Spring RTS Engine

Lindir The Green wrote:Having Spring

1) Store lots of data for each game
2) Send the data to a central server
3) Combine all the data with a complex formula into a single benchmark number for each user

is more complex than having Spring

1) Run a small benchmark application every once in a while

Spring itself only does 1). At this time I would like to draw your attention to the fact that Spring already stores lots of data for each game. Check out the end-of-game statistics at some point.

The lobby client would do 2) and 3). Or the data could even be kept on the host machine for added simplicity. (The more I think about that alternative the more I like it.)

Heck, 3) is just for convenience. You could just have the lobby display average CPU load where it current displays CPU speed. That's about as simple of a change as you can get.

Wouldn't it be easiest just to create some kind of test script (like the ones that's displayed when you just start spring.exe, commanders, air test and whatever they are called), make it run for 1minute or something, collect all data and base the benchmark value from that?

The downside would be that you'd have to run that benchmark every time you change your settings then... or perhaps it could save different profiles... so each settings setup would have its own benchmark values... (i.e. I switch a lot between 1024x768 windowed and 1280x1024 full screen depending on if I'm specing or not)

This sounds like an awful lot of work just so some people can show off their compu-p33ns. A C2D shows as 2.3 or something, but so what?, any CPUs above 2 or slightsly less is fine anyway for spring and would never get kicked by a sane host. The current system works fine if you want to be sure never to play with people who have 10year old comps or whatever.

no.

no.

Hey guys this is Drolito just chiming in to say you can lock this thread now


Thanks in advance

Zydox, that would force people to run some lengthy benchmark.

Recording FPS of existing games costs nothing and is completely transparent to the user..

Ishach wrote:Hey guys, just popping in to say hi


Its me, Video gaming personality Chris Taylor


Thats me, Chris taylor

Ishach is not me ... its Chris Taylor :D

I am so happy to have an admiror who try to copy me ^^

Licho wrote:Zydox, that would force people to run some lengthy benchmark.

Recording FPS of existing games costs nothing and is completely transparent to the user..

That's true, though the FPS varys alot between different battles... atleast for me... depending on how the battle is played... doesn't different units consume different amounts of resources... ?

Or does 300 radar planes consume the same resources as 300 fleas?

And how are the results effected if I get killed 5min in, tab out, and return after 30min, so that most of the battle is executed in minimized mode, without rendering... ?

Zydox wrote:

Licho wrote:Zydox, that would force people to run some lengthy benchmark.

Recording FPS of existing games costs nothing and is completely transparent to the user..

That's true, though the FPS varys alot between different battles... atleast for me... depending on how the battle is played... doesn't different units consume different amounts of resources... ?

Or does 300 radar planes consume the same resources as 300 fleas?

And how are the results effected if I get killed 5min in, tab out, and return after 30min, so that most of the battle is executed in minimized mode, without rendering... ?

That's why you use CPU load. Which is just as low-cost and transparent to record.

edit. So my reworked proposal would be for Spring to record the average CPU load for each battle into some history file on the host's hard drive, use, for the sake of argument, a 20-battle trailing average* of CPU loads, and display that number along with # of cores (if the lobby can get that info?) in the lobby, in place of the CPU speed in GHz.

This would be a much more accurate indicator of ability to run Spring.

*Average of the CPU loads for the last 20 battles.

I still don't like the id├â┬®a of using dynamic CPU captures... if I'm a newb and have only been playing 1v1 battles against some AI, I should get rather low CPU values... though I'm rather sure that would change if the player would join a 5v5 or 8v8 battle...

I think we need a fixed measurement in order to get some usefull information out of it...

Also... doesn't it require different amounts of CPU power to host / play / spec a battle?

In progress: Spring mark

system info program for Spring.
it will upload this info to an public online database
(the upload function is still in development)
AFlobby will use Spring Mark in the near future. i cannot guarantee that TASclient will use it too.

soon people can view players system info online.
people can host big games for fast computers owners only without lag, while the slower ones can have some fun playing smaller games, without getting kicked because they are lagging a big game.

***note: This program respects your privacy, no personal information will be send to the internet, only a part of the info that shows up in springmark, and only when you press the button :D ***

Using an exe for system benchmarking for a game thats multi-os nicely though out.
edit:- also getting all clients to use uft character encoding should be a better thing to get done. Not some benchmarking tool, that would be better handlied via unitsync or spring itself.

Using an exe for system benchmarking for a game thats multi-os nicely though out.

Since every os handles the system info in an unique way, its better to make different system info programs for different os's.
just like spring has one windows version and one linux version.

also getting all clients to use uft character encoding should be a better thing to get done.

? you mean UTF?

Not some benchmarking tool, that would be better handlied via unitsync or spring itself.

go ahead! till that time, SpringMark will do the job.
btw, SpringMark is not benchmarking, only system information

So basically, sometime in the near future, you're gonna have to download a standalone benchmarking program... in order to play massively oversized games on shitty metal-free maps?

Uh, thanks, but I think I'll pass.

Let the low spec players on your team and then .kick .take when they stop responding for 2 seconds

So basically, sometime in the near future, you're gonna have to download a standalone benchmarking program... in order to play massively oversized games on shitty metal-free maps?

for a lot of people a spring game is never over-sized.
btw, spring even lags with 500 units on big maps for some players
spare me your feelings about greenfields.
its a very popular map, with its own strategy rules, and not without reason

That program doesnt benchmark anything... only copies your CPU info -.- Doesnt tell me anything, i thought you were making program which generates only one number which tells how good your computer is for spring...

Zydox wrote:

Licho wrote:Zydox, that would force people to run some lengthy benchmark.

Recording FPS of existing games costs nothing and is completely transparent to the user..

That's true, though the FPS varys alot between different battles... atleast for me... depending on how the battle is played... doesn't different units consume different amounts of resources... ?

Or does 300 radar planes consume the same resources as 300 fleas?

And how are the results effected if I get killed 5min in, tab out, and return after 30min, so that most of the battle is executed in minimized mode, without rendering... ?

Yeah it differes, you would need tens of games averaged to produce reliable results...

I'm not sure how spring measures CPU load, but I suspect it's some reverse value of FPS anyway Any normal game is using CPU to 100% - to display as many FPS as possible..

It would also be unreliable as it'd be highly dependant on the mod, map, and programs in the background. Even the players tactics could affect the results.