have the fasted avaible conect at 100 Mbits and OFP only uses 5 mbits

its CPU over load on the HT side of the cpu

we have the Xeon since we also run a high end Web server off of it, OFP does not hamper any web fuctions since there are 3 other cpus to use, plenty of ram, and ban width.

Win32 based servers are more stable and heavly supported compared to linux, but as the same time linux is better when runing a web server.

next month our goal is to switch over to 2 P4's 3.0s or 3.4 and over clock the shit out of them but im at a loss as to when OFP2 will come out. 1 server will be win32, the other server wil be linux for the web aps - this would actually save us $50 US a month and allow 2000 mbits instead of 1000 mbits (OFP may use 60 gigs in a month) xfer per month and give us 4 p4s 3.0 or faster

the fact that if OFP2 server IS multy threadable would mean i would again switch back to Xeon because at that point they would kill the P4 ten times over with out trying.

so knowing this i keep putting off the change over and our OFP CTI maps suffer because of it, the lack of raw power and FSB sucks ass big time

hopefully biz can tell us

[benu 1]

As ofp2 will be out in a year or something, why base server decisions on it now? You can always switch servers in a year from now...

[darkpeace 0]

Thanks too:

~~~~~~~

RN Malboeuf,

I thank you for your honest and frank reply,

Also cheers for the info on the player stuttering being related to HyperThreading, those logical CPUs are no where near as fast as a real one (load balancing on single core).

That is logical, but I never tried it before.

In return, although you prob already know this, there is a command "MinErrorToSend=0.001" you can add to flashpoint.cfg, which may help ALOT on the MFCTI maps,

(It gets a mention in the change log that no-one reads)

Enjoy,

Other Dodgy Comments:

~~~~~~~~~~~~~~~

So you net admins just run Xeons cuz you host web servers, etc, aswell....makes sense, bummer 'bout the speed hit.

Soon enough I'll get one of those PreScotts to a damn high speed, on a FSB with RAM to suit (PC4000 or so, etc)

Although the Athlon 64FX, may end up being a good choice (I suspect flashpoint benefits from short pipeline CPUs, with fast x87/FPU units, and makes some basic use of MMX, and maybe SSE- Athlon XP2000+ was beating my higher clocked P4s at the time, but its long gone)

Well good thing I am a FlashpointLAN gamer, so we can just use a high FSB, highly clocked P4 []

I am bookmarking this one for others.

Still, where are the coders offering a threaded server ?

(I ain't paying AU$250+ for VBS1, but I would pay $250 for a well threaded Flashpoint: Resistance server update for MFCTI use at LANs, etc)

[darkpeace 0]

Now is anyone going to try and tackle the 32fps vs 50fps server performance issue, without palming it off to Windows timing, or Quanta's as their known ?

Also does a server at 100%, running at 28/32fps mean that if it was 50fps idle, it would run at 44/50fps or 28/50fps ?

(There are many fools out there who have no idea, I am prepared to read alot of replies until I get the fix I am after)

Has anyone been able to re-create the problem, then isolate and fix it ?

Has anyone made a reghack, or other simple solution available to server admins ?

Bummer about the threading, least now I am more certain of why a P4 Northwood - 2.4/800@2.9/1066 beats a Quad Xeon 2.66/533 (uugghhh half the FSB). - and yes I run uber RAM at 166/200th of the FSB speed, as I can't find any 533Mhz DDR1 DIMMs anywhere with decent timings.

Also with the server frame rate, even when playing basic maps, like Pistol DM with 5+ people, it would be nice to see #MONTIOR 1 report 50fps, and SFA CPU usage, rather than 32fps and even less SFA CPU usage.

[aka: HLDS PingBooster, http://www.udpsoft.com, where is Resistance 1.96RC PingBooster I ask?, expecting no reply]

I'd love to see the server at 60% or more load running the server physics engine (well thats almost all it does really) at say 120+ or even 300-600 fps

Also a threaded server could handle much more AI, and even animals running around and the like (free ones, not extracted from VBS1)

Well I wonder if the great mystery will ever be solved ?

[RN Malboeuf 12]

what does the server report in lobby or briefing? it should be 50fps, 37 in game seems fine

as for 50 fps and watching the CPU usage,it wont matter what it tells you really, OFP just uses as much of the CPU it can for CPU's lower then 2.8 ghz (on small maps), then you may see 65% on a small map with faster CPUs and some time you will see 60-99% on CTI maps.

[RN Malboeuf 12]

As ofp2 will be out in a year or something, why base server decisions on it now? You can always switch servers in a year from now...

Cost efficency was the reason, why pay for two seperate servers (web, game) when 1 server will do both (reason why we got the Xeon), but then we found better pricing that would double our lines from a 100 mbit to x2 100 mbit and xfer rate of 1000 gigs per month to 2000 gigs per month simple because they are on two lines and two servers for $50 less a month

Since two RN have web based Hosting Comapines it was cheaper to combine all three needs into 1 server, but now this new deal is too hard to pass up so we'll be going back to 2 P4s

CTI and HT based Zeons were just not ment to be, as soon as i realized Linux was share loading a 100 maxed OFPserver thread to the HT part of the CPUs (you can watch it doing it) i was pissed since the HT end of it just can not handle it

picture a garden hose of data flow and you squeezing it for 40 seconds every 40 seconds and tell me it does not cause dysync, i noticed this dysync in heavy load CTI maps where our old P4 3.6 would have rocked along with out any unless a player lagged on his end and then only he would get dysync

but dont forget the older 2.4/2.8 xeons with out HT, they wont have this problem but i still dislike the share load linux does with an active server but thats minor

[darkpeace 0]

Sorry I wasn't 100% clear on the above,

Some servers (OGN, Australia, Melb), my own server, and others I have seen only do 32fps max (give or take up to 10%)

So #MONITOR 1:

-Shows a peak speed of of 32 fps,

-Shows an idle/loby speed of 32 fps

-Shows 32fps most the time, when cpuload is only 64% or less

Is there a way to make it 50 fps , so it uses more CPU time

(above 90% would be a nice starting point)

Sorry I just thought everyone was aware of the 32 fps vs 50 fps issue,

Although it is good to hear your server is 50 fps,

Perhaps it to, is realated to HyperThreading ? (unlikely)

Just want "as smooth as possible" play on LANs.

Also: Is anyone run any Athlon XP/MP, AMD-64 or Opteron servers out there ? Wouldn't mind finding a way to benchmark them on an equal playing field, and compare the results.

[RN Malboeuf 12]

so you are saying different servers are only getting 37 fps? must be your config settings, post your whole config for us (minus pw ect)

as for 64 bit there is only 1 OS that supports those machines and thats the linux 64bit version and since ofp is 32 bit you wont get any help with it, im out of date on that but im pretty sure

[-IT-Q- 0]

hi m8

ofp multiplayer server is such a great mystery

check out our config, maybe u can see any difference

Fps (from #monitor)

----

Angus from RockRaiders told me to deactivate HT on the p4 machines. he cant tell why but it seems to slow down the server. but it seems obvious to me.

lots of germans have p4 3.2 ghz 2gb ram (well cant tell anything about the setting) root servers, but they do perform really bad with wargames. they've got huge problems with most maps with 20-24 players (fps below 10), but the americans with their 2-2,4 xeons dont.

well i do still want to do more research first because config and other things still may influence this and the ingame stats arent really that great. but so far thats our impression.

[darkpeace 0]

Many servers out there limit at 32 fps (no reason I can see)

(Not 37 fps, 32 fps and thats in the lobby)

Checked flashpoint.cfg and server.cfg, nothing stands out, even looked at the users cfg file too.

==Server.cfg=====================

// add a line password="{password}" here to assign a password for this session, which everyone who wants to join the game will need in order to connect

passwordAdmin = "****"; // Admin password

hostname = "GarageLAN - ACT";

// Server Restrictions

MaxPlayers=64;

// Note : Dont use " between wider quotes

voteThreshold=0.5001;

voteMissionPlayers=0;

motdInterval=30;

motd[]=

{

"messages here";

};

reportingIP="master.gamespy.com"

// reportingIP="207.38.8.34"

// End of Firehazzard II config

==Flashpoint.cfg=====================

Light_Explo="1";

Cloudlets="1";

Cockpit_Textures="256";

Limit_LOD="0.019000";

CPU_Benchmark="3200";

3D_Performance="1600";

Frame_Rate_Pref="750";

VehShadows="1";

File_Heap="16";

Object_Textures="256";

HW_Type="Direct3D";

MaxObjects="256";

Shadows="0";

Language="English";

Quality_Pref="500";

Resolution_W="1024";

Adapter="0";

Texture_Heap="16";

Resolution_Bpp="16";

Resolution_H="768";

Reflections="1";

MaxLights="32";

Sky="1";

Animated_Textures="256";

Light_Missile="1";

Landscape_Textures="256";

Textures_Drop_Down="4";

Shadows_LOD="0.050000";

LOD="7.500000";

Total_Memory="512";

Light_Static="1";

Product="Resistance";

MaxMsgSend=1024;

MaxSizeGuaranteed=1024;

MaxSizeNonguaranteed=512;

MinBandwidth=3000000;

MaxBandwidth=20000000;

MinErrorToSend=0.001; (changing this value, trying to find sweet spot)

==UserInfo.cfg=====================

frameRate=80.000000;

visualQuality=0.150000;

objectShadows=0;

vehicleShadows=1;

cloudlets=1;

viewDistance=1000.000000;

terrainGrid=25.000000;

volumeCD=7.000000;

volumeFX=10.000000;

volumeSpeech=8.000000;

singleVoice=1;

enableEAX=1;

enableHW=1;

frameRate=80.000000;

visualQuality=0.150000;

objectShadows=0;

vehicleShadows=1;

cloudlets=1;

viewDistance=1000.000000;

terrainGrid=25.000000;

volumeCD=7.000000;

volumeFX=10.000000;

volumeSpeech=8.000000;

singleVoice=1;

enableEAX=1;

enableHW=1;

frameRate=80.000000;

visualQuality=0.150000;

objectShadows=0;

vehicleShadows=1;

cloudlets=1;

viewDistance=1000.000000;

terrainGrid=25.000000;

volumeCD=7.000000;

volumeFX=10.000000;

volumeSpeech=8.000000;

singleVoice=1;

enableEAX=1;

enableHW=1;

=======================

I removed all the crap from the above (eg: MOTD, server names, used Server.cfg vs actual file name, etc)

Nothing up there stands out, as "limit this server to 32fps".

Apparently it is a very common problem, just wish there was a real fix for it (Timeslice / Quanta's RegHack anyone ?)

Maybe it is caused by Direct X 9.0b (was running a or just plain DX 9.0 before), or the 1.92 - 1.96 inclusive series of patches (It used to do 50 fps, then one day, no cfg edits performed either, it didn't anymore)

I suspect it is a MS thing,, Suma said in another post it was timeslice related, (remember back in Win 3.1 or OS/2 you could specify them), however no actual fix was mentioned.

That's what I am seeking (for the community)

Also working on a Radeon Z buffer flickering fix (RegHack) this weekend (Australia TimeZone), so I can play without distant objects looking like shite.

[darkpeace 0]

mmmmm, SuSe Linux is looking mighty good there.

(about to roll it out at work for Notes, Domino, Servers with the over 1024 thread patch, 5x the speed of a MS based Domino server, with 2000 users)

SuSu is funny, I think the FPS fluxuations you see are normal, however getting capped in the ass at 32 fps (give or take 5-15% depending on fluxuations) is pretty shit, as both a PEAK and IDLE (loby) server speed.

Why can't it just use 80% (or more) of a single CPU ? (HT disabled) and just run at whatever FPS it can manage with 80% - 99% CPU load ?

For a "game" server it is very weird.

And for what became a "high end" simulation, I would have expected MultiCPU/Threading in the server, esp for AI, and Scripting, as each AI group could run on a dedicated CPU if there where enough (Solaris Flashpoint box anyone ?)

[shrike 0]

Our current server is a dual Xeon 2.2 (note Xeon not Zeon   ) in this machine with 2.5 Gigs of RAM on a 100 Mbit connection, running on SuSE Linux Enterprise server.

In late 2002 when we started we were running it without HT enabled, performance was ok but less than we expected.

Later we installed a HT-enabled Kernel and while we noticed a performance increase on large coop maps with lots of AI, it seemed to have gotten slower in large adversial settings with 28+ players. It's still very well playable but lots of high ROF APCs or choppers can choke it pretty badly.

Maybe I should add that we run 2-3 OFP servers on that machine simultaneously, this is where a 2nd CPU comes in handy. You can easily play a 15+ player coop on our addon server with a full public game (25+) running on the other.

However it's still frustrating to realize what kind of machine one would need to play, let's say, a 15 vs 15 CTI. I don't think money can buy such a server atm.

We will switch to a Dual Opteron 250 system in the near future if everything goes well. I expect a noticeable performance increase due to the high memory bandwidth of the Opteron CPU's. Using 64 bit is not an issue (yet) but the Opterons should perform very well in 32 bit conditions, too.

It's gonna be a Dual machine again because a) it's cheaper in Germany to use 1 server only as you pay per height units in the rack + bandwidth and b) we hope BIS will provide  multiprocessing support in OFP 2.

To Darkpeace: Our server still tops at 50 FPS in the lobby but it's Linux. As far as I understood you run a windows server.

[darkpeace 0]

Yep, dodgy Windows, looking at changing that soon.

VM Ware, or not, I am downloading SuSe now.

(http://www.suse.com/us/business/products/server/)

Also looking at Dual Opteron Server if all goes well (read: OFP2 better be threaded, and well, or I doubt I'll touch it, might even as http://www.es.com what simulations they can offer civilians - lol)

Suprised Suma hasn't replied to this post yet to be frank.

Generally he has his say on such things fairly quickly.

Anyways, if you guys do have a Beta threaded server, fire me an email, or reply here, I keep tabs on this forum.

[RN Malboeuf 12]

MsgSend=1024;

MaxSizeGuaranteed=1024;

MaxSizeNonguaranteed=512;

MinBandwidth=3000000;

MaxBandwidth=20000000;

<span style='color:red'>is supposed to be in your server.cfg and not your Flashpoint.cfg</span>

this is ours, dont put your minbw so low, and remember making them to high will cause CPU load as it says in the manual

MaxMsgSend=1024;  // was 512

MinBandwidth=7680000; // was 5120000

MaxBandwidth=10240000;  // was 5120000

as for HT slowing machines down that is only with linux share loading when it moves the thread from cpu to cpu, when the server thread is on the HT part of the CPU it cant handle it, it will act like a lesser CPU big time, Windows wont have this problem, We cant wait to get way from linzux and back to win2k on our Zeon

[benu 1]

MsgSend=1024;

MaxSizeGuaranteed=1024;

MaxSizeNonguaranteed=512;

MinBandwidth=3000000;

MaxBandwidth=20000000;

<span style='color:red'>is supposed to be in your server.cfg and not your Flashpoint.cfg</span>

Actually, the DS-Admin.rft says it has to go to the flashpoint.cfg, although many admins seem to put it into the server.cfg. But i guess i trust the official docu more. Have you any official sources saying it has to go into the server.cfg?

[RN Malboeuf 12]

holy shit you're right it does say that lol, for almost 3 years every change i have made as been to the server.cfg and i have seen the changes, I've seen countless other server's config files and they too have been in the server config.

KLan-host server are like this, citadel, and many others

[Certa 0]

I never examined how OFP uses multiple threads but I doubt a dual processor server is going to make andy difference in performance. Yes it will make a difference if the server is used for other stuff than OFP too. Like if it's a combined OFP. Web or database server.

Anyways, I wrote some technical information about threading and multiple CPU's. I know a lot of people are interested in the subject. I haven't covered hyperthreading in this post, I haven't looked it up and tested it fully but I believe the theory about threading in computer games reamins the same.

Xeon got superior memory access.

When running dual P4 you will lose 10-20% of speed on the memory access and memory bus coordination. This is logical when two processors must share the same memory bus. Also, running dual processors makes memory write caching hard. Luckily there are some very cool formulas to ease these problems.

Judging from my own multi-threaded test program I ran on a quad Xeon it outperformed the single processor Pentium processors in memory access and speed (not just by 4 times, more like 5 times the speed of the single processor) . The power of Xeon is not just in it overdimensioned price but true power comes out when running two or more Xeons in one machine.

If you want cycles per dollar, go for AMD, Celeron or Pentium. If you want 2-4 processor running in parallell without any performance loss then Xeon is it. But Xeon is not for home gaming, I will describe why below.

The ideal number of threads on a computer is always equal to the number of processors. 1 thread = 1 processor. One problem is called task switching. It takes time for one processor to switch from one thread to another. Each thread is executed a few milliseconds at the time. A small percentage of the time is used for switching in between the threads.

The threads are lined up in a que, the thread of the top of the que is executed for a given time intevall and then moved back to the bottom of the que. A thread with a higher priority is moved up the que faster than the other threads for each task switch.

When running on a multiple processor system, the system tries to run the same thread on the same processor, if possible. This enhances cache hits and gives better performance in general.

Threads in a waiting state, like when waiting for a synchronization event to happen (could mean waiting for input or other threads), will simply directly give their time-slice to the next thread and move back to the bottom of the que.

Normally a thread executes for 5-20 milliseconds depending on configuration and system. NT Servers tend to use a higher intervall and workstation uses a lower for better UI response.

The problem with multi-threading and computer games

Multi-threading in client side of computer games are generally a bad idea.

Using multiple processors in your new monster gaming machine is equally a bad idea and most likely a waste of your money. Unless you like running WinZIP, WinRAR or any other process chewing software in the background while you are playing you are never going to get full use for your two (or more) processors.

Computer games wants to produce rendered images about 30-70 times per second. 50 frames per second means a time intervall of 20 milliseconds per frame. Now, multiple threads needs to be synchronized. One thread might work on logic and movements for the next frame to be rendered while one thread is rendering the current frame. These two threads must at some point be synchronized: The fist thread flags that it's done with logic, it's ok for thread 2 to start rendering. Synchronization in between two threads can take up to 20 milliseconds due to the latency of the thread que! This is way too slow and would create jerky animations on both dual and single processor machines.

This is the reason why very few games uses multi-threading. When buying or building yourself a gaming machine you are best off buying one fast processor put the cash is the graphics card, memory or even a RAID card for your harddrives.

However, on the server side, programming multi-threaded servers and using multi processor machines do make sense... Perhaps not for OFP though.

In a true client/server environment (not like OFP where the modell is more peer/peer and the server mostly serves as HUB) the server sends updates to the clients. The logic threads are running at the server and the client is more or less just doing the rendering, input and sending updates back to the server.

The server can not send the client all logic information needed for 50 frames per second animations. The networks are in general not fast enough for this speed (perhaps it would work at pure LAN play). The server will most likely send logic information 5-20 times per second (200-50 ms). The client will have to fill in the animation between with approximations based on last known object heading and speed and then synchronize with the latest information from the server when available.

For the player on the client the other network players normally seem to move smoothly, though they may jerk a bit when lagged when the client synchronizes with the true position of the network player.

The fact that the server doesn't have to generate 50 frames per seconds opens up for multi-threading. The threads do not have to deliver every 20 milliseconds and there is plenty of time for synchronization in between threads. A process intensive game like OFP could really benefit from this.

Lets hope the OFP2 server is a true server (not just a hub) and is multi-threaded.

[darkpeace 0]

Normally a thread executes for 5-20 milliseconds depending on configuration and system. NT Servers tend to use a higher intervall and workstation uses a lower for better UI response.

The server can not send the client all logic information needed for 50 frames per second animations. The networks are in general not fast enough for this speed (perhaps it would work at pure LAN play). The server will most likely send logic information 5-20 times per second (200-50 ms). The client will have to fill in the animation between with approximations based on last known object heading and speed and then synchronize with the latest information from the server when available.

The fact that the server doesn't have to generate 50 frames per seconds opens up for multi-threading. The threads do not have to deliver every 20 milliseconds and there is plenty of time for synchronization in between threads. A process intensive game like OFP could really benefit from this.

I don't know if it really matters if it is server.cfg or flashpoint.cfg, but yeah, as said above, the official docs say to use flashpoint.cfg for those particular commands, so thats what *I* used.

Perhaps adding the MaxPlayers command to each, and checking the results, would confirm this dilema. Otherwise I can see alot of Admins updating their "flashpoint.cfg" on their servers.

Quote[/b] ]

The server can not send the client all logic information needed for 50 frames per second animations. The networks are in general not fast enough for this speed (perhaps it would work at pure LAN play). The server will most likely send logic information 5-20 times per second (200-50 ms). The client will have to fill in the animation between with approximations based on last known object heading and speed and then synchronize with the latest information from the server when available.

Am pretty much talking about a LAN only server anyway.

Mainly with CTI, and other "huge" missions. Most online servers, unless local to an ISP, and only local customers of that ISP join, will be not be as smooth for other players to join (obvious for CTI to most, but not all)

(Most of us dedicated players have GigaEthernet switches ready for LAN backbones)

Quote[/b] ]

The fact that the server doesn't have to generate 50 frames per seconds opens up for multi-threading. The threads do not have to deliver every 20 milliseconds and there is plenty of time for synchronization in between threads. A process intensive game like OFP could really benefit from this.

So, if the server spends 20ms (1000ms\50fps) processing would that not add an "effective" (not actual network) 20ms to each players latency. (so with a server ping of <10ms, the server is spending more time processing the physics of multiplayer than it is shooting packets back and forth)

(Not to mention that the slower a given users frame-rate and connection, the less frequently that given player is sending updates, but lets all assume the clients are near identical machines on GigaEthernet, just for discussion purposes, we've all gone over this bit 1 million times)

Anyways, to prove the above point, use a crudgy server to host CTI maps at say 6-8 fps, all the players will notice.

(1000\8fps=125ms processing time)

Then use a faster one that sits at 32 fps max/peak/idle/loby, the players will notice it is smoother. (1000\32fps=31.25ms processing time)

Then use a server which does not have the 32 fps "bug", and runs at the 50 fps that Resistance servers "should" run at, the players will notice it is smoother again. (1000\50=20ms)

By graphing the above one can quite easily conclude, that if the server ran faster (eg: didn't cap itself at 32 or 50 fps), and thus used more CPU time then players are happier (in case that wasn't bleedingly obvious)

Once the server hits 125fps (8 ms processing time) no player should be able to notice, unless there are so many players that network/pci/southbridge latencies come into effect which is beyond the scope of this "simple" forum.

(Then again 64 players is the maximum, is it not ?)

I agree with Certa, but am sure we *all* want to see our Xeons / HyperThreaded P4's, and other decent servers showing what they can really do. By (as above multiple times) having the OFPR server benefit from HyperThreading, and Multiple CPUs to run even faster, or in the case of CTI just not as slowly - , say 50 fps constant would be nice to see, and surely possible with current hardware.

Still can't help thinking about the server.cfg vs flashpoint.cfg edits for those commands -   - least we all learned something today (Just wanted to add, thanks Benu for backing me up there)

If MaxPlayers works in both .cfg files, then it likely doesn't matter, still I followed offical documentation just in case.

[benu 1]

I don't know if it really matters if it is server.cfg or flashpoint.cfg, but yeah, as said above, the official docs say to use flashpoint.cfg for those particular commands, so thats what *I* used.

There are TWO config files for a purpose. For two different purposes actually. Try putting your motd into the flashpoint.cfg and remove it from the server.cfg and see if that works

Quote[/b] ]Perhaps adding the MaxPlayers command to each, and checking the results, would confirm this dilema. Otherwise I can see alot of Admins updating their "flashpoint.cfg" on their servers.

Adding anything to BOTH configs proves nothing, except if one of the configs can discard input it can make no sense of. Only add it to one config to test for sure.

Quote[/b] ]Still can't help thinking about the server.cfg vs flashpoint.cfg edits for those commands - - least we all learned something today (Just wanted to add, thanks Benu for backing me up there)

I put it into the server.cfg for quite some time too. First i made my flashpoint.cfg from the ds-admin.rtf, then month later there were discussions in this forum where it was in the server.cfg and i couldn't remember why i put it into the flashpoint.cfg, so i put it into the server.cfg (so many admins can't be wrong ). After checking into the ds-admin.rtf again i corrected that.

Quote[/b] ]If MaxPlayers works in both .cfg files, then it likely doesn't matter, still I followed offical documentation just in case.

As i said, try putting maxplayers into the flashpoint.cfg ONLY to try it. And even if one or more commands would work in both configs thats no guarantee that all commands would work in both configs. Try starting the server without a flashpoint.cfg at all or without the server.cfg to see what i mean.

[RN Malboeuf 12]

i'll do some testing

all in all we have to wait to see what the OFP2 server will be, i cant see why BIZ wont tell us

but we know how behind they are with ofp2 and im just going to ahead and switch to a P4

[darkpeace 0]

Was hoping for a "yep, the server is multi-threaded now" post, but to no avail

Damn, might need to wait until OFP:2

[benu 1]

Are there any multithreaded gameservers for other games at all?

[darkpeace 0]

The World War II online server might be threaded. (?)

The stuff at http://www.es.com is most likely threaded very much aswell. (and on the client side too I suspect)

Are there many games server sthat could benefit from HyperTheading is a better question.

eg: Flashpoint with over 2048 AI, and 64 human players, with heaps of vehicles, etc.......well that would benefit....alot.

(or get 1-5 FPS on the server "world/physics" engine)

The point is, once you you 100% of current CPUs (for any task), you really need 2 or 4 CPUs to go any faster, and that requires much more software engineering to be aware of the multiple cpus, etc

Eg: Video Encoding, Large Servers for various uses, File Compression, and Flashpoint Server would all benefit.

(Most of the above are already threaded btw)

[suma 8]

holy shit you're right it does say that lol, for almost 3 years every change i have made as been to the server.cfg and i have seen the changes,

Then you must have seen the famous placebo effect. Those values are read from Flashpoint.cfg (and only from there).