69782 performance and stability feedback

[suma 8]

How the patch 69645 works for you compared to previous versions (meaning 63826 or older)?

[alef 0]

It's broken: http://dev-heaven.net/issues/10340

Edited April 27, 2010 by alef
Reinstalled patch, fixed

[EricM 0]

It works here. Perfs are largely unchanged, maybe slightly better but it could be placebo. Same benchmark results as 69645 but without the white/transparent ground bug for me.

Question about the poll : should we compare to vanilla 1.05 or the previous beta ?

Edited April 27, 2010 by EricM

[stun 5]

The white texture bug is gone. Nice and smooth when flying. I am losing 5-6 fps compared with 63826 which is a bit of a worry though. I'm getting the same benchmark results as 69645. The best frame rates by benchmark have been with vanilla 1.05.

Benchmark 1 with 69782

41-44 fps

Benchmark 1 with 63826

45-48 fps

system: C2D E6400 @2.68 Ghz, 8800gt (factory overclocked), 2Gig Ram. Drivers tested 196.21 and 197.45. I get about 3 more fps with the 196.21 drivers on both Betas.

Edited April 27, 2010 by stun

[Pauldarrington 9]

Performance wise is about the same as previous beta i did suffer from quite a few CTD more then i normally would without using the beta

[alef 0]

It seems the "Receving ..." bug has changed to this "Error: Device freed".

I've got it more times yesterday playing online.

A way to reproduce it offline is to run Benchmark 01 and 02 a couple of time.

Full report and files at http://dev-heaven.net/issues/10342

[MavericK96 0]

Doing my standard routine of low, fast flybys over Chernagorsk with an F35B, I got massive slowdown after about 3 passes and had to close the game using CTRL-ALT-DEL. Looking in the .RPT file gave me a bunch of this stuff repeating:

Error during v-buffer creation: error 8876017c
Virtual memory total 2047 MB (2147352576 B)
Virtual memory free 88 MB (92520448 B)
Physical memory free 3128 MB (3280846848 B)
Page file free 4095 MB (4294967295 B)
Process working set 1288 MB (1351180288 B)
Process page file used 1510 MB (1583894528 B)
Runtime reserved 719 MB (753926144 B)
Runtime committed 586 MB (614985728 B)
Longest free VM region: 7536640
Small mapped regions: 8, size 32768 B

Also, after that repeated several hundred times, there was some of this:

Memory store: Failed mapping, already mapped 3980 KB, error 8
Virtual free 23416832 B, page free 4294967295 B, physical free 3275751424 B
Memory store: Failed mapping, already mapped 3984 KB, error 8
Virtual free 23416832 B, page free 4294967295 B, physical free 3275751424 B
Memory store: Failed mapping, already mapped 3988 KB, error 8
Virtual free 23416832 B, page free 4294967295 B, physical free 3275751424 B

Etc, etc. And then, the very last thing in the file, oddly enough, is this:

File read error: addons\plants2_tree.pbo,ERROR_NOACCESS
Error 3e6 reading file 'addons\plants2_tree.pbo'
ErrorMessage: There’s a problem with the disc you’re using. It may be dirty or damaged.

Which makes no sense, because the disc was never in, since it's not required.

Hopefully these error messages are helpful, though.

Performance seemed about the same as 1.05, with slowdown over the cities, though that is probably due to the memory leak or whatever this is occurring, since the slowdown got worse each time I flew over.

[stun 5]

I just got the same bug "Error: Device Freed". It was while playing a custom mission on Utes.

Mods: CA;beta

Distribution: 1465

Exe version: 1.05.69782

Error: cannot load sound 'ca\dubbing\global\radio\male05\en\engagingtargete.wss', header cannot be loaded.

*** Device lost - Driver Internal Error

Warning Message: No player was selected!

Warning Message: No player was selected!

Warning Message: No player was selected!

Warning Message: No player was selected!

Link to a007785e (Obj-239,256:94) not released

ErrorMessage: Error: Device freed

[MavericK96 0]

UPDATE: Turns out, doing the same test in 1.05 over Chernogorsk with the F35B yields similar results, though a different error message:

Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Vertex Buffer too small (17075<16384)
Vertex Buffer too small (22059<16384)

So it seems that turning off Threaded Optimization did not entirely fix the issue with Default Video Memory. :( Though the error is slightly different (now has that bit about the Vertex Buffer), it still gets the v-buffer creation error over and over.

Turning off Threaded Optimization definitely improved it, since before when I was doing the same test with a Cobra instead of the F35B, I couldn't get it to crash. Presumably this is because the Cobra is slower and didn't put as much of a strain on texture streaming, but it was really easy to get it to crash before in the Cobra, while turning off Threaded Optimization made it much more difficult, so it did have some effect.

Also, it's worth noting that doing the same fly-over test in the F35B, the beta actually did seem a lot smoother over the cities, until the slowdown started occurring due to whatever is causing the error messages.

[EricM 0]

[68694] Reduced parallax map artifacts on sloped terrain.

This one isn't working well for me.

Everything was fine in vanilla 1.05, but now I see the ground suddenly popping right under my feet when climbing the slopes of Utes, its very disturbing. It was a lot better before.

It seems that the radius of the circle defined to draw the parallax mapping on slopes is now way too small and not progressive enough.

Repro : Open Utes, place a man on the beach down stelka( 044019) and climb the slope to the village. You'll see the rocks suddenly growing off the ground, while it was much smoother and subtle before.

[h - 169]

FPS wise no diffrence between .63826 and this one but texture loading is definitely better, ArmaMark score difference is quite big, .69782 scores about 500 points higher..

Don't see any mentionable stuttering when flying over large cities etc on Chernarus..

Good stuff this far :)

Win7 Ultimate 64bit

4Gb DDR2 (800Mhz + OC)

Q6600 @ 3.2Ghz

ATI 4890 (OC'd) 1Gb / Cat. 10.3

Abit IP35 Pro motherboard

[ericfr 0]

1.05 -> Beta .69782

Bad

Official Benchmark 1

- Lose 16 % FPS

- More stutter

- White texture flash on trees

- Texture loading more longer

Radeon HD4670 1G, last drivers

[Dwarden 1125]

1.05 -> Beta .69782

Bad

Official Benchmark 1

- Lose 16 % FPS

- More stutter

- White texture flash on trees

- Texture loading more longer

Radeon HD4670 1G, last drivers

AMD.ATI Catalyst 10.3 WHQL or 10.4a preview ?

[ericfr 0]

AMD.ATI Catalyst 10.3 WHQL

[Dwarden 1125]

AMD.ATI Catalyst 10.3 WHQL

thanks

[Inkompetent 0]

Recommend trying out the 10.4a drivers. At least for my 4850 they've been quite pleasant.

[ericfr 0]

I'll try 10.4a drivers later Thanks

New test (10.3 drivers HD4670 1G)

Benchmark 2

Memory on default = missing texture ground (like the previous Beta) on the hill and artifacts

[jw custom 56]

I can confirm that the dissapearing ground issue is gone, performance is the same as with any other beta!

UPDATE: Turns out, doing the same test in 1.05 over Chernogorsk with the F35B yields similar results, though a different error message:

Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Error during v-buffer creation: error 8876017c
Vertex Buffer too small (17075<16384)
Vertex Buffer too small (22059<16384)

So it seems that turning off Threaded Optimization did not entirely fix the issue with Default Video Memory. :( Though the error is slightly different (now has that bit about the Vertex Buffer), it still gets the v-buffer creation error over and over.

Turning off Threaded Optimization definitely improved it, since before when I was doing the same test with a Cobra instead of the F35B, I couldn't get it to crash. Presumably this is because the Cobra is slower and didn't put as much of a strain on texture streaming, but it was really easy to get it to crash before in the Cobra, while turning off Threaded Optimization made it much more difficult, so it did have some effect.

Also, it's worth noting that doing the same fly-over test in the F35B, the beta actually did seem a lot smoother over the cities, until the slowdown started occurring due to whatever is causing the error messages.

I tried flying over Chernogorsk about 30-40 times with F35B/A10 and everything ran smoothly as it always does my end, not better not worse!

[MJK-Ranger 0]

Hi.

+ no more texture bug.

+ texture load little bit faster

- lost overall 5 - 7 FPS everywhere

- can't use FRAPS anymore. Totally laggy record down to 1 FPS. Never ever had this problem before with older beta and stable version

[AnimalMother92 10]

Ground texture bug is gone (thank you). Overall I seem to have lost some fps though, will test more.

[rundll.exe 12]

I just ended a long MP coop, and there were no problems. The performance seems to be equal in FPS, but the feeling is much smoother when quickly looking around.

Didnt do any measurements, but it seems to have improved generally

[HitmanFF 6]

Still having trouble with hang ups. Quite reproducible (CTF map when flag is being taken, ArmA2 freezes consistently). I will try to come up with a addon free map that can reproduce the freezes.

[MavericK96 0]

I tried flying over Chernogorsk about 30-40 times with F35B/A10 and everything ran smoothly as it always does my end, not better not worse!

What are your specs/settings? I've been having the Default Video Memory crash issue for a long time now, and so have some others it seems.

[metalcraze 290]

Need to do more thorough testing but it seems this patch is an improvement.

The issue with the disappearing ground seems to be fixed, turning head seems to be smoother and flying on a plane through Chernogorsk-Elektrozavodsk is smoother too now. Whereas previously I had stuttering there that was freezing the game for 5-6 seconds now it is either non-existent or takes 1-2 seconds (even though with a quickly appearing/disappearing "post initialization processing" screen) - which is a noticeable improvement. The HDD thrashing still takes place but doesn't seem to affect the game (RAM used by the game never goes higher than 1 GB though which is strange)

BFG GTX260-216 OC'ed to GTX280 level, 196.21 drivers, Win7 32 bit. Phenom II X4 3.1 GHz, 2 GBs RAM

[Alex72 1]

Tested it a bit on Win7 64 and it felt pretty good. No issues so far but i cant really compare since im on a brand new system now. Testing goes on, but feels good so far.

Click sig for specs.

Edited April 27, 2010 by Alex72