Benchmark and autoconfiguration

[darkpeace 0]

What Flashpoint 1.96 (and later) really needed was a benchmark / autoconfiguration tool.

For example, one that runs through a 30min (or less) benchmark going through texture sizes, for smoke, vehicles, cockpits, terrain, weapons, explosions, etc

The ATi FireGLs can only do 8 (full) hardware lights, which is pretty good, although the setup allows up to 32 lights (good, but dynamic lighting didn't scale well in hardware)

Smoke, Explosions, etc, well smoke need only be a 16x16 alpha effect (looks the same when layered), explosions could be changed, or done at 128x128-256x256, etc

Well, run through the combinations of various settings, and advise the user what to expect (min/max/avg/delta min), if they enable say, feature X (2048x2048 textures) or say feature Y (32 dynamic lights vs 4 or 6)

Give it a cool name like a FlashpointMark 2005 etc, and submit the program/demo/benchmarker to a few PC mags (atomic, pc user, APC, etc), to a few websites, and the like.

Do a database of what video cards get what score, with information on the particulars (like how slow 32 dynamic lights are, or layered alpha effects that actually are computed as (layer1+2+3+4+5+6+nth alpha effect DIV nth = pixel colour) which hurts performance, since they are all the same colour 'smoke cloud' it could be done at (total smoke layers+final non smoke layer DIV grand total alpha layers).

eg:

smoke, background = 50%

(smoke x 2 layers) background = 2/3rd alpha - 1/3rd backgnd

(smoke x 24 layers) background = 24/25th alpha - 1/25th

Cuz damnnnnn, alot of smoke alpha textures can drop me to 6-12 fps on a RADEON 9800 XT card with P4C-3000+HT

Maybe have a setting that limits the total smoke, alpha effects to X many smoke puffs, and include results in benchmark (in increaments of 4 or so)

It would be quite simple to code,

1 -Benchmark available settings in CFG file (with min/max framerate available to tweak, as it is in user.cfg)

2 -Show the results, and let the player decide what features are worth having on, warn them if it will give under 30fps or under 15fps, as some idiots like to break their PCs)

3 - Write the CFG file, let them post the FlashpointMark 2005 score somewhere.

In all my years of computing, have noticed that benchmarks (however meaningless) are always popular, if a good one cropped up (I eg: read above) and was distributed (as above, and to www.tomshardware.com, etc) I am sure Tom (and every other PC Mag, reviewer, out there) would start running it.

And it is a damn cost effective way to use word of mouth on a global scale and get publicty for the game.

(if it was possible to do this for FP1985 I would seriously try coding one up, but it doesn't support logging the frame rate)

Also gives players a sneak peak at what to expect the game to run like on say GeForce MX420/440 and the like, what detail they can 'pull off' at 30fps +)

I am sure others will post their thoughts and ways to improve or even write the program.

I am off to play at 60fps now, after much MANUAL tweaking of files that the autoconfig does not even touch

Still, Flashpoint is to tactial war sims as Half-Life was to first person shooters - Good Work guys

[darkpeace 0]

lol....doh

And after all that I forgot to mention the same thing for the network setup.

Eg: if players need / want to host MFCTI / or normal maps, etc, they need between 64kbps and 384kbps+ PER SLOT in the UPSTREAM direction.

I am sure that many FOOLS on ADSL seriously think that a 1536kbps connection is 1536kbps BOTH WAYS, even though down here in Australia the ADSL consumer speeds are (and will be until 2007) 1536/256, 512/128 and 256/64.

Now 2 players in MFCTI is possible to host on 1536/256 if everyone is on the same ISP, the other 12 players can be on a LAN.

Also seperate LAN and Internet traffic shaping in the server, if you need to cap it at 220kbps you don't want the 12 LAN players been capped to 220kbps/12+ as then their is ARTIFICIAL network congestion to the Internet players....

(This may have already been fixed, just exclude 192.168.0.x and the other LAN classes from bandwidth shaping)

Do a similar benchmark Flashpoint.net 2005 (just a lame name)

-Show how many players they can host

-Do not let them go under 48kbps PER SLOT

Or.....

Would it be possible to 'shoutcast' the same data to all players, with a tag to those players that are far away from a battle that it doesn't matter if they drop 'that' packet, where as for a battle somewhere else it would be the vice-versa situation (assuming a psuedo shoutcast technology, that can selective 'lose' packets to players in different areas)

This way a 1536/256kbps connection could host (in theory) unlimited players, as they would all get (a psuedo selective) shoutcast stream from the server at 256kbps (which is a shitload off bandwidth for gaming, on a per player basis)

Damn, should've posted this under the netcode