Patch 69645 - changes under the hood

[jw custom 56]

Never actually used that one but I am very partial to GB and I've experienced it on all of the boards I've used (and several different i7s).

It's not a big deal at the end of the day as adding 'cpucount' takes all of 1 second but I'd really like to know what changed in that patch that borked it on the 980 because I definitely did not have it before I applied the patch.

Are you on a 64 bit Win 7. I never got it under 32bit XP so maybe that is the catalyst?

I'm on WinXP 32Bit :)

[bangtail 0]

I'm on WinXP 32Bit :)

Ahhh, definitely the best OS for ArmA 2 if you want the least possible problems.

I really should have stated that I was referring to Win 7 64 I guess, DOH!

Seems to be a problem with the way that Win 7 handles the virtual cores.

Edited April 25, 2010 by BangTail

[dslyecxi 23]

Having used the new beta for our Saturday session, we noticed a great deal of crashing above and beyond the normal occurrences. The vanishing terrain bug showed up a good amount (including a mission that took place on Everon that we had to abort due to it basically breaking the mission) as well. I ended up crashing twice myself, which is very unusual - each time A2 would freeze up and become unresponsive, and when I got back to windows (via ctrl-alt-del, task manager (in W7)) it would say that "ArmA had stopped responding". People who normally do not crash were crashing, and it seemed that we had a high percentage of crashes when loading into missions or ending them.

I plan to revert us to the prior patch for now - any performance gains that came from this latest one have been offset by the much more frequent crashing that is in it.

Crashlogs will be sent in shortly.

[nkenny 1057]

On my home computer (a phenomx4, GF285, winXP 32b) the patch hasn't introduced any performance changes so far. On my laptop however the change is staggering. (Intel 2ghz, GF<something or other> winVista 32b)

Arma2 runs blazingly fast on my laptop-- The benchmark1 has increased some 4-5FPS to 30+

-k

[Drew 10]

Thanks Alex :)

[Nazul 10]

Good performance increase for me. Win7/i7/6gig/gtx295.

I haven't played for (since Bad Company2 was released) a while :)

I ran my usual test mission Counter Attack and was getting no texture swapping like usual, could spin around and no texture lag.

Got an error loading the mission- No entry "bin\config.bin/CFGmarkers.faction_CDF" and there was no markers on map screen at intro.

Loaded the benchmarks and got missing ground textures. :(

Edited April 25, 2010 by Nazul

[MavericK96 0]

Ahhh, definitely the best OS for ArmA 2 if you want the least possible problems.

I really should have stated that I was referring to Win 7 64 I guess, DOH!

Seems to be a problem with the way that Win 7 handles the virtual cores.

Just for the sake of clarity, what does the cpuCount=4 actually "fix" for the i7 920? I have the same CPU and I don't use the attribute, but I haven't really noticed any problems, per se. Does it just increase performance? I feel like I've adding the command before but not really seeing a difference in core usage or anything. I'm also running Win 7 64, same as you.

[astral54r 10]

69645 creat new artefacts :)

[MavericK96 0]

(Cross-posting to this thread also because I meant to actually post it here)

Just did some minor testing with Benchmark 2:

Settings:

AA = Low

Video Memory = Default

VD = 3905.

All else, Very High

Tests:

ArmA2 1.05, no command line modifiers: 20 FPS

ArmA2 BETA, no command modifiers: 21 FPS

ArmA2 1.05, using -cpuCount=4: 21 FPS

ArmA2 BETA, using -cpuCount=4: 20 FPS

Conclusion:

So, pretty inconclusive. Or rather, shows me that the beta didn't make any real framerate difference for me, nor does using the cpuCount attribute on my i7 920 seem to do anything for framerate.

Notes:

-My HT is turned OFF. This is because I'm overclocking and having it on raises temperatures considerably. I turned it off a long time ago because after reading benchmarks for gaming and normal use, almost nothing takes advantage of HT with 8 cores.

-In the beta, the game ""seemed"" like it had more stutter. Of course this is subjective, as it's not a framerate issue, but it seemed to stutter a lot more than 1.05.

-The beta also had a glitch where the mountains in the far distance would sort of disappear at the beginning of the benchmark and then reappear as you got closer. This did not occur at all in the 1.05 version. On the second run of the beta (with cpucount=4 enabled), it was less noticeable, but they still seemed to "warp" slightly.

-I didn't get any sort of ground textures disappearing or white boxes at all in either beta run.

[maddogx 13]

Played a 3 hour MP game yesterday evening. Performance is great, experienced no crashes or other major issues but did come across the disappearing ground bug for the first time. Only happened twice very briefly over the course of the whole game though. It seems to occur in specific places and only from certain viewing angles, as if sections of the ground were being mistakenly identified as "off-screen" and thus being culled.

[suma 8]

Issue created for "disappearing ground" issue. Please, add repro at http://dev-heaven.net/issues/10319

[EricM 0]

Repro is an easy one for me : reboot the game, open benchmark mission 1 with vd@1600m, watch the ground disappear on the slope of the hill and at the end of the camera movement near the house. It's systematic.

Edit : I've recorded a simple video showing how it appears again exactly at the same time, at the same place, each time in the benchmark mission 1 for me. No user input needed, 100% reproductible for me.

Edited April 26, 2010 by EricM

[CarlGustaffa 4]

I think performance issues for those with the bug can be ignored. When this happens, a lot more stuff is often visible, such as tree roots and additional landscape behind the scene, and units hiding behind a ridge not so well anymore :) All these new polygons due to a few lacking ones that have little cost to render.

Yes, it does remind me of a culling problem. One pixel change in view direction is enough for toggling the bug.

[jpinard 10]

Issue created for "disappearing ground" issue. Please, add repro at http://dev-heaven.net/issues/10319

Suma - Just wanted to mention how awesome you guys are. Thanks so much for an amazing game. :)

[AnimalMother92 10]

Same here, I can repro with Benchmark 01. I added a 7zip with screen shots and the requested files to the ticket.

[EricM 0]

Repro video link added to bugtracker.

[stun 5]

I'm getting the white ground bug on my system - where you can see the tree roots etc. Also A drop in FPS and a strange kind of stutter. This stutter seems different to that which I have encountered previously. It feels like a very quick surge and is evident when using ground units particularly when I'm side stepping and look in another direction to where I'm moving. I don't think I'm imagining it - overall something feels wrong when I'm walking about.

I'm using xp sp3, E6400 C2D @ 2.68 Ghz, 2 gig ram, 8800gt (factory overclocked) with 197.45 drivers.

Benchmarks

Vanilla 1.05 (62017)

Benchmark 1 45 fps

Benchmark 2 17 fps

previous 63826 Beta

Benchmark 1 50 fps

Benchmark 2 17 fps

New 69645 Beta

Benchmark 1 44fps

Benchmark 2 16 fps

My results show that I was getting 6 fps more using the previous Beta. So no gain for me. I'm hoping the loss of fps is down to the bug with the ground.

For now I'm reverting back to the previous betas.

Edited April 26, 2010 by stun

[pchaxor 0]

Missing textures "white textures" and transparent ground textures happened every time I used the beta. Vanilla and ACE 2 - made no diff.

Upgraded to latest Nvidia Drivers and played with settings to no avail. The gameplay is smooth and stable with no crashes.

Thanks!

(I reverted back to the latest release due to missing/invisible textures; not gameplay or stability issues.)

- I would be interested in testing more should the texture issue be resolved.

[WillaCHilla 10]

-testet the new beta yesterday with some of my mates during private coops and online-mp (domination and warfare in this case).

impressions:

performance:

In general, the performance seems to be the same overall, but the increase of the viewdistance isn´t effecting it so hard anymore. I was able to set it to 4000m-4500 with the same smoothness as it was before the patch with a Vd of 3000.

Important one I noticed, if I start the game (default world:Chernarus) it needs a bit longer to get to the main-menu as before the patch was not installed.

graphical bugs:

For me, I wasn´t able to detect any visual glitches, etc (hardware listed in my sig). 50% of the others, who were on ts (pc collections of Duals and Quads, mostly with ATIs VGAs (4890, 5850)) had the graphical bugs which are already mentioned above.

AI-performance:

mh-seems the ai follow the orders a bit faster (e.g. movement in dangermode), but they still have their own head (don´t really quitting the danger-mode in combat).-But this needs a bit more testing by my side!

Have to correct my previous statement:

Got the ground texture-bug yesterday on Chernarus and Isla Duala as well.

[NoRailgunner 0]

Tested yesterday on a duo core @3.0Ghz with nvidia 9800gtx+ and vista32. Viewdistance set to 2500m, gpu memory = default, AF+AA+shadows = off, rest set to normal.

The ground disappear or changed into white on many places - visible in default A2 Missions:

- Benchmark1

- the mission (dont remember the name yet) where you had to lase the BRDM2 and secure the village commanding other units.

The average drop with this beta build is 5 fps (with A2 benchmarks too).

Stuttering units still visible especially if they disembark and move away from vehicles and planes.

AI are a bit better but they still need more fixing and improvements.

[CarlGustaffa 4]

Resolved in 69686, hoping we'll get a new update pretty soon.

[bangtail 0]

Just for the sake of clarity, what does the cpuCount=4 actually "fix" for the i7 920? I have the same CPU and I don't use the attribute, but I haven't really noticed any problems, per se. Does it just increase performance? I feel like I've adding the command before but not really seeing a difference in core usage or anything. I'm also running Win 7 64, same as you.

It alleviates what is probably attributable to 'cache thrashing'.

It seems to be unique to Win 7 (and maybe to 64 bit versions but I've never run Win 7 32 so I can't confirm that).

Some people actually have to disable HT in the bios in order to get rid of it although I have always managed to circumvent it with just the command line switch.

[Zipper5 74]

Resolved in 69686, hoping we'll get a new update pretty soon.

Very nice. Good job, BIS. Looking forward to the next update. ;)

[ScareCroweb 10]

Just remove the grass and everything is fine, to be honest BIS should force the grass off for everyone. The grass is evil and makes NOBODY happy! Anyway good job BIS for fixing the bug

[jpinard 10]

It alleviates what is probably attributable to 'cache thrashing'.

It seems to be unique to Win 7 (and maybe to 64 bit versions but I've never run Win 7 32 so I can't confirm that).

Some people actually have to disable HT in the bios in order to get rid of it although I have always managed to circumvent it with just the command line switch.

Disabling HT in the BIOS is the recommended option for gamers. Not a single high fidelity 3d game benefits from virtual cores as it puts too much strain on the processors. 3d titles suffer a substantial performance hit with it on, increases heat and electrical load, plus it lowers your overclocking potential.