Bug: FPS loss with Video Memory on Default rather than V.High

[doveman 7]

I did some testing and found the following http://forums.bistudio.com/showthread.php?85124-ArmA2-OA-(low)-performance-issues&p=2230207&viewfull=1#post2230207

These are on Default:

Straight ahead - 69 FPS

Scoped - 63 FPS

Zoomed scoped - 62 FPS

Straight Up - 155 FPS

Behind - 85 FPS

and these are the same on V.High:

Straight ahead - 83 FPS

Scoped - 78 FPS

Zoomed scoped - 75 FPS

Straight Up - 195 FPS

Behind - 102 FPS

As you can see, there was a substantial increase with Video Memory set to V.High as opposed to Default, which is weird as everywhere recommends to use Default as that supposedly allows the entire available VRAM to be used, whilst V.High limits it (to 1024MB it seems from my tests). In that test, it used around 550MB and 100% GPU on either setting, so I don't know what else that setting's doing that makes it run so much better on V.High.

Anyway, I made a ticket about this, so if you want to see if you can confirm this for yourselves and vote for it, it's at https://dev-heaven.net/issues/53313

Note I also tested on E08 where it didn't make any difference but perhaps that's limited by something else. I also tested on a PR server the other night and got maybe 5 FPS extra on V.High. Probably the best way to check is in the same City 34 (in the N-NE) of ‘51km Desert’ terrain that I got the above results on and similar settings which were as shown in this screenshot (and given below as well):

http://imageshack.us/a/img692/6158/arma2oa2012092600500310.png

VD 3000, Texture Detail - V.High, Video Memory - V.High (or Default), AF - Very High, AA - Low

Terrain Detail - Low, Objects - V.High, Shadow Detail - V.High,, HDR - Normal

PP - V.Low, ATOC - All Trees, PPAA - SMAA V.High, Vsync - Disabled

[kklownboy 43]

anything over your Displays Refresh rate doesn't matter.

[.kju 3245]

use the pvpbenchmark suite - it has no AI but stresses your GPU a lot

[doveman 7]

anything over your Displays Refresh rate doesn't matter.

Well most of the time I'm lucky to get over 30! That test just helps to show the lower FPS using Default instead of V.High, which shouldn't be the case.

---------- Post added at 18:53 ---------- Previous post was at 18:52 ----------

use the pvpbenchmark suite - it has no AI but stresses your GPU a lot

Those tests I did on 51km desert had no AI either. I'm not sure I need to do any more tests, I think the ones I've done clearly show that Default has a problem.

[.kju 3245]

if you think an empty desert terrain, is a valid test, so be it

[doveman 7]

if you think an empty desert terrain, is a valid test, so be it

It shows that I got considerably higher FPS on V.High rather than Default doesn't it, which is the whole point. Why should Default, which is meant to allow ArmA to use all available VRAM, perform worse than V.High, which is meant to only allow it to use some (1GB?).

As I mentioned, I've tested in a couple of other situations and noticed an increase (not so large) there as well.

And why are you suggesting an empty desert terrain (it's not actually empty, I was in city 34 with plenty of buildings) is an invalid test but suggesting I test with the no-AI pvpbenchmarksuite?

[chrisb 196]

If your into making your own missions or missions for groups, then I would recommend anyone use city 34 on the ‘51km Desert’ terrain for larger city style missions/scenarios. With over 100 ai in combat in the city, fps can still be 50+.

Its an ideal place to set this type of mission.

Not many cities in arma that could boast that.

Fallujah/Bagdad are similar, lots of ai and still reasonably good fps.

No good fighting clutter with your fps, much rather fight enemy ai..

Plus testing, it opens up your gpu/cpu and gives you a really good idea of how they are performing without the overboard clutter..

With the arma series you have to compromise where performance is concerned, find the best places to set your missions prior to making them, there are lots of places that give good performance.

Regards Vid Mem, always on V/High, always have with a2, why, just because it gives me better performance, whether it should or not. This is arma don’t forget, the impossible tends to be possible and vice versa, well that’s what I have found anyway.

Performance including fps is never a concern for me or the group I play with, make the missions to suit the machines being used, in a location that gives the greatest performance for the style of mission you want to make..

Do the same with testing, test on this and any other terrain to get a real idea of how your pc is going to handle the game for the missions you want to have, makes sense..

[ast65 10]

It shows that I got considerably higher FPS on V.High rather than Default doesn't it, which is the whole point. Why should Default, which is meant to allow ArmA to use all available VRAM, perform worse than V.High, which is meant to only allow it to use some (1GB?).

As I mentioned, I've tested in a couple of other situations and noticed an increase (not so large) there as well.

And why are you suggesting an empty desert terrain (it's not actually empty, I was in city 34 with plenty of buildings) is an invalid test but suggesting I test with the no-AI pvpbenchmarksuite?

well, I noticed the same behaviour (a bit faster on 'very high') but I get some texture popups/not fully loaded textures on 'very high' on the other hand.

This apparently applies only for 'texture heavy' scenarios and not empty desert terrain.

All in all I prefer 'no popups' at the cost of some fps by choosing 'default'.

[doveman 7]

well, I noticed the same behaviour (a bit faster on 'very high') but I get some texture popups/not fully loaded textures on 'very high' on the other hand.

This apparently applies only for 'texture heavy' scenarios and not empty desert terrain.

All in all I prefer 'no popups' at the cost of some fps by choosing 'default'.

True, just because it used the same VRAM on 51km Desert doesn't mean that will be the case on other terrains/missions and if it uses more VRAM on Default that's likely to help avoid popups. Still would be nice to know why we get better FPS on V.High though.

[ast65 10]

True, just because it used the same VRAM on 51km Desert doesn't mean that will be the case on other terrains/missions and if it uses more VRAM on Default that's likely to help avoid popups. Still would be nice to know why we get better FPS on V.High though.

seems logical to me since on 'default' the amount of data to process at once is bigger, thus a higher strain on the system?

just a guess, though :P

[doveman 7]

seems logical to me since on 'default' the amount of data to process at once is bigger, thus a higher strain on the system?

just a guess, though :P

It's logical in theory but in the test I did I don't think there was much difference in VRAM used yet the FPS was much lower on Default.