Johnny,
What may be the "mother of all threads is here.
Proper "sync" is an absolute must, and most are barely (if at all) cognizant of how it works.
Danny Danzi did a video tutorial at one time, can't remember if its in the above thread or not.
My Tascam FW-1884 is, of course, a firewire device and has always been spot on; Sonar 7 through
to Platinum; well except at 2048, then it is one sample off.
I'm running the UFX on the USB buss, its "spot on" 48 to 2048. But that's RME's forte; drivers, drivers, drivers.
There were sync issues beginning in X1, but this was a Sonar problem if you "Looped and Punched In".
Each successive "take" was placed later and later. Other than that, sync is "tight as a tick"!
I never could find the rhyme or reason; the "hidden" buffers maybe?
As to the "search" function of the forum; I'll leave that to another thread.
Its always baffled me that Cake does not have a definitive white paper on the "how and why" of proper sync.
Maybe it would just create a huge "support" issue?
For the most part the forum members seem to help one another to get it sorted.
Thanks for bringing it up; we need a "sticky" that lays out the "facts" and loses the "noise".
T

Please check my sig.
Cheers:)

So for all you Sound Blaster fans out there I have good news in ASIO mode it surprised me and passed the test. WDM mode was a train wreck. WDM mode is now officially banned from my list of audio driver options. read on
I had not bothered with the Centrance or Oblique Audio RTL test as what is documented by people is it will only show you what Sonar will also show you so really doesn't add much. Only thing is it will do this without the need to even open Sonar. All you need is the loop back cables and hit the button. There is some fussing with selecting the proper in/outs. 
Any how it's free and works outside Sonar, you can use it to compare Sonars results. 
http://www.oblique-audio.com/free/rtlutility
 

 
It showed 5804 samples and Sonar showed 5860 that's 1 ms off. 
 

Now Notice the Buffer size of 2980 but the Oblique test shows this at 2880. Is this our hidden buffer?? 
 

 
But oddly enough the Audigy was actually within a few samples when I zoomed in. I counted 28.  

 
So ASIO was totally fine and it was stable over a 3 minute period. No so with WDM. 
As you can see above it was early which seems to be normal for WDM but it drifted after only 15 measures  by 24 samples, It kept drifting throughout the 3 minutes. No 2 hits were the same. 

 

Mine was set ages ago, but this is a good reminder to check it again since I've just this past weekend finished rebuilding windows on a new OS drive for this computer.  I doubt my latency offset was transferred to the new install of sonar, so I need to check that.

Splat
Please check my sig.
Cheers:)

Grem
What if I'm on a mobile device and I can't "see" your sig? Any conversation based on what ever is in your sig is lost to me and others on mobile devices.

What if I can't see my sig as I'm on a mobile device ;)

Maybe somebody could tell me what my settings are ;)

Splat

What if I can't see my sig as I'm on a mobile device ;)

Maybe somebody could tell me what my settings are ;)

Oh what a burn!! ;) LOL

I didn't think you were on a mobile device!!

Thank you Robert. What interface is it?

I finally installed Platinum and revisited this since it occurred to me that I slide clips around without giving it a second thought (just considered it part of the work-flow). Internally I could care less about offset since I am rarely synching that to anything, but then it occurred to me that the delta on my Saffire may be from it sending the "expected internal" offset rather than the "expected external" offset to SONAR (or possibly that it sends both and SONAR is using the wrong one).
 
Anyway... I also found another "non-intuitive" thing in SONAR... since I was 76 samples late, I inserted a manual offset of "-76" and it ended up doubling it! Changed it to "+76" and the loopback was sample accurate. This is actually a very cool thread and is something anyone should consider trying.
 
I also wanted to point out another thing I would "judge" an audio interface by... simple engineering definitions of "precision" and "accuracy"... Thinking in gun shooting terms, "accuracy" is how all-over-the-map one gets shooting at center... "precision" is putting a bullet through the same hole in the paper each time regardless of accuracy. When precision never wanders, accuracy is a simple sight (offset) correction... but when precision wanders, you are pretty much screwed. That said, my Saffire is 100% precise, which offset will correct accuracy on... but a few of the examples above show "walking precision" which makes them "highly inadvisable" for an audio interface.

Wow! Thanks Tom. That ought to make me go right out and buy one! LOL Actually, I'm waiting for a reply on a FF800 (though I'd need a FW PCIe card, too). I talked a buddy into getting one and wonder why I never did! SHEESH Using Budget as a rationalization is always bad! LOL
 
Michael

I guess we should all bite the bullet and get RME or Motu and be done with it. But we knew that before I started this thread. What I'm stil after is those el cheepo interfaces.
 
So getting back to el cheepo interfaces.. ( tascam us1641 $199  with Cubase LE 5)  
 
A bit of a surprise today. in my first post I originally tested my Tascam us1641 and the results were less than stellar. 400 samples early.. That's around 10 ms @ 44.1.  I never thought that the performance of a driver should be much different from computer to computer but I just found out this is not true. The computer must make a big difference. 
I originally tested the Tascam on a new machine I built as an office computer. It's good stuff but with a wimpy processor.  But the rest is not a heck of a lot different than my DAW. 
DAW= Gigabyte Z97X-SOC-  SSD- 16 Gigs RAM - Intel Core i5 I5-4460 Haswell 3.2GHZ Processor 
 
Office = ASRock H97 PRO4  -7200 HD- 8 Gigs RAM - Intel Pentium G3220 Dual Core 3.0GHZ 
 
Both are the same OS windows 8.1 and all tweaking was the same. 
I have a 4 piece  Jazz combo coming in next Monday so will have to use the Tascam as I need 8 channels. So I installed the new drivers from Dec 2014 on my DAW  and fired it up. On a whim I did the loop back test and what the heck? It's only about 16 samples early? This is  a big improvement over the 400 samples on the office computer. So now I'm seeing a new aspect to the building of a DAW.... I wonder what will happen if I test the Tascam on a top of the line i7??  
So this is good news for Tascam owners, if these new drivers they are writing in house are much better. 16 samples seem to be close to what most are reporting. So now I will test the Scarlett on the office build and see if it performs is worse. 

I revisited this based on this post by Brundlefly and downloaded the "CEntrance ASIO latency tester" It gives identical results and is quicker to check/modify routing and buffer settings on the interface.
 
One thing that did occur to me with mine is the internal DSP. When internally looped back, it seems the DSP routing is bypassed completely (simple in->out engagement), but the extra buffer is used to accommodate the internal DSP processing (and apparently not reported to SONAR). I tested several buffer sizes at all rates and the offset was identical for each. Internal loopback is off by 6 samples, external 76.
 
I did not check the internal VRM feature because I was too lazy to find my splitter to get the headphone out routed back in, and that is a "mixing only" feature anyway.

Yes you can use the Centrance or the Oblique testers they both do the same it seems. 
The visual loop back test is more involved but the advantage is it shows you fairly quickly that something is off. 
 
Out of researching for this thread I've certainly found that my Tascam us1641 is now at least not the door stop I thought it was. I recorded and mixed a 10 song Jazz 5 piece band on Monday and Tuesday without a hitch. So I'm still getting my $200 back from it ten fold. To bad it caused so much grief with the old drivers. Thank you Tascam for taking the time to update drivers for your old stuff. 
 

Michael, I will see if I can switch to 48 with the Tascam. I seem to remember that this was an issue for me too. My old Sound Blaster projects where all 48 and I think they barfed with the Tascam.  Because my Yamaha 01V digital mixer only works in 44.1 I'm sort of stuck at that clock rate  so  have not used 48 for a long time now. I'm moving towards changing this as I kinda like 48. Having the extra inputs of the Tascam and it now working nicely changes my game plan so I might not need the 01v anymore. But if the Tascam is also stuck at 44.1 that's a bit of a bummer.   
 

+1, if SONAR has no project with existing audio loaded, it will allow you to shift sampling rates inside SONAR. Once audio data is present, SONAR locks itself to match (and will also lock the interface if they talk correctly).
 
There are a few third party tools which will batch convert more quickly; but within SONAR itself, it requires exporting the audio and doing a SRC on export to shift rates. I have never actually done this, so not sure if this also requires a new project to be created and have the audio imported. In such a situation, I will work the project through and then SRC the final product (audio for video use in another program).