PDA

View Full Version : Sounds like a popcorn machine! Please Help



Warren
11-24-2008, 07:39 PM
Hi Bob:

I was wondering with all the Output buffer dropping issues I have, Is it possible to use Microsoft Process Monitor somehow to see what processes might be stepping on the RME HDSP9652 audio card? If so how would you set it up to monitor for this kind of issue.

SAC is running with 24 channels (all dev 1) active on FOH, and 6 monitors pre-fader from FOH no Native or VST effects at all and no EQ, COMP or GATES active it will be stable at 50***37; then after 5 minutes the output buffers will quickly start to runaway and the CPU usage ramps up to100% a GTR playing through it sounds like a popping flanger and a force shutdown is required.

(1) I have used DPC Latency Checker and it***8217;s rock solid at 5us no spikes at all.
(2) I setup the Matrox P750 with the latest BIOS and drivers.
(3) PCI Latency forces HDSP to highest priority
(4) SAC runs at Realtime priority. And it***8217;s the only process at realtime.
(5) No change in either SINGLE CPU or DUAL CPU.
(6) All your suggested tweaks have been applied.
(7) Minimum services applied, only enough to network. (No change if NIC enabled or not)
(8) No IRQ***8217;s or DMA***8217;s are shared with AUDIO card.
(9) I have checked for viruses and even rebuilt entire OS no change
(10) Buffer settings offer very little change.

IWILL DH800 Dual 3.2 Gig Xeon 880 Mhz FSB latest drivers and bios
4 Gig ram DDR400 PC3200
Matrox P750 triple head video card, Latest drivers and bios
2 SATA drives
2 IDE drives

Thanks Bob and all others for any help in this matter, which has plagued me with this machine since it was built.

Chris

Leadfoot
11-24-2008, 08:18 PM
Only thing I can think of is, does it have an older VIA chipset? I seem to remember some incompatibilities with that. Just a thought, not much left to go on. Some motherboards just don't play nice with pro audio. Can you try the 9652 in another box to rule out the card?

Tony

Warren
11-24-2008, 08:36 PM
Only thing I can think of is, does it have an older VIA chipset? I seem to remember some incompatibilities with that. Just a thought, not much left to go on. Some motherboards just don't play nice with pro audio. Can you try the 9652 in another box to rule out the card?

Tony


875P chipset is on the DH800.
HDSP works in other slower machines and is perfect in TOTAL MIX on this one

Bob L
11-24-2008, 08:50 PM
The runaway buffers sure sounds like a floating point plugin issue... are you certain you did not setup a default mix template at one time that has a plugin buried somewhere on some channel?

You might try a complete fresh SAC install in a new test folder and try the test there and see if anything changes.

Bob L

Naturally Digital
11-24-2008, 10:34 PM
it will be stable at 50***37; then after 5 minutes the output buffers will quickly start to runaway and the CPU usage ramps up to100% Chris, I assume this is 50% cpu reading in SAC, not task manager...

1) If you run this repeatedly, is it always *exactly* the same amount of time before things go south?

2) You say it's stable for 5min... Right after starting up (since you're only at 50% cpu) are you able to add another 3 or 4 monitor mixes with stable performance (even if only for a brief period)?

Bob L
11-25-2008, 12:39 AM
What about clocking... Is the RME set as Master clock... or in AutoSync?

If its not the master... where is the clock coming from?

Could the clock be set at a different rate than the SAC rate? Make sure they are the same... because SAC can not control it if the RME is not the master.

Try setting it to be the master and your converters to slave to the optical signal from the RME... see if that helps.

Bob L

Warren
11-25-2008, 12:59 AM
What about clocking... Is the RME set as Master clock... or in AutoSync?

If its not the master... where is the clock coming from?

Could the clock be set at a different rate than the SAC rate? Make sure they are the same... because SAC can not control it if the RME is not the master.

Try setting it to be the master and your converters to slave to the optical signal from the RME... see if that helps.

Bob L

Bob:
This happens even on a new install of SAC and no effects.
I will check into the clock issue The RME is not the master in my setup because I use two Tascam MX2424's for converters ADDA, Matthis (RME) stated that the MX2424's have better clock than the HDSP. But I am willing to try anything, What SAC wants SAC gets if I can get this thing to stop this motorboating runaway buffer trash.
No mention of process monitor? am I off base here?
By the way the hdsp config software states that ALL devices are SYNC, but from what your stating SAC need the aduio card to be master, I will give that a try.

Dave:
Yes, its in SAC that I am getting the 50***37;.
5 min is an approx. time not an exact time.
Once at 50%, and running, most any window change, effect added or just waiting a few moments, it will eventually go south. SAC CPU% will ramp up till crash of SAC not XP.

Thank you all for the help tracking this down, its been a pain in the bum for awhile.

Chris

Bob L
11-25-2008, 07:51 AM
You did not mention what your rate settings were... is the Tascam clock at 48k and SAC at 44.1k?

That could be the problem.

Bob L

DominicPerry
11-25-2008, 10:19 AM
I was wondering with all the Output buffer dropping issues I have, Is it possible to use Microsoft Process Monitor somehow to see what processes might be stepping on the RME HDSP9652 audio card? If so how would you set it up to monitor for this kind of issue.



Chris, it might be worth using normal task manager. Open it before you start using SAW but minimise it. Perhaps change the update speed to slow (so it doesn't use too much CPU). Add some columns in the Processes tab, using View -> Select columns -> I always add CPU usage, CPU time, Memory Usage, Peak Memory Usage. That way you can see the CPU in use but also which processes have been greedy with CPU time since you started the box - this is important, it's best to reboot before you test so that the results show only the time you are using SAW/SAC, not the time since you started the box including surfing, doing your accounts, whatever.

Hope this helps.

Dominic

Warren
11-25-2008, 11:05 AM
Chris, it might be worth using normal task manager. Open it before you start using SAW but minimise it. Perhaps change the update speed to slow (so it doesn't use too much CPU). Add some columns in the Processes tab, using View -> Select columns -> I always add CPU usage, CPU time, Memory Usage, Peak Memory Usage. That way you can see the CPU in use but also which processes have been greedy with CPU time since you started the box - this is important, it's best to reboot before you test so that the results show only the time you are using SAW/SAC, not the time since you started the box including surfing, doing your accounts, whatever.

Hope this helps.

Dominic

Hi Dominic:

I am sorry if I stated somewhere that it was SAW its SAC, (In SAW I get buffer underruns DATA not recieved fast enough or something like that) I have used TM and SAC is the only resource hog at the time,(not that I have issues with SAC its the one using the highest load) everything else is steady. Changing the refresh rate of TM actually adds a bit to the sputtering, the faster TM refreshes, the faster the sputter ,but stoping the refreshes does nothing to stop the buffer runaways.

I have added these columns as you stated but have not seen anything that appeared concerning, but I can do that again. This is why I had asked about microsoft process monitor because it appears as though it might dive deeper into what going on within the processes and even hidden ones.

Thanks for the ideas.

Chris

Warren
11-25-2008, 11:11 AM
You did not mention what your rate settings were... is the Tascam clock at 48k and SAC at 44.1k?

That could be the problem.

Bob L

Actually it is set on all machines to 44.1k 16bit, I was trying to see if the lowest load would help resolve the problem, no change. later this morning I will change the master to the RME as you suggest.

Fingers crossed, even though crossing never seemed to help.:)

Thanks
Chris

Harold Thompson
11-25-2008, 12:27 PM
Chris

You might want to try and substitute another video card. I had to ditch my P750 for a cheaper video card because of the same sort of problem. If have another card or can borrow one to check it out it could possibly be it. I have used the Matrox cards for years without any problems but the P750 definitely caused a lot of dropped buffers.

Harold Thompson

Warren
11-25-2008, 12:57 PM
Update on this issue:

Bob:
I have moved the Master clock over to be the RME HDSP and now I get a crackle sound when moving the fader on TOTALMIX, (Which is not such a big deal since I don't use it for anything)but it sounds fine when its not moved, at least for now SAC appears to be output buffer dropping free, I have only been able to test for about 15 min, but I have added effects, changed screen resolution, and a few other thing that otherwise would have crashed SAC on my system for sure.

I will continue to test to be sure this is not a one time thing.

Thanks for the help

Chris

Warren
11-25-2008, 01:03 PM
Chris

You might want to try and substitute another video card. I had to ditch my P750 for a cheaper video card because of the same sort of problem. If have another card or can borrow one to check it out it could possibly be it. I have used the Matrox cards for years without any problems but the P750 definitely caused a lot of dropped buffers.

Harold Thompson

Hi Harold:

My P750 video issue was solved due to allowing the audio card to have a higher priorty than the video card.

Chris

Warren
11-25-2008, 03:44 PM
NEXT UPDATE:

The Word clock master on audio card (RME HDSP9652) did positively affect the system, but no cookie. It still after about 8 min. will slowly drop output buffers then over about 4 more min. runaway to 100% (SAC).

DCP LATENCY CHECKER shows 5us solid all the time even when dropping buffers.

Still fighting this issue with lots of hope because when it is running it just works and sounds so sweet.

Chris

DominicPerry
11-25-2008, 03:52 PM
Chris, I'd still be interested to see the task manage process window after you've run your test to the point where it all goes horribly wrong. Can you do a screen print?

Dominic

Bob L
11-25-2008, 04:10 PM
Chris... is this just 1 RME card with 3 converters hooked up to it for 24 chans?

What are the converters and did you change them all to now sync optically... rather than be the master or use word-clock?

If you only have 8 chans worth of converters make sure to not assign any more than that in the audio device setup window.... SAC uses every chan it sees in the device setup list... and if you are missing converters or they are not clocked correctly, they are still involved in the live engine loop and will cause problems.

Bob L

Warren
11-25-2008, 05:22 PM
Chris... is this just 1 RME card with 3 converters hooked up to it for 24 chans?

What are the converters and did you change them all to now sync optically... rather than be the master or use word-clock?

If you only have 8 chans worth of converters make sure to not assign any more than that in the audio device setup window.... SAC uses every chan it sees in the device setup list... and if you are missing converters or they are not clocked correctly, they are still involved in the live engine loop and will cause problems.

Bob L

Hi Bob:
Just one HDSP9652 hooked up to two MX2424' one for 24 channels in and the other for 24 channels out. The RME is now the Master and is a bit more stable , Clocking is via BNC, but make no difference when ADAT sync is enabled. I get a rock solid lock either way but it still sputters.

Chris G

Warren
11-25-2008, 05:56 PM
Bob:
I don't know if this is anything but when it started to sputter I turned down SAC and turned up TOTAL MIX and it was clean, I went back to SAC turned it up and it still was sputtering.

Does that point anywhere?

Chris

Warren
11-25-2008, 07:32 PM
Update #3

OK I had a PS2 wireless keyboard and mouse setup that I removed and installed a hardwired mouse and keyboard, now I can get about 20 min of clean play before buffer drop, but buffers dropped are not audable.
Appears to be a big help as did RME as Master clock.
Seems as were getting there. SLOOOWLY but hey I'm a lot happier. :D

Dominic:

Sorry this machine is very stripped down so no paint program, but again of the processes that can be monitored via Task Manager (TM) there is no change when output buffers are dropped.

Bob L
11-25-2008, 07:32 PM
I'm still questioning the clocking... if the RME is now the master... you must have both decks now in optical sync... or slave mode... they cannot also still be the master clock... and... they must have a mode to sync to the ADAT optical, or to BNC wordclock... which you then must be feeding properly from the RME card... and you must be able to feed both decks and properly terminate the word clock signal at the last one.

Sputtering and popcorn descriptions sound more and more like a clocking issue to me.

Bob L

IraSeigel
11-25-2008, 07:45 PM
I'm still questioning the clocking... if the RME is now the master... you must have both decks now in optical sync... or slave mode... they cannot also still be the master clock... and... they must have a mode to sync to the ADAT optical, or to BNC wordclock... which you then must be feeding properly from the RME card... and you must be able to feed both decks and properly terminate the word clock signal at the last one.

Sputtering and popcorn descriptions sound more and more like a clocking issue to me.

Bob L

Chris, is it possible for you to take a screenshot of your RME card settings and post it here? (For example, the Firefaces have a "Fireface Settings" dialog box in which you choose your clock source and turn on/off phantom, choose gain in/out, etc. Does your HDSP card have something similar?)

Ira

Warren
11-25-2008, 08:56 PM
I'm still questioning the clocking... if the RME is now the master... you must have both decks now in optical sync... or slave mode... they cannot also still be the master clock... and... they must have a mode to sync to the ADAT optical, or to BNC wordclock... which you then must be feeding properly from the RME card... and you must be able to feed both decks and properly terminate the word clock signal at the last one.

Sputtering and popcorn descriptions sound more and more like a clocking issue to me.

Bob L

On the MX2424, you set it up for external sync when its a SLAVE either WORD CLOCK or ADAT SYNC and INTERNAL when it used AS the MASTER WORB CLOCK or ADAT SYNC.

If you think it might help I will disconnect any clock cables when ADAT SYNC is engaged.

Ira.:
I don't have paint program on this unit as it is stripped down. The HDSP is only ADAT and WORD CLOCK w/SPDIF so its no as involved as your setup is and I have a solid sync on all three ADAT units and SPDIF when it is in use.

Please don't give up on me though, we are making progress, and thank you all for that.:)

Chris

IraSeigel
11-25-2008, 10:04 PM
On the MX2424, you set it up for external sync when its a SLAVE either WORD CLOCK or ADAT SYNC and INTERNAL when it used AS the MASTER WORB CLOCK or ADAT SYNC.

If you think it might help I will disconnect any clock cables when ADAT SYNC is engaged....

Chris

Chris, as in most troubleshooting scenarios, I would strip down the setup to the bare minimum and get rid of all extra connections. Start with SAW and RME only, no other cables. Go from there.

Also, I believe Matthias or Daniel has said in the past that ADAT should NOT be used to provide sync; always go with word clock. I might have misunderstood your post above.

It sounds like you're describing a clocking issue, which I've experienced in the past, as well.
Ira

Warren
11-25-2008, 11:41 PM
Chris, as in most troubleshooting scenarios, I would strip down the setup to the bare minimum and get rid of all extra connections. Start with SAW and RME only, no other cables. Go from there.

Also, I believe Matthias or Daniel has said in the past that ADAT should NOT be used to provide sync; always go with word clock. I might have misunderstood your post above.

It sounds like you're describing a clocking issue, which I've experienced in the past, as well.
Ira

Hi Ira:

Thanks for trying to help here, I have indeed rebuilt the &^%*& a few times at least. Yes we are seeing Word Clock issues here for sure and I have read what you have by Daniel and Matthias. I think that Bob is mainly stating that the audio card should be the Master and that sync be either by ADAT optical sync or BNC connection WORD CLOCK, not both if I understand him correctly.

At any rate since I went to having the audio card be the master, it has been much better.

Were getting there.

Chris

IraSeigel
11-26-2008, 12:17 AM
Hi Ira:

... that sync be either by ADAT optical sync or BNC connection WORD CLOCK, not both if I understand him correctly.

...
Chris

As RME stated, sync should NOT be supplied by ADAT. Definitely use the BNC if it's available.

Nitey nite.

Ira

Bob L
11-26-2008, 12:36 AM
Sorry Ira... but I believe you may be misled here... Every stable system I have in the field and here in my production studio is working by using the RME as master and simply using the converter clocked to the optical pipe... not ADAT sync... but the ADAT Optical connection... the RME is supplying the clock down the optical line and the converter uses that with no problems whatsoever.

Perhaps word clock can provide a more stable overall performance... but it can also generally be much more finicky... and if things are not properly terminated, can cause problems as Chris is describing.

I am suggesting he try something simple and see if it corrects the problem.

Bob L

Warren
11-26-2008, 02:07 AM
Sorry Ira... but I believe you may be misled here... Every stable system I have in the field and here in my production studio is working by using the RME as master and simply using the converter clocked to the optical pipe... not ADAT sync... but the ADAT Optical connection... the RME is supplying the clock down the optical line and the converter uses that with no problems whatsoever.

Perhaps word clock can provide a more stable overall performance... but it can also generally be much more finicky... and if things are not properly terminated, can cause problems as Chris is describing.

I am suggesting he try something simple and see if it corrects the problem.

Bob L

Bob I think that the thing to do at this point is to exchange the MX2424's for a couple ADA8000's for further testing, because so many other things have been eliminated here. I have two ADA8000,s somewhere around here I will check out tomorrow.

I will keep you informed of any progress I have, I think your right sync may appear to be established here but I am afraid I don't totally trust what I see.

Also in the morning I will have a better additude. :D

Thanks to all for you help with this. GREAT SUPPORT!!!

Chris

Bob L
11-26-2008, 07:36 AM
Ok... on the ADA8000 set the sync switch in the back to ADAT IN position and connect both light-pipes (in and out). Set the RME to master.

Also... if you only have 2 ADAs, then make sure to set up your audio device setup options with only 8 stereo devices for a total of 16 chans... both in and out devices.

Bob L

Warren
11-26-2008, 06:40 PM
I hate to say this but, after connecting the ADA8000's (two each both set to slave ADAT in) to the RME HDSP9652 (Master) with Sync via ADAT Litepipe, and SAC only setup for Devices 1-16 (ADAT 1 and 2) input and output. I still after about 5 min. start to hear a few dropouts then as I check what is up it starts to runaway. So I still feel its a CLOCK issue but have yet to track it to either a unit or toslink cable.

But I am still confused as to why is sounds GREAT when monitoring a input with Totalmix only, but when SAC is running it slowly hangs.

I am now going to check each ADA8000 one by one to rule out a defective unit. because I had noticed that one is a bit noisy compared to the other.

Still hopeful I hope. :p

Thanks All!

Chris

Warren
11-26-2008, 07:15 PM
OK I just tried it with the seemingly good ADA8000 and RME only
and it sounded a sooth as hot butter on a bald monkey. then (there always seems to be one of those) a few pops and runaway gerrrrrrrrrrrrrrr :eek:

The last thing I want to see is a bad HDSP9652, how about Toslink those are cheap!

You can see by the way I am writing I am going nuts, since it appears I am writing to myself.

Any Ideas from here anyone?

Chris

Craig Allen
11-26-2008, 07:40 PM
I don't believe it's a clocking issue. If it works with Total Mix, then it would point to something on the PC. Total Mix mixes at the hardware, so it doesn't run audio through the PC. But it does go through the converters, so if it's not popping, no clocking issue. Somewhere you have something stomping on the soundcard - it may be an IRQ or PCI bus latency setting issue. I'd start by trying to disable things (like a network adapter) and see what happens.

Warren
11-26-2008, 08:56 PM
I don't believe it's a clocking issue. If it works with Total Mix, then it would point to something on the PC. Total Mix mixes at the hardware, so it doesn't run audio through the PC. But it does go through the converters, so if it's not popping, no clocking issue. Somewhere you have something stomping on the soundcard - it may be an IRQ or PCI bus latency setting issue. I'd start by trying to disable things (like a network adapter) and see what happens.

Hi Craig:

I did disable th NIC, changed Video card and forced the SAC to realtime.

Chris

Craig Allen
11-26-2008, 09:43 PM
I've read the post, but don't remember everything that has been done so far. Did you change the PCI slot the RME is in?

Bob L
11-26-2008, 09:56 PM
Well it definitely has to be something going on in the PC then... Totalmix works at the hardware level so does not depend on Windows and other driver issues.

SAC does... obviously... it is a program app fighting with all other services and apps in Windows for CPU time...

This one is a mystery... and it might be one of those things that you may never solve... so far... this machine is just not a good one to use for a SAC system.

For all the time you are spending... it may just be easier and cheaper to try another machine. :)

Bob L

Naturally Digital
11-27-2008, 01:11 AM
I did disable th NIC, changed Video card and forced the SAC to realtime.If you haven't already done so, I would suggest you get another drive and do a fresh install of windows, RME and SAC. Don't tweak anything and see how it runs. Use the 2.94 version of the RME driver. I suggest you start without the Matrox driver installed and see how it runs. Give it a go first, then install the Matrox driver, just so you can rule that out.

Before you start the install reset your BIOS to default settings but just make sure to turn off any kind of cpu throttling (demand based switching) or similar features. I suggest you leave Hyperthreading on for this test.

Again, resist the urge to tweak. Then report your results.

Naturally Digital
11-27-2008, 01:19 AM
This one is a mystery... and it might be one of those things that you may never solve... so far... this machine is just not a good one to use for a SAC system.Bob makes a good point here. I thought the 875 chipsets were ok for audio but OTOH I never heard of the 875P (dual cpu) version of this chipset. It's a bit of an unknown I fear.

You may be able to get it to work but you should also prepare yourself to swap the motherboard for another dual xeon version with a different chipset if you really need the absolute best performance.

I blew my brains out trying to get a 9652 card to work at low(ish) latencies on a 925x based machine (notoriously bad chipset for audio use) and there was absolutely nothing I could do to get it stable with small buffers. In your case you may not be that far off of best case for that mb.

DominicPerry
11-27-2008, 06:44 AM
Xeons and their associated MoBo's are both very expensive still. It would probably be cheaper, if you decide it is a MoBo problem, to just buy a Core2Duo and a cheap iP35 MoBo.
But I would rule out a faulty HDSP card first. Can you try the same (or similar) test with the inbuilt sound card and/or try the HDSP in another box?
I'm still interested to see the Task Manage output before and after the CPU ramping. Something, is swallowing CPU. Once you know what, you can try to find out why.

Dominic

IraSeigel
11-27-2008, 11:52 AM
Chris,
I've sent you an email regarding this issue to the gethelp@themasterstracks.com address.

Ira

Warren
11-29-2008, 02:01 AM
What I have done until I find what the heck is up this dually is take the RME out and put it in a Pent 4 3gig HP and remote with the dual Xeon to it. I can now get 24 in's, FOH and 6 monitor feeds out, w/ 3 verbs and echo at 32x2 90% solid for hours. The whole thing has left me a little depressed with Iwill motherboards and Xeon processors.

At least I have a rig I can use until I can get another rackmount system.

Hey! Thank you all for all the great help you have offered.

Chris