|
what map(s) are you referring to?
i think it is safe to say that most of the tpw maps have more aesthetic detail than the average ladder maps. (the newer ladder maps being alot less detailed than the original release day map pool as well)
that of course increases the performance costs and can cause a significant drop in fps (not actual lag) if you are playing on an older/weaker computer or very high graphic settings.
of course this is something we keep in mind when working on maps and make sure it still runs smooth on our machines. but we don't actually have a number of different testing systems available and depend on feedback from the community.
if there are indeed problems on our maps that is something that we take seriously, but it is also something that can rather easily be fixed by reducing some detail.
i also wouldn't rule out the possibility of players who experienced lag presuming it is the map that is causing it since they haven't played it before and the map being a possible cause of it (even tho it might not be related).
|
sometimes ladder maps like metalopolis or abyssal caverns lag (fps drops), i have no idea why, it has happened a few times only
|
yeah, metalopolis top starting position lags (at least it has before it was out of the mappool)^^
|
I have a pretty shitty pc, so I lag in maps lag Antiga and Metalopolis (I think is due the smoke)
and Dual Sight, Crevasse etc, they have A LOT of details all over the maps
Late game is almost no-micro
|
The smoke is an abslotule evil. My PC could run SC2 perfectly (on lowest, ofc), if it weren't for these stupid smokes. I don't need to see them moving, I could even have just a solid wall of grey, I don't care - but you can't make this happen with graphic settings. So it adds for me a new level of strategy - I must avoid areas with smoke for fights.
Considereing that almost anything runs well, but the smoke, I strongly suggest to the mapmakers to avoid using smoke effects altogether. I think you will make life better for many people with a very simple thing.
|
XNC had quite a number of issues resulting from the hole in the center of the map, which caused Fps drops on less powerful computers.
|
I lag almost every game on District 10, 4v4 map. I don't lag any other time in SC2 except this map. I'm sure the massive amounts of stuff going on, and units on the field in 4v4 has something to do with it, but I don't lag on any other map.
|
I have to veto Xel Naga Caverns (a favorite of mine) because whenever I play it, the damn thing lags like a mofo. Whenever I tell people in tournaments and stuff they think I'm bullshitting because I don't want to play the map but I reallly actually want to play XNC.
I have a piss poor GPU that I am in the process of replacing. I never thought XNC had a lot of doodads or special lighting?
|
|
|
Bnet syncs everybodies sounds, so custom sounds can make maps like mafia lag a lot even though it isn't taxing the computers resources.
|
xel naga lags for me because of all those flowers ;(
|
The only map I have had complaints about lag was on Marshlands, because there were too many trees (5500-ish) and trees aren't too kind on the GPU as there are lots of layered transparency combined with animations (especially when they were arranged as they were in Marshlands, with tons of trees overlapping each other).
On November 01 2011 11:00 Barrin wrote:
As said before, too many doodads and doodads with a lot of particles can lag a lot.
As a rule of thumb, more than 2000 doodads or more than 2mb map file is an indication that the map might lag.
it can also be a lot of doodads really close to each other
Look at the Overview Manager (Hotkey: F12) to see what is taking space in the compressed map file.
Most of my maps reach about 1.5–1.8 MiB before they are finished, much of that data is not doodads, it is the texturing which even when compressed eats ~95% of the storage space because I try to make it detailed.
So all my maps are 1.5Mb+, even Monsoon Valley which is the smallest competitive SC2 map I've so far seen (smaller than Steppes of War) is really large in file-size because of the textures.
However, I do not think the texturing is causing much of an impact on the FPS, instead there are lots of other factors, some of which already have been mentioned, as particle emitters (eg. fires and smoke).
Rain does not have much of an impact on the fps, on Marshlands the tree heavy areas lowered my fps from ~35 to ~25, and when I removed the rain to test, it only increased the fps by ~2 (on medium graphics, 1920x1200px, which my computer can't really handle in game (ultra+ works fine in editor strangely), I usually play on low). I had to remove about 1000 trees to fix the FPS problem (reducing the amount of overlapping trees quite a lot).
Also, so far I have not received any complaints about Concrete Dreams, which fulfills your rule of thumb really nicely (2143 doodads and 2 MiB file), although it has two custom textures which lower the amount of "real size" to about 1.5 MiB.
On November 01 2011 12:01 Cyber_Cheese wrote:
Bnet syncs everybodies sounds, so custom sounds can make maps like mafia lag a lot even though it isn't taxing the computers resources.
IIRC, Bnet also syncs the triggers, so a trigger cannot execute only on one player's box, it has to execute on all of them simultaneously which can make the game lag quite a lot if there are triggers which are to be run very frequently, especially if someone has a high latency.
But as we currently seem to discuss melee maps this does not really apply as much.
|
FPS tests on NASL Open Maps
In combination with the NASL Open I have made some tests on the TPW maps played there and compared with a reference map from Blizzard (Metalopolis).
All of these were run on a MacBook Pro 17" from early 2009 with an upgraded CPU:
CPU: Intel Core 2 Duo 2.97 Ghz
RAM: 8 Gb 1066Mhz DDR3
GPU: Nvidia 9400M, integrated graphics with shared VRAM
Screen Resolution: 1920x1200px
Starcraft 2 settings: Low on everything, 1920x1200px resolution
I did not use the Nvidia 9600M GT card (as that runs everything on low/medium in SC2 in 60 FPS), instead the integrated 9400M was used to produce larger changes in the FPS numbers. This test only focuses on the performance impact on the graphics card, I did not have a computer available with less ram or a slower CPU.
To measure FPS, the integrated FPS counter in SC2 was used (ctrl + alt + F).
First, before we start looking at the maps, another reference can be used: The Starcraft 2 menu, which had 26 FPS
(yes, it is bad for some reason, in game is better)
All maps were played using the uploaded maps which are on battle.net, while signed in, and using Protoss vs an Easy Computer (this to have the game running as close to "normal usage" while measuring things). I went for a quick Stargate and chronoed out one Phoenix to use to move around the map while watching the FPS counter.
Maps
First out is Metalopolis, to give us some reference to the numbers later on:
Average FPS: 32
Lowest FPS: 23, next to the LoS blockers in the main
+ Show Spoiler +
Then a map which has lots of doodads and one you can expect to have low FPS: TPW Concrete Dreams:
Average FPS: 31
Lowest FPS: 25, next to some fires
+ Show Spoiler +
An older map which has less doodads than most newer maps: TPW One Must Fall:
Average FPS: 30
Lowest FPS: 25, but only when the camera is "inside" a god-ray which is a very small portion of the map
otherwise the lowest was 27, right over the LoS blockers on the map.
+ Show Spoiler +
Now a map which has had some complaints about lag: TPW Ohana:
Average FPS: 31
Lowest FPS: 23, in the middle caused by the numbers of trees there
+ Show Spoiler +
TPW Overgrown
Average FPS: 30
Lowest FPS: 27, northern part of the center
+ Show Spoiler +
TPW Lunar Station
Average FPS: 31
Lowest FPS: 27, in the middle over pipes
+ Show Spoiler +
TPW Emerald Jungle
Average FPS: 33
Lowest FPS: 23, trees and waterfalls
+ Show Spoiler +
TPW Damage Inc
Average FPS: 30
Lowest FPS: 24, map edge with lots of generators and some fires
+ Show Spoiler +Note: This picture is wrong, but the map was the latest and the circled areas are the correct ones, but on the new map ![[image loading]](http://i.imgur.com/KiTKsl.jpg)
Finally I also tested TPW Burning Altar which is among the top 7 the TL Map Contest:
Average FPS: 32
Lowest FPS: 22, caused by waterfalls and trees on the edge of the map, again
+ Show Spoiler +
Conclusions
Short summary: Trees, waterfalls, fires and godrays are bad when clumped up. Especially trees and waterfalls.
The test had some unexpected results, I never thought Concrete Dreams would fare so well (and even being better) compared to Metalopolis, as Concrete has over 2000 doodads and would lag a lot more than Metalopolis according to my guesses. It can partially be explained by the smoke the LoS blockers produce in a clumped up area, and it can also be related to the Korhal buildings, if they require a high polygon count or something similar.
It didn't occur for me either that waterfalls would have a large impact on the FPS, as clearly is the case when looking at Burning Altar and Emerald Jungle. So not only are clumped up waterfalls annoying because they make a racket, but they also lower the FPS when looking at them 
Ohana had low FPS as expected in the heavily forested areas of the map, something I previously encountered with my map Marshlands.
Some final comments:
The belief that TPW maps are overly decorated can be disputed, but according to these tests the extra decoration does not hurt the FPS as much as one would think. So to all map makers out there; continue to make your maps beautiful and do not worry too much about the FPS, as that is only lowered by specific edge-cases!
|
This is very nice to know.
I got a question about trees. Some trees are animated, some are completely static. Most interesting is bel'shir and aiur trees cause they are very similair, the difference being that bel'shir is animated so you can see them blowing in the wind while aiur trees are totaly static.
The maps in which trees caused FPS drops were they all animated trees? I could not imagine static trees would cause FPS drops.
I would assume that if you make a forest/beach/etc style map with a lot of trees(sometimes 1000+) you would want to pick static trees to ensure good FPS.
|
dezi
Germany1536 Posts
|
On November 16 2011 20:27 Archvil3 wrote:
I got a question about trees. Some trees are animated, some are completely static. Most interesting is bel'shir and aiur trees cause they are very similair, the difference being that bel'shir is animated so you can see them blowing in the wind while aiur trees are totaly static.
The maps in which trees caused FPS drops were they all animated trees? I could not imagine static trees would cause FPS drops.
According to my speculation, all trees which have leaves use lots of transparency in their textures. This might lower the FPS when you layer the leaves so there are many levels of transparency which has to be calculated (this is not affected by animations, and is almost completely separate as the total transparency needs to be recalculated when you move the camera, and as that is happening very frequently I do not think it is cached). This is probably why clumped up trees lag, as there are many levels of leaves which has to be calculated for each frame (maybe around 8-16 levels compared to a single tree where you only might have 2-4 levels of textures).
But animated trees should most probably prove to be worse when it comes to FPS, as you need to update the polygons and so on between each frame.
On November 16 2011 20:32 dezi wrote:
Wtf is a godray Oo
See http://en.wikipedia.org/wiki/Crepuscular_rays: "Crepuscular rays (/krɨˈpʌskjələr/; also known as God Rays)"
I don't recall the name of the doodad at the moment though.
|
dezi
Germany1536 Posts
You could just have written sunlight ...
|
I've had a few weird issues on ladder maps. On a fairly decent setup (Q6600, 5770 ATI card) the Slag Pits would always give me crappy FPS and obvious delay (Lag Pits) and Twilight Fortress in 2v2s. Some versions of some maps seem to lag some computers, I remember Sheth having to cycle through various versions of Xel'Naga to find one that didn't give him dodgy FPS.
As a few have commented, FPS is technically different to lag. So many SC2 people seem to use them synonymously, and I've even got into the habit of it myself.
|
In Jungle - Environment you find "Sunrays", that's the doodad.
Do anyone know if mist doodads, or sand storms cause FPS drop? It's particle effects like smoke, but I asume they are less taxing, beacuse they cover a larger area. I havn't been able to test
|
On November 16 2011 21:18 scruffeh wrote:
I've had a few weird issues on ladder maps. On a fairly decent setup (Q6600, 5770 ATI card) the Slag Pits would always give me crappy FPS and obvious delay (Lag Pits) and Twilight Fortress in 2v2s. Some versions of some maps seem to lag some computers, I remember Sheth having to cycle through various versions of Xel'Naga to find one that didn't give him dodgy FPS.
As a few have commented, FPS is technically different to lag. So many SC2 people seem to use them synonymously, and I've even got into the habit of it myself.
Note: This is only speculation concerning FPS vs Lag as I have no idea about how the SC2 engine works, or actually graphics engines in general (I have not made any programming for 3D games, but I've read a tiny bit about it).
I guess the engine uses a "standard" run-loop style as it uses one CPU core as much as possible (although a few tasks can be run in parallel even when one task is hogging one core by itself). This means that the engine most probably works something like this (extremely simplified):
1 Sync simulation with other players
2 Step the game simulation one step forward
3 Update the models and scene data combined with the camera
4 Render the scene and pass it to the screen
This will force the game to wait for the scene to be rendered (step 4) before going back to step 1, meaning that if you have low FPS it will most probably impact the game's ability to simulate the game and sync with the other players. This is probably what is causing the "Foobar is slowing down the game" message when the person's graphics card has a hard time rendering all the stuff on screen, as then the game has to wait for him. It can also be caused by the CPU being loaded by lots of triggers or units to keep track of as that too will slow down the loop (step 2).
Obviously the game has some stuff built into it to prevent the impact of this kind of lag, probably some kind of parallelism to prevent the game speed of slowing down too drastically when the FPS reach something around 10-ish, but also something to prevent too much lag with other players over networks (step 1 is probably not run for each frame and the same goes for step 2, but instead the game simulation is probably step-based, like in Supreme Commander which simulates the game at 16 FPS (iirc), while interpolating the graphics for the frames in-between).
So you can have lag from several sources:
- The game taxing the GPU too hard resulting in long render times which can interfere with the run-loop and subsequent syncing with other players (it also makes the game somewhat unresponsive as you have a larger input lag).
- The game using lots of CPU during the simulation step causing the game to wait for the simulation to be finished before syncing state over network.
- Finally network can also cause lag as that will force the game to wait for the sync to finish before continuing with the simulation (if we assume the 16 simulation steps per second and a naive implementation of the game simulation loop the highest working ping will be around 62 ms before causing lag).
I will have to note yet again that this is only speculation. If anyone actually have some experience with 3D engines and multiplayer programming, feel free to correct me.
|
|
|
|
|
|
EPT![[image loading]](http://i.imgur.com/FV0J4.jpg)
![[image loading]](http://i.imgur.com/ZtlyQ.jpg)
![[image loading]](http://i.imgur.com/MYuWU.jpg)
![[image loading]](http://i.imgur.com/VDJTm.jpg)
![[image loading]](http://i.imgur.com/7Zth7.jpg)
![[image loading]](http://i.imgur.com/CPMh7.jpg)
![[image loading]](http://i.imgur.com/Lz3iq.jpg)
![[image loading]](http://i.imgur.com/KiTKs.jpg)
![[image loading]](http://i.imgur.com/TST2K.jpg)
