Testing netcode settings

[Adammo 0]

Can some of you verify my findings:

On the dedicated servers there is a file called

flashpoint.cfg

This files hold some of the customizable netcode settings for the server.

By default these settings are not in the file. Please run this test and see if you are getting better performances with the dedicated server.. I know that I saw an huge decrease in lag in version 1.46 thru 1.49.

Place the following lines in your flashpoint.cfg at the top:

MaxMsgSend = 64;

MaxSizeGuaranteed = 512;

MaxSizeNonguaranteed = 64;

Then save the file and start your server.

Remeber you created the backup of your flashpoint.cfg file.. or just remove these lines to get it back to default.

Please let me know if this improves your game play and reduces your lag.

Thanks,

Adammo

[rekrul 7]

This is dependant on your Internetconnection tho, so I don't see why you wanna know.

-Rekrul

[Adammo 0]

These settings are not dependant on the internet connetion "type". They are settings that relate to DirectX and how the streaming of packets are handled. The setting that is dependant on the bandwidth I omitted from this thread. The bandwidth setting is a setting based on the "type" of connection.

The reason that I need to know is that the testing I do in this game may directly effect what you get in the next patch and I need to know if these settings will help the games latency or not for consideration in future patches.

-Adammo

[BigQEd 0]

Adammo who are you? Do you work directly with BIS?

Why isn't SUMA or other Moderators asking this question if it as you say:

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote (Adammo @ Mar. 22 2002,08:29)</td></tr><tr><td id="QUOTE">The reason that I need to know is that the testing I do in this game may directly effect what you get in the next patch and I need to know if these settings will help the games latency or not for consideration in future patches<span id='postcolor'>

I mean if it will effect future releases i would think they would be asking these questions.

[Adammo 0]

I do testing for Bi studios on the "Internet" side of MP testing, that is why I am asking these questions.

Suma is at the EU gaming convention right now and that is why he is not posting at the moment.

Now, can anyone please let me know if the settings reduce lag in 1.49? If players are not seeing any decrease in lag then I will retest the netcode setting and try to get better numbers to retest with. Some changes int he netcode settings where made in the 4.x patch that helped the MP game for dedicated servers. I believe that these setting might need to be re-evaluated for the changes made in the 1.47-1.49 patch. I could be wrong but I have seen a great deal of decrease in lag on the 1.49 patch using the above mentioned settings.

Adam Moore

(check game credits)

[BigQEd 0]

Hmmm... sorry. I figured you would have a MOD or Admin status on the board.

As far as 1.49. It is better than 1.45-1.48. However we still see a lot of lag. Especially on larger Missions.

I HOPE something can be done to reduce the lag... it is not fun to get 10000+ Desync on servers that are more than capable with more than enough bandwidth (Athlon XP 2000 1GB Ram SCSI 15000 RPM HD, and Multiple OC3 Bandwidth).

[Placebo 28]

Would be nice if there were some settings we could play around with to reduce the lag for those of us who host games not on a deddy, the netcode of the latest patches seems to have taken a lil step back in time compared to early 1.4x.....

[Espectro (DayZ) 0]

Doesnt any1 have a dedicated server they can test it on?

[gonk 0]

Are they (BIS) playing with the netcode for non-dedcated servers ? ie normal MP games or is it too in-grained in to the game engine ? if you need a good example of a game that handles lag very well...check out IL2.

[Adammo 0]

Things learned from the dedicated server can be used in the future for non-dedicated clients.

-Adammo

[gonk 0]

ok...thanks for the reply

[MitchSchaft 0]

Am I wasting my time by putting these settings in my flashpoint.cfg?

MaxMsgSend=256;

MinBandwidth=300000;

That's all I have concerning my dedicated. I have a 3Mb download and 360Kb upload. Cable.

[suma 8]

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote (Adammo @ Mar. 24 2002,07:56)</td></tr><tr><td id="QUOTE">I do testing for Bi studios on the "Internet" side of MP testing, that is why I am asking these questions.<span id='postcolor'>

I can confirm Addamo is really heliping us beta-testing netcode. I received his settings via e-mail and I just wanted to post the same question here - seeing he posted it first, I can just backup it - if we have enough feedback those setting may make things better, we may make them default.

The same would be true for any setting that are discovered by anyone, not necessarily "official beta tester". If someone will provide us idea that is simple to implement and significanlty improves or fixes gameplay, we will definitely consider it implementing (see bugfix list in readme.txt - there is Cessna fix in 1.43 implemented by config modification as suggested by WKK Gimbal).

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote </td></tr><tr><td id="QUOTE">These settings are not dependant on the internet connetion "type". They are settings that relate to DirectX and how the streaming of packets are handled. <span id='postcolor'>

Can you provide me some more insight on how is this related to DirectX packed streaming? I understand MaxMsgSend may be related to size of some DX internal buffer - although I did not see any such limitation in DirectPlay docs (not that it means much, as DP docs are often very incomplete, especially when talking about how things are done). I do not understand how is MaxSizeNonguaranteed=64 improving connection. I think in most situations it will prevent message merging, which I would expect to make performance worse, not better.

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote </td></tr><tr><td id="QUOTE">Now, can anyone please let me know if the settings reduce lag in 1.49? <span id='postcolor'>

I would like to see some answer too. Until now it seems it generated only more questions. Any feedback based on real server testing?

[MitchSchaft 0]

I went ahead and put those settings in my flashpoint.cfg.

[Lt_Damage 0]

Assuming these settings are in 1.49, the following results were achieved on a dedicated 1.49 server with 1.46 clients:

*Vehicle lag is significantly reduced over 1.46 netcode, delay upon entering vehicle is usually less than a couple seconds, where as 1.46 it could be up to a minute.

*Driving vehicles is also much less laggy.

*Kill lag is also improved.

*Server timeout's have virtually been eliminated.

*"No message received" message rarely appears, even when server connection under strain. (This might be bad or good..)

Only issue I have noticed is crash to desktop, which I am working on troubleshooting. As soon as I have more information I will cover this area more in depth.

Edit - Typos

Regards,

[suma 8]

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote </td></tr><tr><td id="QUOTE">Assuming these settings are in 1.49, the following results were achieved on a dedicated 1.49 server with 1.46 clients:<span id='postcolor'>

Those setting are not in 1.49. Addamo asked to check if something will improve when you will use them for 1.49 by inserting them into your Flashpoint.cfg.

It looks we will have to wait for some answers a little longer.

[Lt_Damage 0]

Sorry, bit unclear there. Assuming a person uses these settings in 1.49, as we had done, the results are as follows.....:

Basically I mean the same settings in versions 1.48 and lower do not seem to produce the same results.

1.49 + these settings turned out to be a good combination for us.

Hope that clears it up.. no doubt you will need more data to establish whether these settings improve performance accross the board.

Regards,

[moerty 0]

bump ^_^

[Espectro (DayZ) 0]

m000

[Lt_Dan_Sweden 0]

Hi

I have used

MaxMsgSend = 64;

MaxSizeGuaranteed = 512;

MaxSizeNonguaranteed = 64;

since your post. I usually have a higer MaxMsgSend around 128-256.

The problem is that its very hard to eveluate the settings. Maybe thats why you are not getting any posts back.

Please suggest a uniform evaluation method.

My observation/speculations:

I think the MaxMsgSend is strongly corelated to the CPU speed. In my case i think i get slightly better performance increasing the MaxMsgSend to 180 or so. The MaxSizaXXX

I have no idea on the impact.

If we play a big map, with choppers, 12-14 player the flops go down to 8-10 FPS, which is a shame. but the players keep coming back since my server has really good PING and interesting maps and ease of becoming ADMIN

My Dedicated OFP PC is a P3 850 Mhz

256 RAM

2 Mbit upp and 2 MBit Down

OFP Dedicated V1.49

Try it at IP 62.119.194.214 during evenings at CET

[BigQEd 0]

Computer SPECS: Athlon XP 2000 with 1 GB Ram and 15000 RPM SCSI drives.

Bandwidth SPECS: Multiple OC3 connection. Fast Ethernet on computer. (thus a 100Mbps true Internet Connection)

With MaxMsgSend I have tried:

MaxMsgSend=64

MaxMsgSend=128

MaxMsgSend=192

MaxMsgSend=256

MaxMsgSend=512

MaxMsgSend=768

MaxMsgSend=1024

With MinBandwidth I have tried:

MinBandwidth=768000

MinBandwidth=1024000

MinBandwidth=2048000

When the MinBandwidth is set to high we seem to lag out more and the bandiwdth usage hovers around 750-2000 no matter what setting we apply.

When the MaxMsgSend is set above 256 the server runs awesome until we get over approx 20 people connected... at which time we see massive desync and lag... and even crashes in extreme situations. If this setting is to low you see more delays when playing such as entring a vehicle... thus you have a constant lag... lower settings do seem to compensate if you wish to host a bunch of players but the overall performance is poor.

Conclusion:

With all these settings even though we have way above the norm in Bandwidth and above spec on cpu. It seems the best settings are STILL MaxMsgSend=256 and MinBandwidth=768000. Overall it just seems to perform the best with these settings... no matter what Bandwidth someone has (because OFP will not utilize more than a certain amount of bandwidth no matter what setting you have... we KNOW because we have a network that can handle more than OFP could ever push). So I truly believe it is a Netcode limitation... in that once you reach a critial point in CPU usage your done and no tweaking these settings is going to get you beyond it. DirectPlay has to go...

[suchey 0]

</span><table border="0" align="center" width="95%" cellpadding="3" cellspacing="1"><tr><td>Quote (BigQEd @ April 08 2002,21:29)</td></tr><tr><td id="QUOTE">It seems the best settings are STILL MaxMsgSend=256 and MinBandwidth=768000. Overall it just seems to perform the best with these settings...<span id='postcolor'>

We have found this to be true in our servers case as well...

[Angry Radish 0]

Agreed.

We're running with

MaxMsgSend = 256;

MinBandwidth = 768000;

On a 2.2ghz P4 with 1GB of ram on a full T-1 serving ONLY OFP, no other apps/clients/servers, and it seems these settings are pretty much the sweet spot for performance.

[Lt_Dan_Sweden 0]

I have changed to your settings MaxMessege=256

and Min Band 768000, but my opinion still stand:

It is difficult to tell the effect. Please suggest a

standard way of evaluating the result of the changes?

4 vs 4 players all with ping between 50-100 ms

A standard CTF

measure FPS?

Suggestion?

[karhu 1]

What are the best settings for 100 Mbs lan game with 4 computers and server?