Low cell map

[dmitri 0]

I'm working on a 6 cell map. Due to the low cell count, it's only playable at a very low viewdistance of 500. Above that performance degrades massively. At 700, ArmA crashes immediately. I realise performance will be low on such a map (though for my own goals, the performance hit is justified, and the low viewdistance does not effect gameplay..). I've given up trying to find a solution for the crash, I realise it's the terrains increased poly count. Just wondering if anyone has experience with low cell maps?

I've seen others here saying they're working on 7 cell. Curious about how the engine is handling your maps, whether it'd be worth it for me to move from 6 to 7. I have already recreated the portion of the map that is complete (roughly 700 square meters) in 6 cell, it was originally at 5.

Another problem, mentioned in a previous thread with no solution found, is that AI drivers are crippled at the lower cell count. Even after all this, it's really hard giving up on the dips, gulley's and other small terrain features that lower cell counts allow. I love it.

Seam12

[Nicholas Bell 0]

Can't comment on use of small cells myself, but your island certainly looks good!

[fitzee 7]

Hi.

The map I made to learn Visitor 3, is 2560x2560 meters with a 5 meter cell size. Its a desert stlyle map with lots of mesas, valleys and rolling bumps. Heights ranging from -1 meter to 230 meters. Im not experincing the problems you mention. I get fps ranging from 30-40 in the extremely cluttered town up to the 70's out of town.

Ive had battles in and out of the town with close to 200 units, without problems.

View distance 1500-2000.

I know this info dosent help you but just letting you know 5 meter cells do work.

My system is rather weak by ArmA standards.

Athlon64 3200+, 2 gigs ram, 7800GS AGP video.

[bravo 6 0]

the most important is that the map is fluid and produces no lag. Else it will be worthless.

Is it a map or an island?

[fasad 1]

Ahh, so you mean small terrain grids/cells? As in 6 meter cells, rather than 6 cells? "Low cell map" implies the total number of cells is very low.

Terrain/texture grid sizes and impact on performance have been briefly discussed in this thread. I found that using a terrain grid with cells sizes of less than 8 meters reduces performance significantly, with little/no change in performance for all cell sizes above that. The testing only took about 10 minutes, so you may want to do some investigations of your own too.

[dmitri 0]

Ahh, so you mean small terrain grids/cells? As in 6 meter cells, rather than 6 cells? "Low cell map" implies the total number of cells is very low.

Yeah, sorry, I need to amend my language. I meant "low" as in 6 cells as opposed to 10. Of course that would actually be a high cell count, as you've pointed out.

Hi.

The map I made to learn Visitor 3, is 2560x2560 meters with a 5 meter cell size. Its a desert stlyle map with lots of mesas, valleys and rolling bumps. Heights ranging from -1 meter to 230 meters. Im not experincing the problems you mention.

View distance 1500-2000.

I know this info dosent help you but just letting you know 5 meter cells do work.

My system is rather weak by ArmA standards.

Athlon64 3200+, 2 gigs ram, 7800GS AGP video.

Hmm. That kinda goes against everything I've read\experienced.

Does your island have a lot of objects? Now that I think about it though..when I initially created the terrain for my island without any sat map, I could achieve a high viewdistance (2000), though performance was obviously quite poor. The minute I included a sat map\objects, the crashes above 600m occurred. I just wrote it off as the cell count causing it, and headed back into photoshop. Will do a simple test (remove sat map) and see if that has an effect.

Thank's for the feedback, Fitzee.

My map's performance is at about 50fps on a fairly good machine. It is lower on average than Northern Sahrani. Seems to be BIS's bush objects as much as the cell count effecting performance (long as viewdistance is at 500m, above that the cell count effects performance exponentially. Which is fine, I don't need it any higher than 500).

[ProwlerWolf 0]

Would a low-poly plant mod help performance?

[fitzee 7]

I didnt realize you could get an island/map into the game without a satmap.

No I dont have a lot of objects. just 1460, all inside the town, except for the road network.

If your satmap is effecting performance maybe its too high res? I used the formula from BI wikey 1 pixel = 1 meter.

I followed BI's directory structure and my map is binarized with working ladders etc. Not sure if the availble tutorials cover that stuff properly.

I hope you can get it sorted because your map is looking great.

[dmitri 0]

Would a low-poly plant mod help performance?

I think so, yes. It'd have to take things a little further than the Low Plants mod though.

I'm thinking purely gameplay here. I like the idea of having very dense grass\bush areas. To get that without a huge drop in performance, I'd be happy suffering compromised plant complexity.. Plants consisting of only a few poly planes with high res grass images with alpha channels mapped on.

VTE's amazing catalog of Islands are a good example. In the current alpha it's all converted from OFP, so in parts it looks a little dated, but the objects are very efficient performance wise, and awesome to play on when the AI decides to engage.

I didnt realize you could get an island/map into the game without a satmap.

Yup, if you skip the "import sat + mask" or move your layers folder out of the project directory. It'll show completely white in game. Not all that practical for actually discerning finer terrain features, but there you have it.

edit:

Did some tests, it's the sat map causing the crashes at 700m. Without it, I can increase the viewdistance to a poor performing 2000m before..the crash again occurs, likely due to the cell count. Cell count is set in stone, need to focus my attention on the sat map and the object density.

[dmitri 0]

At 700, ArmA crashes immediately. I realise performance will be low on such a map (though for my own goals, the performance hit is justified, and the low viewdistance does not effect gameplay..). I've given up trying to find a solution for the crash, I realise it's the terrains increased poly count.

While there is no cure for the high cell performance hit..just want to report back for anyone interested that there is a solution to the crash\viewdistance issue. It isn't caused by the high cell count. Rather the resolution of the sat\mask images. They need to be multiples of the grid size. I had assumed (..or not read through the Biki carefully enough) that multiples of 512 were fine. Which is a shame since my island's grid is 1024.

[rocket 9]

More polys = More Ram Usage + More CPU Usage.

There is no escaping that. Removing the vegetation nearby will help, but ultimately, by dropping below 10m cell, you're really talking ALOT of polygons. It would appear to me that the terrain geometry LODs, and this helps greatly. But below 10m cell size (even below 30m), you're rendering most of the area anyway so LODing won't help.

Because ArmA uses this geometry for working out so many things, such as where the drivers should go and how their vehicle will be affected, increased geometry for terrain probably has something close to an exponential impact on performance due to vastly increased mathematical calculations.

[dmitri 0]

Ah yeah, I am aware of all that.

Just wanted to update the thread in case someone else was experiencing a similar crash.