Jump to content

4 Fps Issue After Multiple Plays

v1.0.134

94 replies to this topic

#81 Buzz313th

    Member

  • PipPip
  • 36 posts

Posted 01 November 2012 - 07:26 AM

Ok, an update on my side to confirm what Chozo has said about video memory being saturated and not released between matches...

Using GPUZ to monitor GPU ram here are my results all the way to the 4fps bug.

1st match : In Mechlab before launch, GPU Ram usage is a steady 32mb... Durring match, GPU usage is ~ 750mb

2nd match : In Mechlab before launch, GPU Ram usage is a steady 350mb... Durring match, GPU usage is ~ 1000mb

3rd match : In Mechlab before launch, GPU Ram usage is a steady 550mb... Durring match, GPU usage is Saturated at 1024

4th match : In Mechlab before launch, GPU Ram usage is a steady 750mb... Durring match, GPU usage is Saturated at 1024 (This where graphical glitches start and the "Isolated 4fps" bug pops up ingame)

5th match : In Mechlab before launch, GPU Ram usage is a steady 850mb...

Restarted client...

1st match : In Mechlab before launch, GPU Ram usage is a steady 32mb

If I lower Ingame graphics options, the GPU ram buildup is slower. If I raise Ingame Graphics options, GPU ram buildup is faster.

The writing is on the wall

For those of you with graphical glitches after multiple matches... test for yourselves.. GPUZ can be downloaded for free here.. http://www.techpowerup.com/gpuz/

Edited by Buzz313th, 01 November 2012 - 07:28 AM.


#82 Cochise

    Member

  • PipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 642 posts
  • LocationAustin, Texas

Posted 01 November 2012 - 08:04 AM

View PostBuzz313th, on 29 October 2012 - 05:08 PM, said:

This has been happening to me over the last week.

After 2- 4 matches without restarting the client, if I pan my view towerds the center of the map, I will have my FPS drop from roughly 30-40 FPS smooth to an unplayable slide show anywhere from 2-4 FPS. If I pan away from the center of the map, FPS goes back up to an ultra smooth 30-40 FPS.

My solution is to restart the client and everything will be fine for a few matches.

Heres my specs...

I5 2600k at 3.3
8 gigs of ram
Nvidea GTX 560TI at stock speeds
Win xp pro 64 bit
Latest vid drivers

I have tried numerous things to fix the issue, including changing driver settings in the Nvidea control panel, deleting the shader cache, and a complete reinstall of MWO. Nuthing has helped.


This IMHO is a complete gamestopper for me.. I have searched the forums and have not found anything where a developer has chimed in.

Could a developer please reply to this post and let myself and others that might be having this issue that they are aware of it and looking into a fix please.

Thanks


That's interesting, I just restart the client every four matches. Been doing this since the beginning. I always thought it was because I only had 4 gigs of ram but you have 8 gigs.

I also use the EVGA GTX 560Ti, I have two of them setup as SLI. The SLI doesn't work yet but I have never had probs with the game not recognizing my cards. Interesting.

View PostBuzz313th, on 01 November 2012 - 07:26 AM, said:

Ok, an update on my side to confirm what Chozo has said about video memory being saturated and not released between matches...

Using GPUZ to monitor GPU ram here are my results all the way to the 4fps bug.

1st match : In Mechlab before launch, GPU Ram usage is a steady 32mb... Durring match, GPU usage is ~ 750mb

2nd match : In Mechlab before launch, GPU Ram usage is a steady 350mb... Durring match, GPU usage is ~ 1000mb

3rd match : In Mechlab before launch, GPU Ram usage is a steady 550mb... Durring match, GPU usage is Saturated at 1024

4th match : In Mechlab before launch, GPU Ram usage is a steady 750mb... Durring match, GPU usage is Saturated at 1024 (This where graphical glitches start and the "Isolated 4fps" bug pops up ingame)

5th match : In Mechlab before launch, GPU Ram usage is a steady 850mb...

Restarted client...

1st match : In Mechlab before launch, GPU Ram usage is a steady 32mb

If I lower Ingame graphics options, the GPU ram buildup is slower. If I raise Ingame Graphics options, GPU ram buildup is faster.

The writing is on the wall

For those of you with graphical glitches after multiple matches... test for yourselves.. GPUZ can be downloaded for free here.. http://www.techpowerup.com/gpuz/



Seems like the shaders are just sitting in memory instead of being dumped after each match or something like that.

#83 GL Chozo4

    Member

  • PipPipPip
  • Veteran Founder
  • Veteran Founder
  • 87 posts

Posted 01 November 2012 - 02:26 PM

Just to let you all know whom think they aren't aware... PGI IS aware of the issue and is apparently working on it. Got a reply from my support ticket regarding it and here was the following reply:

Quote

Hello GL Chozo4,
The problem you have encountered is a known issue and we hope to have it fixed soon.
This ticket is now considered to be closed, however this does not mean that we consider the issue resolved as bugs are tracked in a separate database. You can reply to this mail if you have pertinent information to add and the ticket will be reopened. Thank you for contacting the MechWarrior® Online™ support.
Regards,
<Redacted>
GameMaster
MechWarrior® Online™


EDIT: Seems this thread got moved and renamed so.. the old notifications are not sticking to this new one. Liking my own post in hopes it resets the notification for this thread.

Edited by GL Chozo4, 01 November 2012 - 05:05 PM.


#84 Rainlyte

    Member

  • PipPip
  • 39 posts

Posted 01 November 2012 - 03:37 PM

I've gotten this bug 3 or 4 times yesterday. It seems very random unfortunately... as in it's hard to reproduce and there's no error message to help either. You're playing games and everything is fine, with a stable 30 or so fps, and then suddenly one drop the FPS drops to a slideshow. For me it seems to be when facing enemy mechs, even if they're on the other side of the map. This of course makes it even more infuriating because when you get someone in visual it starts lagging. But other times it seems it doesn't matter where you look.

Other times yet it seems to start stripping models from the map. I had this once on river city and the buildings disappeared on top of the 2 or so FPS I was getting. Fairly certain it's not a graphic card overheat problem because I have a monitor for that. Also it doesn't happen consistently enough. Sometimes it's after 3-4 games, sometimes it's after 20.

#85 Stealth Monkey

    Rookie

  • 1 posts

Posted 02 November 2012 - 01:57 AM

Hi guys,

New player here, I've been playing for about 2 hours and appeared to have this bug. Really started to screw around the river city map, but the night time map was also doing it to a lesser extent.

Windows 7 32 bit
AMD 6950
4gb RAM

Not particularly interested in fix methods, I'm simply leaving for 3 months to give you time to get your stuff together. Just thought some additional info might help, and to let you know that this bug is costing you players.

Edited by Stealth Monkey, 02 November 2012 - 01:58 AM.


#86 John MatriX82

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 2,398 posts
  • LocationItaly

Posted 02 November 2012 - 02:22 AM

this bug is annoying but as much as I could monitor it, i've not seen the behaviour posted by Buzz

Using GPUZ and with task manager opened I haven't noticed nothing out of the ordinary, ram is most of the times the same and gpu vram usage seems to be stable as the previous matches.

i5 3570k
z77a g45 (msi)
2x4gh geil 1600
GTX 560 (not ti/no se)
Win7 x64 Pro

Alt entering or running around zoomed seems to soften the issue for a while so to me it's not related to hardware resources but a problem within the game engine itself, there's something broken in the shader or how the engine fills gpu pipelines.

A fix? I tend to reach this bug after around an hour of play, so consider to restart the client at some point to avoid it. Other than that i do really hope they fix this asap, because scoring kills with 4fps and stutter is kinda hard..

Edited by John MatriX82, 02 November 2012 - 02:23 AM.


#87 Devious Redbeard

    Rookie

  • The Resolute
  • The Resolute
  • 4 posts

Posted 02 November 2012 - 03:18 AM

This is not just an annoyance; it has practical implications as well. Among other things, it amounts to financial punishment. When I consider the cbills and mechs lost with the last patch and the cost of mech repairs/rearmament that have been implemented along with it, I find it particularly frustrating when I end up losing a mech in combat due to the 4 fps bug leaving me essentially defenseless and knowing I’m going to have to foot the bill to replace the gear when that mech get destroyed.

Yes I realize we’re still in Beta, but I hope an “official” resolution is in the works because nothing helps reduce enthusiasm for the game then the combo I mentioned above.

And while I am not a Founder, I do have money invested in this game - because I want to see it work. Although to be honest, I don't think it should matter when it comes to contributing input to the effort.

Edited by Devious Redbeard, 02 November 2012 - 03:27 AM.


#88 Captain Commander

    Member

  • PipPipPip
  • Philanthropist
  • Philanthropist
  • 57 posts
  • LocationDeep South of the Upper North Americas

Posted 02 November 2012 - 09:44 AM

Add me to the 4FPS bug group as of today. First time I've had any serious performance based issues with the game, beyond CTD's. Running a upper middle class PC, i7 3.2x4, 12 gigs ram, 1.5gig GTX 580, running everything maxed. Seems it only really drops out if I look at the red target indicators/get any combat animations on the screen,

-CC

#89 Wireball

    Member

  • PipPipPip
  • FP Veteran - Beta 1
  • 59 posts

Posted 02 November 2012 - 09:26 PM

I'm running MSI Afterburner and Process Explorer, and noted down the results when this started happening in closed beta.

A few things I noticed:
  • At all times during a match, including when the game is working fine, my video card memory is maxed out at 1018-1021MB, even with all the graphic detail settings on low.
  • During the closed beta, CPU usage seemed to be reduced to one core (25%) instead of two cores (50%) when the framerate drop occurred. I'll have to check next time it occurs to see if the CPU usage still drops.
  • GPU speed is consistently at 880 core, 1760 shader, and 2100MHz memory clock throughout. No downclocking here.
  • GPU load was less than 100% much of the time. I wish I could find my original post - but if I recall, it looked like it might have been waiting for data to be swapped in and out of video memory.
  • Usually takes at least 4-5 matches for the 4.0FPS bug to show up, like many of the posters here, and it typically only occurs when looking towards the center of the map (in one game, I won despite the issue because I avoided facing the center of the map while fighting enemy mechs).
  • Restarting the client fixes the issue.
  • If I forget* to restart the client and get into another game, it suffers the same issue immediately - oftentimes worse (e.g. I can't look anywhere without it lagging).
    *-I can play, it just takes a lot of guesswork and doesn't work too well if the enemy mech is not moving in a predictable pattern ;)

Core 2 Quad Q6600 @ 3.2GHz
Nvidia GTX560 Ti 1024MB w/ 306.97 drivers (it also happened with 285.62 drivers)
4GB RAM
Windows 7 64-bit
DirectX 11 according to dxdiag, but the game always runs in DX9 mode.
1920x1200
High settings as I recall.

#90 GL Chozo4

    Member

  • PipPipPip
  • Veteran Founder
  • Veteran Founder
  • 87 posts

Posted 03 November 2012 - 02:20 AM

Thinking I've started to narrow down the actual source of the GPU-Memory Leak and can be easily replicated without entering a game. Take note of your GPU-Memory the moment the game launches...

1) Enter MechBay
2) Click any mech be it trial or otherwise...
3) Do this a few times [even the same mech works] and then watch your GPU-Memory.

You'll have taken note that depending on your settings your GPU-Memory has been consumed an additional 10.1xMB each and every time you click a mech. Keep clicking the same mech on the 2x2 square on the [home] screen or others in mechlab and you'll simply run out of Graphics Memory and cause the client to glitch out and/or crash. It's rather interesting to further note how on the [home] screen that selecting mechs will further trigger this since they are not being displayed unless you go into the MechLab.

Point I'm noting: The source of the problem is most likely the BattleMech models not being released from memory. This would coincide from other tests of a leak of roughly 170.6xmb every match meaning 10.1xMB of GPU memory for each of 16 Battlemechs per match which are not being released from memory. You can further see this as well while waiting for players to join in game with the same increase per battlemech.

First Marker is upon first launch. The second line is the result of simply clicking the same mech over and over in the [HOME] tab before glitching out and then killing the client.
Posted Image

I have carbon-copied this result to support to aid in their patching of this GPU-Memory leak in the event they are still trying to narrow down the source. While they may be aware of the issue - the reason for it may not have been clear.

EDIT: Got a reply and it has been forwarded directly to the devs... hope it helps it get in the next patch.

Quote

Hello GL Chozo4,
I will pass this on to the devs, thanks for submitting it!
Regards,
<redacted>
GameMaster
MechWarrior® Online™

Edited by GL Chozo4, 03 November 2012 - 07:59 AM.


#91 Mackensen

    Member

  • PipPipPipPipPip
  • FP Veteran - Beta 1
  • 155 posts
  • LocationSweden

Posted 03 November 2012 - 12:37 PM

I also get 4 fps or very unstable fps after a few matches. Very irritating!

i5-2500K
16 GB
Nvidia GTX 680

I normally should get at least 60 fps on Very High. I´m frustrated and tired about the fps bugs in MWO. I will not play until they fix this.

Edited by Mackensen, 03 November 2012 - 01:13 PM.


#92 Indeed

    Member

  • Pip
  • 14 posts

Posted 04 November 2012 - 05:25 AM

View PostGL Chozo4, on 03 November 2012 - 02:20 AM, said:

Thinking I've started to narrow down the actual source of the GPU-Memory Leak and can be easily replicated without entering a game. Take note of your GPU-Memory the moment the game launches...

1) Enter MechBay
2) Click any mech be it trial or otherwise...
3) Do this a few times [even the same mech works] and then watch your GPU-Memory.

You'll have taken note that depending on your settings your GPU-Memory has been consumed an additional 10.1xMB each and every time you click a mech. Keep clicking the same mech on the 2x2 square on the [home] screen or others in mechlab and you'll simply run out of Graphics Memory and cause the client to glitch out and/or crash. It's rather interesting to further note how on the [home] screen that selecting mechs will further trigger this since they are not being displayed unless you go into the MechLab.

Point I'm noting: The source of the problem is most likely the BattleMech models not being released from memory. This would coincide from other tests of a leak of roughly 170.6xmb every match meaning 10.1xMB of GPU memory for each of 16 Battlemechs per match which are not being released from memory. You can further see this as well while waiting for players to join in game with the same increase per battlemech.

First Marker is upon first launch. The second line is the result of simply clicking the same mech over and over in the [HOME] tab before glitching out and then killing the client.
Posted Image

I have carbon-copied this result to support to aid in their patching of this GPU-Memory leak in the event they are still trying to narrow down the source. While they may be aware of the issue - the reason for it may not have been clear.

EDIT: Got a reply and it has been forwarded directly to the devs... hope it helps it get in the next patch.

This is really a superb piece of investigative bug finding... ness. I don't know how to say it, other than excellent work! I can scarcely imagine how much work you probably saved the bug fix team with this. I also look forward to being able to play more than 4 games in a row in the future.

Again, really great work.

#93 Phatel

    Member

  • PipPipPipPipPipPip
  • 442 posts

Posted 04 November 2012 - 08:58 AM

View PostBuzz313th, on 29 October 2012 - 07:12 PM, said:

I'm the OP..

My specs are as above in the OP...

Since we were in Beta for the last few months, I did not think much of the unsuported graphics card message I get every time I start MWO.. And furthermore I did not worry about it since the game ran fine all the way up until the last patch, which started the 4FPS bug.

Anyway, here is a screen of the message I get everytime I launch MWO.. I never thouight to report it until I talked to someone tonight who is running the same exact card as me and he is not getting this message...



Maybe this has something to do with my 4fps bug.. But remember that the game runs perfectly fine untill roughly the 3rd game..Posted Image


My son got that on his laptop. I had to go into his nvidia settings and force the game to use his graphics card, it was attempting to use his HD and ignoring that he had a card installed.

#94 Glowhollow

    Member

  • PipPipPipPipPip
  • Moderate Giver
  • Moderate Giver
  • 126 posts

Posted 04 November 2012 - 12:09 PM

Hey Guys, i have nearly the same issue.

My Config.

AMD FX-4100 3,6 Ghz
6 GB RAM
64-Bit Windows 7
ATI Radeon HD 5750 (1GB VRAM)

The last time this error occured, was in the closed beta. But - there - i had the 4fps bug - when i am starting the game (so the game was unplayable).

The really new with this bug is - that it occurs from time to time.

I am running in a group of enemy mechs - going into fight - what happens. 4fps bug for 10-15 Seconds. After this, it running normal. It occurs randomly - sometimes 3-10 times in a game.

I hope this will be patched soon, for more information contact me.

#95 Buzz313th

    Member

  • PipPip
  • 36 posts

Posted 04 November 2012 - 01:09 PM

View PostGL Chozo4, on 03 November 2012 - 02:20 AM, said:

Thinking I've started to narrow down the actual source of the GPU-Memory Leak and can be easily replicated without entering a game. Take note of your GPU-Memory the moment the game launches...

1) Enter MechBay
2) Click any mech be it trial or otherwise...
3) Do this a few times [even the same mech works] and then watch your GPU-Memory.

You'll have taken note that depending on your settings your GPU-Memory has been consumed an additional 10.1xMB each and every time you click a mech. Keep clicking the same mech on the 2x2 square on the [home] screen or others in mechlab and you'll simply run out of Graphics Memory and cause the client to glitch out and/or crash. It's rather interesting to further note how on the [home] screen that selecting mechs will further trigger this since they are not being displayed unless you go into the MechLab.

Point I'm noting: The source of the problem is most likely the BattleMech models not being released from memory. This would coincide from other tests of a leak of roughly 170.6xmb every match meaning 10.1xMB of GPU memory for each of 16 Battlemechs per match which are not being released from memory. You can further see this as well while waiting for players to join in game with the same increase per battlemech.

First Marker is upon first launch. The second line is the result of simply clicking the same mech over and over in the [HOME] tab before glitching out and then killing the client.
Posted Image

I have carbon-copied this result to support to aid in their patching of this GPU-Memory leak in the event they are still trying to narrow down the source. While they may be aware of the issue - the reason for it may not have been clear.

EDIT: Got a reply and it has been forwarded directly to the devs... hope it helps it get in the next patch.


Fantastic work Chozo!

I believe that your attention to detail would be beneficial to the Dev team. Personally I think the devs should consider you for a higher tier beta tester.

I haven't played MWO in a bit, mostly because I have been busy with work and family, but I am definitely looking forward to the progress to be made on the next update.

Again, good job.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users