Howa many players at once?

[ding333]

I know the beta is going to be a 4 player ai battle but how many will be able to be on a map. The standard rts 8 or more? Also are bots going to be usable like in defcon? Also are teams going to be allowed like a 2v2 game?

Edit: Sorry. Any non sarcastic comments are welcome.

[NeoThermic]

Information is currently thin on the ground, and the beta having an NDA will mean that it'll stay that way unless you make the cut. Ergo nothing has been let on as to how many players can partake in a game. If Multiwinia uses the same networking structs as Defcon, then I'd assume max players would only be limited by the map (and possibly colours of DGs). 6 players would be feasible. It can also be assumed that the AI for Multiwinia will be like a game of Defcon; a challenge at first but fodder when you're used to the game. Finally for teams, technically possible, information is lacking

In short, you'll have to wait and see.

NeoThermic

[martin]

I would imagine that AI for MW is easier to make than AI in Defcon simply because MW is more like other strategy games - so maybe it won't be such fodder?

[jaysc]

i guess you would be able to change the colours of the darwinia customly, but if they might have pre colours - blue, yellow, white etc.

The max i guess is 8 or 6.

[Xocrates]

Personally, I doubt there will be more than 4.

Appart from what jelco already pointed out, all development screens seen never showed more than for. Additionally, I don't think that 6-8 player each controlling blowing up hundreds of darwinians would allow for much performance even in high end computers. Think Bioshpere times 4 that's probably what an 8 player game would look like.

It's entirely possible that there are more than 4 though.

[NeoThermic]

Think Bioshpere times 4 that's probably what an 8 player game would look like.

It's entirely possible that there are more than 4 though.

Wha? Biosphere had about 2k darwinians on it. I'd wager an 8 player game would have about 4k, assuming meger resources to make you fight for more control for the larger army. Least not forgetting that you can just as easily ignore the AI states of non-visible DGs (as long as the AI is stage based and one can jump into it at request) as well as not drawing them, so Multiwinia shouldn't have to draw and AI much.

Really, Biosphere was a toddle in Darwinia anyway. Receiver was always the problem. Plus my testing of spawning DGs in Darwinia showed that 10k DGs can be on screen without too much performance loss. This was with an ATi 9600 Pro, so newer cards should do better:
Click for bigger

NeoThermic

[Feud]

I suppose it's hard to tell at this point, but while Darwinia can handle 10k without too much problem, what kind of issues would those numbers raise when trying to stay synced over the internet?

[Montyphy]

My aunt Betty told me that Multiwinia will support 42 unique players at once.

[Feud]

I guess the max is 4. The fact that we only see 4 colours in this pic:

Five actually, the Darwinian on the left is burgundy while the one on the right is candy apple red.

[ewanm]

I guess the max is 4. The fact that we only see 4 colours in this pic:

Five actually, the Darwinian on the left is burgundy while the one on the right is candy apple red.

How did you notice the difference, I would need my Gran (who like cross stitch) to tell me they we different

[briceman2]

...my testing of spawning DGs in Darwinia showed that 10k DGs can be on screen without too much performance loss. This was with an ATi 9600 Pro, so newer cards should do better...

The performance problem is *not* rendering that many DGs -- rendering what are essentially sprites is probably the least of the burdens caused by DGs.

My rough testing has shown me that it is the DG AI that brings systems to their knees. When you have just one team on the map, you can run thousands, as you found. But as soon as they have enemies, or building ports to run to, or officers to follow, or turrets to run away from, then the ceiling becomes much much lower. The game seems mostly CPU bound (AI), not GPU bound (rendering).

Plus there are weird cases where performance suddenly goes to single digit fps for no apparent reason. If you have several hundred DGs in a group and an enemy unit flies overhead (invaders, jellyfish, triffid eggs, flying ants, etc.) then the DGs will try to fire their lasers into the air. For some reason this is EXTREMELY resource intensive. As soon as the DGs stop firing up, and return to firing along the ground, fps returns to normal. I've seen drops of 50fps to 5fps with only a few hundred DGs.

Another performance killer is when lots of DGs are on very steep terrain, like the walls of tall flatten areas. The entire vertical is usually just a single landscape grid cell, stretched, so all the DGs on the wall are essentially on top of each other (darwinia is really a 2D game internally) and DG density skyrockets. If you run the linux version, you can watch server messages which say: "EntityGrid max neighbours set to 100 (time taken 0.00ms)". Once that neighbor number goes up to near 2000 and over, performance takes a steep nose dive on my system.

You can also cause this problem by having an officer gather thousands of DGs into a tight group, or send them all to a single radar dish door. Suddenly the engine can no longer handle all the DGs that were running fine before you gathered them. Unfortunately the EntityGrid never gets smaller, so once you've hit the ceiling you're stuck with sucky performance until you leave the level. Disbanding the DG clusters, leaving and re-enering the level can often gain me +20fps since the EntityGrid will restart at a much lower density / size.

[NeoThermic]

...my testing of spawning DGs in Darwinia showed that 10k DGs can be on screen without too much performance loss. This was with an ATi 9600 Pro, so newer cards should do better...

The performance problem is *not* rendering that many DGs -- rendering what are essentially sprites is probably the least of the burdens caused by DGs.

Clearly my first few lines were outright ignored:

Least not forgetting that you can just as easily ignore the AI states of non-visible DGs (as long as the AI is stage based and one can jump into it at request) as well as not drawing them, so Multiwinia shouldn't have to draw and AI much.

I.E. if there's DGs on the map that are not your DGs, not engaged with yours and not visible on your screen, they can be ignored.

When you have just one team on the map, you can run thousands, as you found. But as soon as they have enemies, or building ports to run to, or officers to follow, or turrets to run away from, then the ceiling becomes much much lower. The game seems mostly CPU bound (AI), not GPU bound (rendering).

That is strange. My rendering test was an officer in the middle of a large map calling DGs over to itself as they spawned. I never experienced such issues. I also tried a test where 10k red DGs were pitted against two converging teams of yellow and green DGs of 5k each. Moving them was not a problem, the main performance issue came from the second after a grenade impact; my theory being the game was sorting through the list of DGs and removing those who were dead, while adding to the sprite list the DG pieces to animate for their deaths.

Further still, in what looks like a huge lightsaber battle, two teams (red vs green) pitted it out on the test terrain. When the greens started to lose, I gathered them up for a proper attack. These caused no fps issues past drawing the souls, reds and greens.

Plus there are weird cases where performance suddenly goes to single digit fps for no apparent reason. If you have several hundred DGs in a group and an enemy unit flies overhead (invaders, jellyfish, triffid eggs, flying ants, etc.) then the DGs will try to fire their lasers into the air. For some reason this is EXTREMELY resource intensive. As soon as the DGs stop firing up, and return to firing along the ground, fps returns to normal. I've seen drops of 50fps to 5fps with only a few hundred DGs.

Again, this is strange. I experienced reverse issues. taking 50k DGs on an island, I then spawned a squad and fired a grenade into the centre of the bunch. The FPS improved, since the whole group had found something that their AI was set to avoid. (this being the before http://www.neothermic.com/darwinia/stre ... nshot2.png ) As soon as the nade went off, however, the performance dropped further than having 50k DGs idle. As above, removing the dead and replacing them with their bits consumed more time.

You can also cause this problem by having an officer gather thousands of DGs into a tight group, or send them all to a single radar dish door. Suddenly the engine can no longer handle all the DGs that were running fine before you gathered them. Unfortunately the EntityGrid never gets smaller, so once you've hit the ceiling you're stuck with sucky performance until you leave the level. Disbanding the DG clusters, leaving and re-enering the level can often gain me +20fps since the EntityGrid will restart at a much lower density / size.

Again, I never had such issues. I'd assume, then, that these are bugs native to the Linux version?

Don't forget though, that rendering issues can cause more problems than AI issues. Defcon is a perfect example of that; the rendering order is suspect at best and causes a performance drain. Even though they are sprites.

NeoThermic

[martin]

Out of interest is there any particular techincal reason that the graph largely levels out at 10FPS? some kind of lower FPS limit beyond which darwinia starts making quality sacrifices to keep the FPS up maybe?

[NeoThermic]

Out of interest is there any particular techincal reason that the graph largely levels out at 10FPS? some kind of lower FPS limit beyond which darwinia starts making quality sacrifices to keep the FPS up maybe?

The graph doesn't level out at 10fps If I were to plot further, it'd eventually hit 0. What you're seeing is exponential decay.

NeoThermic

[martin]

Out of interest is there any particular techincal reason that the graph largely levels out at 10FPS? some kind of lower FPS limit beyond which darwinia starts making quality sacrifices to keep the FPS up maybe?

The graph doesn't level out at 10fps If I were to plot further, it'd eventually hit 0. What you're seeing is exponential decay.

NeoThermic

I know what exponential decay is ^^

However, what I meant to say is that why does the rate of decay suddenly slow down at 10ish, because it certainly doesn't look like it's decaying at the same rate to me