Play Buffer Overrun Data Not Proceseed Fast Enough

[DennisC]

I am getting this error periodically. I can't make it happen on command or even as often as I can the display problem I am having during playback.

I have read several old threads on this error message. I have the following set-up and hard drive test results:

Intel Core 2 Duo E6700
Asus P5W DH Deluxe
2 GB Crucial Ballistix RAM
(2) UAD cards
Running UAD and other plugins
RME FireFace 800

MTU/CPU load ~38%
SRC load ranges from 0-25% Mostly 0% and only flickers to 10-25%

Bob's HDtest Results:
Audio Drive - Read 70+K KB/sec ( New WD 500 GB 16M cache)
C Drive - Read 23+K KB/sec. (this is a 2.5" mobile drive for keeping noise down)

There doesn't seem to be any CPU or hard drive performance challenges here unless the marginal C Drive could be a problem. What do you think?

One idea I thought of was, do you think my using the integrated firewire port on the Asus board could be an issue? Like maybe SAW is trying to shove the digital output to the Fireface but the port can't process fast enough? It is only 2 channels of digital output though, so I don't know.

Thoughts?

Thanks,
Dennis

[DominicPerry]

Dennis,

What buffer sizes and number of buffers. And what drivers - ASIO?

Probably just too few or too small.

Dominic

[AudioAstronomer]

Are you sure your HD tests are in kb not mb? Otherwise I think that's the problem :D

[DennisC]

Dominic - Sorry, I meant to post that: ASIO 1024x6 buffers

Robert - exact numbers: Audio Drive: 70,151 KBytes/sec
C Drive: 24,178 KBytes/sec

I thought Bob had an old post stating the numbers shound be greater than 20K KBytes/sec but I may be wrong..... Do these seem really off?

Thanks,
Dennis

[Bob L]

If you have wireless network cards in the system, consider turning them off... their constant polling for a network may cause random buffer underuns as they can steal the cpu for quite some time.

Anything else like that running in the background can do the same thing.

Bob L

[MikeDee]

If you have wireless network cards in the system, consider turning them off... their constant polling for a network may cause random buffer underuns as they can steal the cpu for quite some time.

Anything else like that running in the background can do the same thing.

Bob L

Disable all of your network connections you are not using (no cable connected); also, if you have FireWire, disable all 1394 Net Adapters you are not utilizing. These unused live connections cause the same buffer issues, including crackles and pops, due to their constant polling for a network.

Best regards,

[trock]

what exactly is the difference between a source buffer overrun and an underrun?? and what needs to be looked at and tweaked to fix either? i used to get the underruns alot, and then LEFT click mouse to continue etc

[Pedro Itriago]

Underruns: Data not getting there in time.

Overruns: Too much data getting there to be able to process it in time.

[DominicPerry]

ASIO 1024x6 buffers


It won't be buffers then. This is huge. You should be able to get down to 64x4 or 64x6.

Have you tried PCILatency? Might need to give the FW port a bit of priority.

My disk runs at about 15,000Kb/s, so that isn't likely to be a big deal - I get 64x4 on the digiface and managed 64x6 on the FF400.

Maybe the UAD needs tweaking - others here have played around alot to get it working ok. Dunno what the fixes are tho'.

Dominic

[DennisC]

Well, I looked into the network suggestions. Even though my bios has the network ports disabled, Windows Task Manager showed the process running for network as well as in the toolbar their were icons "on" concerning network ports that I turned off when Windows was up.

The difficult thing is that the problem has been intermittent. So, after I turned off the networking ports and services, I did not have the problem but I don't know if it is really fixed.

Regarding buffer size and numbers. I have never been able to get the very low settings many of you relate. Perhaps I need to install a dedicated Firewire card and not use the integrated port. But I have plenty of horsepower and the RME FF800, drivers up to date. It hasn't been a huge problem but it has been in the back of my mind.

Dennis

[Ian Alexander]

Dennis,

Have you done the XP Tweaks from SawStudio.com and the Firewire update for XP from MS? I believe the FW update affects FW800 but not 400.

How many processes do you see in Task Manager? Do you know what they are? For example, my office machine has 51 running right now, but my audio machine has only 26. (And I confess I don't know what they all are, but I'm not having any problems.:o )

HTH.

[DennisC]

Ian,

Yes, I have done the XP tweaks. No firewire update as I am running XP SP1.

I am away from my audio PC, but the number of tasks is probably closer to your audio PC versus in the 50s.

I did track for 3 hours last night with no buffer overrun errors. I am getting some pops etc. which are not in the audio. When I back up and replay they do not happen again. And I only had one instance of the playback-display problem with the Full mixer. Still I wish it was "cleaner". there are no show stoppers but it is a little discouraging, especially given my task to convert my Nuendo partner.

I probably just need to do the whole re-build as Perry suggested. Although, given my inexperience with system builds, I don't hold a lot of hope that I, personally, can do it "right" from scratch.

We'll see this summer.

Thanks,
Dennis

[DominicPerry]

Dennis, send me a list of the services running and I'll have a look to see what might be wrong.

Dominic

[DennisC]

Dominic,

Thank you. Here's a list:

taskmgr.exe Dennis
mgabg.exe Dennis
msmsgs.exe Dennis
pdesk.exe Dennis
qttask.exe Dennis
firefacemix.exe Dennis
fireface.exe Dennis
explorer.exe Dennis
EzBackup.exe System
Spoolsv.exe System
svchost.exe Local Service
svchost.exe Network Service
svchost.exe System
svchost.exe System
lsass.exe System
services.exe System
winlogon.exe System
csrss.exe System
smss.exe System
System System
System Idle Process System

Any you can explain that I need or don't need would be great. Especially if there are any which you think might cause some of my symptoms; PLay buffer overrun and/or the full mixer display problems on playback.

Thank you,

Dennis

[DominicPerry]

Dennis,

You might want to try stopping and disabling

qttask.exe
and
EzBackup.exe System

All the rest are benign as far as I can tell (sometimes a bit difficult to be sure with svchost processes.

You should try the PCILatency tool too.

Dominic

[Pedro Itriago]

Dominic,

Thank you. Here's a list:

taskmgr.exe Dennis
mgabg.exe Dennis
msmsgs.exe Dennis
pdesk.exe Dennis
qttask.exe Dennis
firefacemix.exe Dennis
fireface.exe Dennis
explorer.exe Dennis
EzBackup.exe System
Spoolsv.exe System
svchost.exe Local Service
svchost.exe Network Service
svchost.exe System
svchost.exe System
lsass.exe System
services.exe System
winlogon.exe System
csrss.exe System
smss.exe System
System System
System Idle Process System

Any you can explain that I need or don't need would be great. Especially if there are any which you think might cause some of my symptoms; PLay buffer overrun and/or the full mixer display problems on playback.

Thank you,

Dennis

Let me butt in.

First, you have messenger starting up with your system, not too good (msmsgs.exe)

Then, I don't know how good having a back-up program running in the background could be, but personally, I would only run such a thing when needed (EzBackup.exe)

Then, you have the ever present quicktime virus task running at start-up (qttask.exe)

Those three could be disabled.

Other program that could be disabled, although very unlikely to cause problems:

You have the matrox bios guard running. Here's some info I found about it.

taken form http://www.liutilities.com/products/wintaskspro/processlibrary/mgabg/


Process File: mgabg.exe or mgabg
Process Name: Matrox BIOS Guard

Description:
mgabg.exe is a process associated with the efficient running of Matrox VGA cards. This program is a non-essential system process, but should not be terminated unless suspected to be causing problems.

The ohter one is matrox power desk (pdesk.exe). If you're not using any desktop workspace administration or utilization, I'd zap that one too!

But it was not a question directed to me, so wait until the person who asked you about the tasks helps you

[DennisC]

Dominic,

I ran the PCILatency Tool and everything looks fine. Everything was set to "0" exept the two UAD cards ("64") and the firewire driver "96". The Matrox card is PCIe so it is not a factor in the PCILatency Tool.

Pedro,

Yes, I stopped the messenger and the quicktime. FYI, I am running dual monitor with the Matrox and the pdesk was needed. I found that out the hard way....

Dennis

[DominicPerry]

Dennis,

Have you tried disabling the UADs in Device Manager or removing them completely (physically) and seeing if you can get the buffer size and numers down? Can't see any reason you shouldn't get a decent performance from that machine.

Dominic

[Pedro Itriago]

... FYI, I am running dual monitor with the Matrox and the pdesk was needed. I found that out the hard way....

Dennis

The best way to find something out ;)

Keep an eye on that quicktime task. There's a reason why I labeled it as virus, it keeps poping-up uninvited.

[Perry]

Ian,

Yes, I have done the XP tweaks. No firewire update as I am running XP SP1.

I am away from my audio PC, but the number of tasks is probably closer to your audio PC versus in the 50s.

I did track for 3 hours last night with no buffer overrun errors. I am getting some pops etc. which are not in the audio. When I back up and replay they do not happen again. And I only had one instance of the playback-display problem with the Full mixer. Still I wish it was "cleaner". there are no show stoppers but it is a little discouraging, especially given my task to convert my Nuendo partner.

I probably just need to do the whole re-build as Perry suggested. Although, given my inexperience with system builds, I don't hold a lot of hope that I, personally, can do it "right" from scratch.

We'll see this summer.

Thanks,
Dennis

Hi Dennis,

Sorry to hear you're still having problems. You're getting some good tips here.. any improvement yet?

All the best,

Perry

[MikeDee]

Dominic,

I ran the PCILatency Tool and everything looks fine. Everything was set to "0" exept the two UAD cards ("64") and the firewire driver "96". The Matrox card is PCIe so it is not a factor in the PCILatency Tool.

Pedro,

Yes, I stopped the messenger and the quicktime. FYI, I am running dual monitor with the Matrox and the pdesk was needed. I found that out the hard way....

Dennis

Dennis, you might be configuring the PCI Latency Tool as I did when I first began using it...it appears you are giving your highest priority devices either a value of 0 [zero] or a very low value, such as 64. This is incorrect.

In fact, it is the other way around. Give your highest priority devices -- i.e., your UAD cards -- a value of 255. A value of, say, 192 (greater than half, yet, a multiple of 16) might work for your FireWire driver.

All: If I am mistaken in any way, no matter how slight, please feel free to correct. This way I will learn, too...thanks.

HTH,

[AcousticGlue]

Everyone, for processes running, the best tool to download is ProcessExplorer from Microsoft (used to be Systernals). Then you can see what svchost processes that are srvany processes are. It only runs when you run it. Great tool.

[Ian Alexander]

Everyone, for processes running, the best tool to download is ProcessExplorer from Microsoft (used to be Systernals). Then you can see what svchost processes that are srvany processes are. It only runs when you run it. Great tool.
Thanks for the tip. I just went to download ProcessExplorer and was surprised to see a picture of a shiny red apple on the ms home page.

[DennisC]

Mike,

Thanks for the input. There seems to be different takes on the PCILatency Tool settings. By the way, I did not change any values. My understanding of "0" values is that those particular devices can not have their settings changed. The settings I reported were the settings the tool detected.

Regarding high or low values, I have read that lower values provide a smoother peformance for both video and audio. The reasoning being that no one device "hogs" the PCI bus. So, giving the UAD cards a high value might be good for them but it might make the display problem worse. Although it is unclear how my PCIe video card would be effected, if at all, given that the PCI latencies do not apply to PCIe devices. That is why when I saw the values I hesitated to change any. I just can't make either problem happen often enough to reliably test setting changes.

Any thoughts to this are appreciated.

Thanks,
Dennis

[Bob L]

Dennis... Matrox card in dual mode... is it in Clone mode... if so... try shutting that down... I reported a situation like this a while back where for 2 years the lowest buffer size usable on a very powerful machine was 512.... below that would stutter.... discovered the clone mode interfering with the mouse movements and burning up the cpu as a result... turned clone mode OFF... and 2 x 64 became the noemal operating capacity.

Bob L

[DennisC]

Bob,
The Matrox is not in Clone mode. Good news is that I was able to get buffers down to 64x4 with no glitching at all (when I turned off all of the plugins but still had 35 channels using EQ on many, SAW Comp and Gate on several).

BUT - I can regularly make the display problem happen on playback. Large or small buffers don't effect it. I guess it's good that I can duplicate it now pretty much on demand. However, I don't have any clue how to solve it.

Further frustrating is that all my trials etc. now have my PC taking several minutes just to boot-up. Don't say it Perry...:) I know it looks like a reload but without a known problem I don't have much confidence in that either.

Dennis

[DennisC]

Just to clarify and summarize:

I am now experiencing the "display problem on playback" with regularity. I can make it happen pretty much at will within a minute or so during playback.

I have all FX turned off by clicking of the little "chord" plug in the Wide view of each channel that has FX plugins patched.

I am running ~34 tracks, many with EQ, some with Comp. and Gate. I am using an additional 6-8 output channels for submixing. Only 11-14% CPU usage.

Windows XP Pro SP1
Intel Core 2 Duo E6700
2 GB Crucial Ballistix PC8500
Matrox PCIe G550
UAD PCIe
UAD PCI
RME FF800
Firewire off of the Asus P5W DH Deluxe motherboard
XP Tweaks done
All networking turned off
Not running dual monitors in clone mode
Wireless keyboard and mouse (Logitech)
SAWStudio Full v4.0c

Maybe it is related to my Matrox card being a PCIe not a PCI card?

I am beginning to consider replacing the video card. Perry, you have the same MB and processor, what video card are you using for dual monitors?

Other thoughts?

Thanks,
Dennis

[Naturally Digital]

I am now experiencing the "display problem on playback" with regularity. I can make it happen pretty much at will within a minute or so during playback.Dennis, what exactly do you mean by 'display problem on playback'? Am I missing something?

[DennisC]

David,

It has been a problem experienced by others in the past but Bob said he fixed with the v4 release. I believe it was with people using dual monitors and or dual processors. And it looks like this:

When using dual monitors and during playback, while changing Fkey views the Full Mixer window locks up and the content of the Full Mixer window fails to refresh properly, although playback will continue fine from an audio perspective.

Prior to the v4 release of SAWStudio, a workaround was to force the use of a single processor. This workaround does not work for me currently.

Dennis

[Naturally Digital]

When using dual monitors and during playback, while changing Fkey views the Full Mixer window locks up and the content of the Full Mixer window fails to refresh properly, although playback will continue fine from an audio perspective. OK. For me it happens more often with the MT although it's definitely less often these days.

Have you tried reducing the hardware acceleration features of the video card... Turning off bitmap caching and reducing the troubleshooting slider etc.? I don't think you should need to get a new video card but I will say I see this problem less with my nvidia cards than my Matrox G450. FWIW, if you're just looking for a good 2D card I have to recommend the PNY NVS285 (quadroNVS) cards. Doesn't have TV out or any fancy stuff but drives two monitors quite nicely.

[Bob L]

Definitely take the video acceleration down at least one notch... also... try removing (not just bypassing) plugins like DX and VST and UAD stuff... they will still allocate threads when patched even if bypassed and the multi-threads going off in all directions could be causing the issues.

Bob L

[Naturally Digital]

OK. For me it happens more often with the MT although it's definitely less often these days.Actually, I just realized one thing... I've been running in dualview mode (two independent monitors instead of one spanned desktop) for the last 6-8 months. I can say I'm seeing the video lockups far less, if at all these days.

[Bob L]

Since 4.0 I can no longer duploicate the video issues on my duo-core laptop either.

Bob L

[Perry]

........

Maybe it is related to my Matrox card being a PCIe not a PCI card?

I am beginning to consider replacing the video card. Perry, you have the same MB and processor, what video card are you using for dual monitors?

Other thoughts?

Thanks,
Dennis

Dennis I have an ATI card... Asus EAX 1950 Pro. Just to add.. with this I only install the drivers.. not the 'desktop utility'. I still have all the functionality I need though for setting up dual monitors. And I do run this as 2 separate monitors; they're different sizes with different native resolutions. One is a 23 inch widescreen @ 1920 X 1200 and the other is a 19 inch widescreen @ 1440X900. I definitely don't have this problem that is being described.

I've also run a few different nVidia cards with no apparent problems. I do like this ATI card though. And I like the way Asus designed the fan on this one... very quiet.

Perry