I get better results using my CPU for SVP and just recently using Nvidia CUVID via LAV than using OpenCL for SVP and no HW acceleration for video. Even when using my CPU for everything I still have plenty of headroom for SVP's 3c profile.
    As far as using madVR, I do use it quite alot actually with with Lanzcos for all chroma and Jinc 3 for Luma upscaling and Lanzcos for Luma downscaling. What I'm trying to do is to get SVP to work with Jan Willem32's renderer regardless if madVR is superior or not.

I've had some free time so I've been able to experiment some. It seems the problem is related to the renderer I'm using in MPC-HC. I've been trying to use SVP with Jan Willem32's MPC-HC tester builds for internal renderer fixes http://forum.doom9.org/showthread.php?t=161047 . It's basically an alpha build but is a good alternative to madVR. I've used vanilla EVR-CP(via MPC-BE) and madVR and I don't get the delay issues with these renderers at all and both renders are using SVP 3c profile to it's fullest without any unexpected issues. I'll contact JanWillem32 and ask him if there is something he can do on his end. I've done some informal testing for him before so I don't think he'll mind.

I've checked multiple menu items (turning outcrop off, changing video player priority, disabling outer lighting, and etc.........). I even tried modifying the video delay values up to +1000/-1000 ms via the svpmgr config. Believe me, I know what a PITA it is for multiple people to ask the same question over and over again without trying to research and figure out things themselves first, so I will try not to do it to you folks.

Is it really my graphics card? It's not too bad, just really, really old. It only has 112 shader cores but it is 256 bit PCI-E 2.0. It can handle madVR Lanzcos luma/chroma @ 1080p 30 fps with out issue. Besides, I was under the assumption that the settings I have for SVP don't depend on the graphics too much, more on the CPU.

Oh and I am NOT using madVR with SVP.....before you ask. smile .

I see where you're going with this but I don't think that's it. It fluctuates from exactly .98x or 1.02x. I get the same exact figures even when I switch the profile to 1c. I have a AMD FX 4300 Black Edition Vishera(piledriver) CPU (all 4 cores are unlocked by manufacturer) running @ 4.2 GHZ and a crappy Nvidia 9800 GT 1 GB and 4 gb DDR3. Needless to say I'm not using OpenCL only the CPU. When playing 1920x1080p @24 fps content and only using CPU(no video card acceleration ie. CUDA, DXVA). My CPU usage rarely goes beyond 64% on the 1st core, a few percentage lower on the other 3 cores.

The frame interpolation seems to be working as it should be but as time goes on the video delay gets worse and worse. As you can probably guess, adjusting the video delay via the svpmgr menu isn't a workable solution.