I recently uprgraded my audio interface to a Lynx L22 and my CPU to an AMD FX-8350 in an attempt to be able to record at 96KHz and a lower sample buffer size. Unfortunately, even with this current set-up I cannot lower my buffer size below 1024 at 96KHZ without terrible static/crackling. It seems to me that with this rig, I should be able to get much lower without issue. Is there something obvious that I am missing here?

I watch my cores as they are displayed in Sonar and the 1st core is really taxed while the other 7 are hovering in the 15% range. I have tried setting the affinity in task manager to not use the 1st core at all but it still shows up in Sonar and it is when this core hits in the 90% range that the problems and dropouts occur.

Any assistance would be greatly appreciated as I am really new to this. Thanks!

__________________
- Sean McLaurin

AMD FX-8350 OC'd to 4.3 GHz
12GB RAM
Windows 7 64-bit
Cakewalk Sonar X1D
Lynx L22 interface
JBL LSR2325P monitors w/MSC1
Behringer BCF2000
Korg K25 midi controller
FMR Audio RNP mic preamp
Oktavamod modded MCA SP1

bump

Do you have use multiple CPU cores checked in Options -> Audio Menu item on one of the tabs on that group of Windows? Not sure about the overclocking which I missed, if you know that stuff for sure, but perhaps there would be something here about any of that which may help: http://www.guru3d.com/ Otherwise this program to see if anything else is interfering with any of the sounds. http://www.thesycon.de/deu/latency_check.shtml

"Do you have use multiple CPU cores checked in Options -> Audio Menu item on one of the tabs on that group of Windows?" That window is in Sonar where you set up your audio card. Server is not working great here on this forum again, so second post.

I have selected "Use Multiprocessing engine" under "Playback and Recording".  Is this what you are referring to?

SM...

A couple of things.. there is a setting in the aud.ini file (you can do a search) that tells you what variable you need to set to best balance multi cores (all I remember is you should set this value to 2 for QUAD cores).

But, I have another suggestion for you..... believe it or not, this is going to sound counter intuitive but it worked for me.   here goes...

When I tried to get as much horse power to SONAR to track with lower latency, I too had tremendous glitches in the audio... however, the more I thought about it and tried to adjust every parameter in both SONAR and the OS to squeeze out as much CPU dedicated to SONAR, I had a thought... perhaps the OS isn't getting enough CPU to do the required background tasks to support the audio sub system and stumbled upon another parameter in AUD.INI.  This parameter tells SONAR the max number of threads to use for SONAR (not related to the other multi core parameter).... in essence, the default value of ZERO, tells Sonar to go grab every thread for its usage hence giving SONAR its maximum CPU allocation...

So, as a test, I limited the number of cores that SONAR will control to SIX of the total 8 in my system. I am not sure my theory is correct but my low latency glitches all disappeared.  I played with this variable and found that Sonar using 5 wasn't enough for SONAR and SONAR using 7 threads, wasn't enough for the OS... hence my setup is set to use a "maximum thread count" of six. By the way, I think the variable is something like max_thread_count.... you can look this up as well.

Hope my shared experience helps you with your setup.

Jim

Thank you very much, Jim!  I am going to look into this as soon as I get home.  I will let you know how it works out.

I was able to make some progress by setting my max thread count to 6 and "Enable Set Thread Ideal Processor" to "FALSE" (this allowed me to get to 512 samples and none of my cores are hitting harder that the others).  I will keep playing with these settings to try to get closer to 128 or 64 samples.  Thanks again for the help!

I will keep playing with these settings to try to get closer to 128 or 64 samples.

64 samples / 96000 samples/sec = 0.67 milliseconds of buffer. That does not give your computer a lot of time to do much processing, and is certainly well below the latency most human beings could perceive.


To put it in perspective:


speed of sound 340 meters/second X .00067 seconds = .23 meters.


So that small a buffer would add the kind of latency you will experience by having your ear 9 inches from the speaker. 


Remember that what your computer needs is processing time (not samples) and that a buffer of 1115 samples at 96 KHz sampling rate provides the same time as a buffer of 512 samples at 44.1 KHz. 


You do not need a nitro fueled drag racer to get to work on the freeway.

I am running an L22 on a motherboard that implements a bridged (not native) PCI slot. I experienced similar problems, which were solved by REDUCING the buffer size. I don't know about your motherboard, but there is a history of issues with motherboards with CPU complexes that do not natively support PCI. You might also refer to the Lynx support forum: www.lynxstudio.com