69782 performance and stability feedback

[MavericK96 0]

hi Suma,

Just curious if a testbed of different missions would give a better result?

So you could better see where the issues might be.

For example (or as interim workaround) which of the default SP/MP missions could be taken as "testmission"?

Which settings should people use for a proper test eg Viewdistance, Video Memory, 100% (lower/higher) Resolution etc.?

Yeah, it would be nice to have some sort of standardized program that would be used to help the devs better figure out issues. I'm definitely willing to put in the time to get this game running nice. :cool:

[darthmuller_cro 10]

Suma i make mistake in my post...i was thinking on visual preformance while moving mouse in the game. With 69782 i have more microsutters when moving mouse slow than 69645 which goes much fluid at same locations (tested many times), othervise everything is fine. I know how hard for you is and i am very satisfied with this patches, more then ever! I wrote my post to try help you not to insult you .

[langgis08 10]

I am convinced that even if we made no fixes at all, just released a patch with a new version number, there would still be people who would get worse (or better) performance from it.

IMO Placebo-Effect in combination with "we see what we want to see" ;)

The poll results tell the same story: both latest poll results have a big mean variation (statistical).

My little test results so far with newest beta patch 69782 in comparison to 69645

1) "white ground"-Problem seems to be fixed, really (flying, walking, driving -> white spots/areas are filled out with correct textures again)

2) no crashes so far (I never suffered from that in the past, anyway)

3) performance not better, not worse (concerning FPS-Count, but it "feels" smoother)

I know that I cannot expect high fps with my rig @all maxed out, so bug fixing is done very well and that's what matters mostly for me at the moment.

so big thx @all of you BIS, you're doing a great job

and thx for your excellent communication with the community:)

Edited April 29, 2010 by langgis08
statement

[f2k sel 164]

Can't get this beta. Links show this: :confused:

I had the same thing earlier, as mentioned above it's spyware.

I downloaded Malwarebytes http://www.malwarebytes.org/ and it sorted it out.

[InFireBaptize 0]

i don't care about fps, i can adjust the settings to low. My only concern is the micro stutter which is evident when you are the gunner on a moving vehicles (air/ground) and zoomed in. VON i can live without right now since TS3 is filling out the gap. Texture load/pop up is annoying but more of a cosmetic and i'm not sure if it is effecting performance. Best as the guys suggested above me is to designate a heavy resources mission to test the betas there or have good beta testers guys from this community test and report back to you.

[SWAT_BigBear 0]

I am convinced that even if we made no fixes at all, just released a patch with a new version number, there would still be people who would get worse (or better) performance from it.

Thats funny Suma...69782 release.

And now I've had plenty of time to have tested many options/tweaks, results are the same for me.

Edited April 30, 2010 by SWAT_BigBear

[MJK-Ranger 0]

This is exactly what makes PC development a nightmare. The only differences between 69645 and 69782 are a few simple bug fixes. I cannot see anything which could affect performance in any significant way. Even the "missing terrain" bug was a more or less a visual thing, with very little performance impact.

I am convinced that even if we made no fixes at all, just released a patch with a new version number, there would still be people who would get worse (or better) performance from it.

Hi Suma.

Thanks a lot for all your hard work and your team.

I have been testing this new beta and overall i feel much better. :)

Bu it's one thing that really give me BIG problems. FRAPS don't work, record a movie and all stop up with 1 - 2 FPS, my HD don't work like it should when recording.

With beta 63826 no problems at all. :S

I hope u you BIS can look into this issue.

Thanks

[gulag 10]

IMO Placebo-Effect in combination with "we see what we want to see" ;)

The poll results tell the same story: both latest poll results have a big mean variation (statistical).

My little test results so far with newest beta patch 69782 in comparison to 69645

1) "white ground"-Problem seems to be fixed, really (flying, walking, driving -> white spots/areas are filled out with correct textures again)

2) no crashes so far (I never suffered from that in the past, anyway)

3) performance not better, not worse (concerning FPS-Count, but it "feels" smoother)

I know that I cannot expect high fps with my rig @all maxed out, so bug fixing is done very well and that's what matters mostly for me at the moment.

so big thx @all of you BIS, you're doing a great job

and thx for your excellent communication with the community:)

Yes, well said, I got the same results, no white ground, no crashes (so far), same performance but more smoother.

Thanks BIS for your efforts to make this game better.

[stun 5]

Never mind.

Edited April 30, 2010 by stun
off topic

[thr0tt 12]

As Suma said, PC dev is nightmare and with such a complex engine then getting a single test for all may not be possible which is why I suggested sticking to the benchmark tests initially which uses a lot of objects and effects to see if there are any problems.

Unless stated otherwise then looking for the extra FPS anywhere could well be a placebo effect but odder things have happened when bug fixing.

...also this is just a beta so really all that shouuld be possted here are the hard facts i.e. issues with terrain, sound, AI or whatever, you could possibly then post benchmark 1 = +10 fps increase in the test after running it 3 times or -5 fps hit on 2 benchmark.

What also would be useful is those systems specs and settings are posted, I mean, we all must have different settings and if someone using the lowest possible VD, terrain, no shadows etc. is getting a 10 fps increase then that mean nothing to me when I am using all high settings with VD of 3k etc. Dwarden also mention posting your driver version and OS version, which too could be useful to recreate the issue or at least bug fix the beta.

[jw custom 56]

This and the previous patch seems to run fine, but as soon as you deal with a lot of AI and have a lot of scripts running things get real slow.

[EricM 0]

This is exactly what makes PC development a nightmare. The only differences between 69645 and 69782 are a few simple bug fixes

@ suma : the problem here is that the poll is badly named (which earlier versionS should we compare to exactly?).

People may not compare the same things. I voted "stable" because I was comparing 69645 and 69782 and had same fps but no more white ground.

Some people may compare 69782 with vanilla 1.05 or 63826, in which case other factors enter into play.

I'd say next time, say in the poll title exactly which two versions should be taken into account.

[turin turambar 0]

This is exactly what makes PC development a nightmare. The only differences between 69645 and 69782 are a few simple bug fixes. I cannot see anything which could affect performance in any significant way. Even the "missing terrain" bug was a more or less a visual thing, with very little performance impact.

I am convinced that even if we made no fixes at all, just released a patch with a new version number, there would still be people who would get worse (or better) performance from it.

You need to establish some hard guidelines so people can make the same repeteable benchmarks, labeling their hardware specs, software configuration, and everyone using the same tool to measure the exact same thing. You need something that gives a detailed log of framerate in a specific timeframe, and from there you can obtain average FPS, minimum FPS, stutter, etc. In hardocp benchmarks uses graphics like that

http://hardocp.com/image.html?image=MTI3MTAzNjc0MjVDaThvQWU0U3JfM18zX2wuZ2lm

We need something like that, as it can give you average fps, minimum fps (lowest point) and stutter (high variance in fps, instead of a smooth line).

[WillaCHilla 10]

We need something like that, as it can give you average fps, minimum fps (lowest point) and stutter (high variance in fps, instead of a smooth line).

...Exactly + the benchmark-mission should not differ so heavy in results because of the dynamic AI.

f.e. benchmark 1:

1st run:-chopper is not shooting rockets --> higher FPS

2nd run: chopper is shooting rockets-explosions --> impact on FPS

- based on this benchmark-setting it is really hard to get a definite value for a specific setting.

Edited April 30, 2010 by WillaCHilla

[alef 0]

News from Suma at dev-heaven.

Does this still apply for ArmA2?

[MavericK96 0]

News from Suma at dev-heaven.

Does this still apply for ArmA2?

God, I hope it's fixed. That's the major reason as far as I can tell for getting crashes/sudden slideshow lag in this game, is improper virtual memory use.

[jw custom 56]

...Exactly + the benchmark-mission should not differ so heavy in results because of the dynamic AI.

f.e. benchmark 1:

1st run:-chopper is not shooting rockets --> higher FPS

2nd run: chopper is shooting rockets-explosions --> impact on FPS

- based on this benchmark-setting it is really hard to get a definite value for a specific setting.

There's more than FPS to test, things might appear fine and awesome in one environment but really poorly in others.

So just having a standard mission to test with might not give the best result!

[connos 10]

News from Suma at dev-heaven.

Does this still apply for ArmA2?

You can change the Arma2 exe to be large address aware but it doesnt make any difference as its hardcoded not to use more ram. For me the -maxmem=2047 command is giving me crashes.

Edited April 30, 2010 by connos

[otrebla_snake_ita 2]

You can change the Arma2 exe to be large address aware but it doesnt make any difference as its hardcoded not to use more ram. For me the -maxmem=2047 command is giving me crashes.

Yeah, developers tried using largeaddressaware flag but they say it didn't work...probably they have to change a lot their engine in order to use ALL the ram users have...That would be an incredible performance boost

[Proton 11]

;1618888']Yeah' date=' developers tried using largeaddressaware flag but they say it didn't work...[/quote']

Where do you have this info from?

[CarlGustaffa 4]

Eventually voted "Slower - it runs a little bit slower, no new crashes or stability issues" as my best bet, as the "Mixed" text didn't really fit the bill.

There are cases where I feel the performance is greatly increased, such as in open plains where looking around is a lot more smoother. And while flying, except I'm not much of a flyboy.

However, some of the trees (now also young conifers which I haven't had the same degration with before) are now actually worse than before. In order to get my frames back under control, I have to go all the way down to LOW textures (from Normal), and that just looks horrible. It is also unfair since AI will spot and shoot me through the leaves which for me now are big ugly globs of green which I cannot see through.

I may drop from 18-25 FPS (playable, but could be better) to 2-6 FPS (extremely unplayable) when I go from low textures to normal textures in problematic woods.

The patch might be a blessing to flyboys, but I think it has made it worse for the infantry.

Note that all these observations is done while running ACE and rather heavy mission (Domino on dedicated), I simply don't have time to test without.

Edit:

As for stability, it is rock solid as long as memory isn't too fragmented. Most of my friends who are having stability issues, are those with high spec PCs who insist on running with highest texture settings. I haven't dared alt-tab yet from a windowed editor, to see if that still causes hangs though.

I just did. And it hangs.

Edited April 30, 2010 by CarlGustaffa

[Joseph Archer 10]

Great patch for me, (i7@3.8ghz 4870x2 at 1600x1200). More stable in game and in the menu system! Faster/smoother in game and the AI are like new men. Is this a sign of things to come?

[langgis08 10]

Is this a sign of things to come?

This is a sign of how it is :) -> ftp://downloads.bistudio.com/arma2.com/update/beta/ARMA2_Build_69782.log says "current patch candidate"

[mrn 0]

Slower - it runs a little bit slower, no new crashes or stability issues

No crashing, alt-tabbing fine but very erratic framerates. Average on 1.05 25-30 consistently, with this beta average seems to be 20-25 but with occassional (5-10 seconds) drops down to 1-2 fps which make it unplayable. Tested on Chernarus and third party islands in editor and vanilla missions.

Shame because the AI seem much better.

System in sig, ATI Cat 10.2, will try 10.4 see if that helps.

Installed 10.4 and slowdowns have been eliminated. Average is back to 25-30 with beta.

Edited May 1, 2010 by MrN
Update after driver install

[mosh 0]

Runs better than previous beta (same, as before last beta patch, is what I voted). The last beta slowed things down but this one seems to be ok from what little testing I've done. I'll test some more though and report anything major.