Daylaa
I hate to be the bearer of confusing news - but I have tonight attempted to reproduce the issue by changing the record and render settings back to '64' and then pushing up my sound card buffers to 1024.
 
The Pro Channel continued to operate seamlessly - whilst this sounds like a positive thing, it surely means that we haven't yet found the cause of the problem?
 
*Edit: I have experienced days of seamless Pro Channel operation before only to switch on another day to find the problem has returned.

I've already mentioned the PCI card issue. The inconsistency may rule this out but still the card may be right on the edge of functionality (as was mentioned by someone else) and something minor on your system is pushing it over the edge. As I said if you can rent or borrow a solid USB interface (like a Focusrite) to test maybe that's the issue.
 
However... since this is intermittent I'm wondering. How quickly are you starting up Sonar after booting? Do you just plow into things or do you let your system "warm up" a little? Sometimes the system takes it's time finding all the drivers and devices and if you open Sonar too quick things go weird. I've had issues with MIDI controllers too where if I don't use a specific sequence to turn them on they just won't work. Example: If I turn my M-Audio Oxygen 25 on AFTER I open Sonar... it doesn't work. My padKontrol sometimes works, sometimes doesn't depending on how I start everything up. It definitely takes a little while for the system to detect MIDI connectivity (just like it takes a while for it to tell me whether I've got my network cable plugged in).
 
There are just lots of little things that can get messed up if you force the computer to do something it hasn't prepared itself to do yet. Keep track of how you booted up, how long you waited before opening Sonar, what devices you turned on or after opening Sonar and what results in good behavior or poor behavior.
 
Some people look at me a little sideways about my meticulous boot up ritual and device set up procedures but if I don't follow then I am more likely to have something not work or get a crash.
 
Our DAWs are like fancy European sports cars. Sometimes you gotta ease it into gear and pay attention to little nuances as you start 'er up and go ripping down the Autobahn.
 
I however am no tech. I'm still convinced smacking the monitor makes the computer work faster. That isn't nearly as satisfying with a flat panel screen though.
 
;-P

Beepster
Keep your samplerate consistent with the project samplerate. 44.1 is CD quality but most people record above that and mix down (with dithering) to 44.1 for their final product. 

Beeeps. I'm sure you know this but you mis-typed/spoke on the dithering. Dithering is for bit rates (32/24/16/). You dither when you step down from 32/24 to 16.
 
IOW if you record at 16/48 but output at 16/44.1 you would not need to dither.

Actually, Bapu I was indeed having a brainfart and thank you for the correction and apologies to the OP. I used to get all that stuff confused (as in I thought somehow dithering affected samplerate as well) and most certainly had corrected my faulty assumptions but today has been one of those days and it crept back into my rhetoric. In fact that entire post was mostly based on very non technical anecdotal evidence so should be taken with a grain of salt.
 
Still though I've noticed that if I give my system a little time to get itself sorted out before opening Sonar and turn on certain device in a certain order I have far less problems.
 
Basically Beepster voodoo but considering OP is experiencing the "gremlin" perhaps a little attention to such things may reduce the odd behavior.
 
Brainfarts, voodoo and personal ritual aside that PCI card does stand out to me as a potential problem.
 
Cheers.

Daylaa
P.s - My top ratings for 'DPC Count' (Not counting the top placed one):
 
2= rdwm1027.sys (Roland)
3= Intel Rapid Storage Technology Driver - x64
4= NT Kernel & System
5= NDIS 6.20 Driver
6= Virtual Machine Monitor
 
Interestingly, the 'highest execution (ms) was 'Virtual Machine Monitor' followed by Intel Rapid Storage Technology Driver - x64 and NDIS 6.20 Driver.
 

The iastordatamgrsvc.exe under hard page faults is also Intel Rapid Storage. Are you running RAID? If not you may be able to disable this (I believe MS provides their own SSD drivers).
 
Based on your very powerful system, you should not be experiencing problems due to latency, sample rate, etc. And since I believe you said earlier you had the problem just running a sequence into a soft synth, I can't believe it's a disk throughput problem (though if you're using external drives, I'd generally recommend eSATA rather than USB).
 
That leaves things like drivers (sound, video, mouse, network, etc.), installation problems, HW problems (doesn't strike me as likely here), or of course SW bugs.
 
I would disconnect your network, shut down any firewalls and anti virus and also make sure nothing is trying to update or otherwise do anything in the background. Close everything except Sonar. If there's anything not absolutely necessary in the Taskbar notification area, right click on it and shut it down if you can. If you don't have RAID and can shut off Intel Rapid Storage without problems occurring, try that too. Make sure power management is not throttling the CPU.
 
Microsoft has a tool you can download called "Process Explorer" which is sort of like Task Manager on steroids and can show you very detailed information about each process and look at things like GPU usage as well. When I run X2 (64bit) on my much lower powered laptop here, spinning the ProChannel knobs increases Sonars CPU usage by only a few points and has even less impact on the GPU.

Thanks to everyone that contributed to this thread.  It has been very useful.  I built myself a Core i7 Ivy Bridge beast a little while back, and just assuming it would handle anything I threw at it, I had set all bit depths to 64.  I'm a software engineer, I should know better and I guess I just got cocky :)  My machine did actually handle most of everything I threw at it, but it makes absolute sense to set the record depth to 24 and render to 32. 
 
My question is, would changing these settings mid project create any issues?  Maybe this is more for Noel, or anyone that has attempted it.  I have more tracks to record.  If end up having a mix of 24 and 64 bit files, am I asking for trouble in Sonar engine/plug-in land?  (I'm now on the latest X3)
 
Would love to hear a success story...

I've always heard changing bit depth mid project is a double plus ungood thing to do.
 
YMMV.
 

Okay... the following needs to have a serious warning attached to it. I just barely grasp the concepts of samplerate and bit depth and all the various point within the software and mix process. The general concept of picking a samplerate and bit depth and sticking to it until final mixdown is solid but my blatherings about the internal audio engine and what not may be inaccurate hackery. I'm also doing a bunch of other crap so I haven't edited this post properly or thought it out as well as I could have. Just didn't want to see you doing something weird to your work. Hopefully the smarter dudes will correct any mistakes I have made.
 

dman301
Thanks to everyone that contributed to this thread.  It has been very useful.  I built myself a Core i7 Ivy Bridge beast a little while back, and just assuming it would handle anything I threw at it, I had set all bit depths to 64.  I'm a software engineer, I should know better and I guess I just got cocky :)  My machine did actually handle most of everything I threw at it, but it makes absolute sense to set the record depth to 24 and render to 32. 
 
My question is, would changing these settings mid project create any issues?  Maybe this is more for Noel, or anyone that has attempted it.  I have more tracks to record.  If end up having a mix of 24 and 64 bit files, am I asking for trouble in Sonar engine/plug-in land?  (I'm now on the latest X3)
 
Would love to hear a success story...

From what I understand Sonar 64's audio engine does everything at 64bits anyway despite what you have the project set to. Changing your bit depths mid session usually isn't advisable but if 64bit is giving you a problem then you could bounce everything you are working on down to 24bits (which seems to be what most people use as their bit depth settings). Just be sure to use dithering when doing these renders. There are different types of dithering so probably best to read up on that but I thing the Pwr Dither ones are the more quality algorithms. Also it's likely best to disable Fast Bounce so when the dithering is applied it is done in real time to avoid glitches. Cranking up your interface's audio buffers might be a good idea too but I'm not sure if that would effect anything but if the resulting bounce is causing pops or crackles that's a good place to look. I am not sure why you would want to render back up to 32 bits but that seems like a bad idea. You usually don't INCREASE bit depth. You generally work at a higher bit depth (like 24 or 32) and leave it there until the final mixdown then bring it down to 16 bit for CD burning. Even a mixdown you intend to send to a mastering house should remain at 24 bits. Perhaps you mean for internal bounces but I think just using the default settings would probably be fine... but seriously that's a weird thing to say and I'm not sure what you are referring to.
 
Sorry... that may sound scattered and weird and it is because I still find the topic confusing BUT from what I've learned....
 
While tracking, editing and mixing you would use a higher bit depth. 24 seems to be what usually gets used. Leave the project/interface at 24 bits (or whatever you choose as your bit depth) until you are ready to burn some CDs.
 
Once you are ready to burn to CD use the Mixdown/Export dialog to select your new bit depth which will be 16bit (and your samplerate will be 44.1khz).
 
That really is the only time you should have to worry about changing bit depth as far as I know.
 
Frankly unless your 64 bit projects are giving you a super hard time I'd leave them there until you mix down but any new projects you open definitely set them lower than that. Every time you dither down you are yanking out bits of info which is why you try to do it only once.
 
When mixing down to mp3 you don't even need to worry about bit depth in some cases because mp3 can actually run at 24 bits. It's just CDs.
 
And remember no matter what settings you have a 64 bit version of Sonar will process all your audio internally at 64 bits anyway as long as you have the Use 64 Bit Double Precision Engine selected in your Driver Preferences.

Thanks for the words of warning.  What I missed was that record bit depth will physically write 64 bit samples to disc.  I was definitely at fault for not understanding what these internal settings really did.  I just always assumed my tracks were 24 bit in, 24 bit on the disc, and 24 bit out until I dithered down to 16 for Redbook, and all the 64 bit processing was internal to the engine only.  Hell, I never used dither going from 64 to 24 when I printed because of that assumption (although it sounded fine regardless because there probably wasn't significant data in the upper bits anyway... I'm curious if I'll notice a difference when I can test tonight...).  I would like to know more about the function of rendering and how it applies, especially if the engine is always working in 64 bit mode.  Does it even matter what render is set to at that point?  Is this in the documentation somewhere (I'm at work and should probably be working so I will have to check later)? 

If you have one of Scott Garrigus' Sonar Power books he gives a better explanation than the manual.
 
I only recently learned about the internal bounce rendering bit depths a while ago and now understand what you may have been referring to. Like when you do a clip bounce or bounce effects to clip? Is that what you mean? I can't speak to that but IIRC that is indeed independent of the project bit depth just like the internal audio engine processing's bit depth. How that should be set I do not know but I've never had to screw with it and everything has worked out fine. I'd go with whatever the default is for internal rendering until you learn more. Checkout that Preferences > Drivers: 64 Bit Double Precision Audio Engine setting though because that's supposedly where all the quality internal processing magic happens. Aside from that all you should have to do is start you project at 24 bit, do all your work, then use dithering when mixing down to CD quality.
 
I don't even render down to 16 bit for my "final" mixdown. I do the mixdown at the same project bit depth/samplerate and then open the stereo wave in a new project to pretend like I'm mastering it. lol
 
This can be done within the existing project by bouncing to a new track or treating you master bus as the final master but I find that to be too confusing so I treat the mixdown and new project as if I'm sending it to a mastering house... the mastering house being derpy ole me. ;-)
 
Seriously though I'm just starting out. There are tons of articles on this kind of thing and much smarter dudes here that know what the deal is. I still hope you got a little help out of my blatherings.
 
Cheers.

By the way 64 bit does nothing for you unless you have more than 4gb of RAM.