I've been putting off purchasing a new PC for the past 2 years, waiting until I purchase a new video camcorder.  I am now very close to "finally" purchasing a camcorder (probably an AVCHD camcorder) and hence I am looking at also purchasing a PC to go along with it.  So I have some basic hardware questions, and I was hoping to get some views/feedback ...

I'm looking at purchasing a motherboard with an Intel Core i7 processor (with 4GB or more of RAM), but I can't decided if it is important to get SLI functionality.  When I surfed on this, I could not find any specific reference stating SLI (with multiple video cards) provided substantial performance improvements when multiple graphic cards were in place during Non Linear Video editing.  The best claim I saw as a claim of a 5% improvement for NLE.  I dont' see that as worth the extra money, and hence I'm inclined to not include SLI as a feature I want in a motherboard.  Any views on that?  (especially given ffmpeg now supports vdpau).  I concede my research here could be less than adequate, and may SLI has tremendous NLE benefits. I don't know.

And speaking of VDPAU, I am also curious about the value of extra memory on a graphic card for NLE, given ffmpeg can now take advantage of VDPAU.  Are there any views on the use of extra RAM on the video card ?  I read a page where one critic of a nVidia GeForce 9800GTX+ card (with 512MB of RAM) complained that the card manufactuer did not offer cards with more than 512MB, as the critic stated that limited high resolution/anti-aliasing in certain games, and also cancels out the advantage of installing the card in an SLI or triple SLI configuration (though it supports SLI).

How important is extra RAM to NLE wrt high-resolution/anti-alising?  Would a 1GB graphic card be much superior to a 512MB card for NLE?  and is high-resolution/anti-alising the reason why?  As for SLI, I can't even find a reference indicating what sort of performance improvement one can get in NLE with SLI, and hence I can't see why a 512MB card would hurt NLE for SLI reasons. ...

Also, some clarification on VDPAU, from what I understand it does not help encoding.  Rather I read VDPAU helps in offloading the decoding, post-processing, compositiing, and displaying compressed/uncompressed video streams on the GPU, where those aspects are also important to a Non-Linear Video editor.

Do I have that correct?   And any comments?

hi

Another question ... on a c2d is it better to install 64 bits os ?

is it faster for ffmpeg ? 

My goal is just to get the mximum performance for avchd operation ..

Thanks

Nyme

In part to address my own question , ... I asked about possible ffmpeg benefit from SLI (and multiple graphic cards) on IRC chat channel freenode #ffmpeg and the view was ffmpeg would not benefit much, if at all, from multiple graphic processors connected via SLI, even with vdpau.  Reference the benefit of adding memory beyond 512MB on a nVidia graphic card, the view was one would not gain much in a 1GB vs a 512MB card, when refering to 1920x1080 resolutions.

I also note the older nVidia GeForce 8400GS graphic card has a G98 GPU supporting VP3 (Pure Video 3rd generation, and hence superior VC-1 (WMV3, WMVA, and WMVC1 decoding) than the newer nVidia GeForce 9800 GTX+ with a G92/94 GPU supporting VP2 (Pure Video 2nd generation).  Pure Video, of course, being the Windows equivalent to VDPAU.  Although the H264 support by the G98 and G92/94 GPUs are the same.  It made me ponder if for Video deconding for a non-linear-video editor if the GeForce 8400GS is superior (due to its VC-1 support) but I think think that the more expensive 9800 GTX+ because of its faster GPU clock speed should be the superior (ableit far more expensive processor) for Non-Linear-Video editing, especially wrt H264 (ie AVCHD).  Of course this is assuming absolutely no games are played (as the 8400GS is NOT a gaming chip, while the 9800 GTX+ can be used for games).