Why? Better quality, standardized format, more detail.

Higher resolution video has better detail than the lower, because detail is there to begin with.

If you means because the screen on N900 isn't large enough, then so what? My DSLR doesn't have a 10 Mpx screen.

--

I didn't edit my post because they're both quite large.

Observations:
* ATTO's figures are quite close to mine. That is, over decent buffers throughput's what limiting the writes.

* Note that at 400-500K/s average, you need at least 2K buffers on the EMMC and SD. Any lower and it stutters. Once it drops a frame, flushing buffers for at least half a second means you need a peak of 2 M/s to recover otherwise the segment is lost, that means a minimum of 4 K(?)

* I tried both overlapped IO, direct IO, etc. All figures were withing 5% of each other on lower buffers, and identical on higher buffers (4K+)

* Further testing showed that panning isn't the most load on the camera. From 3200 Kbps, it peaked at 3600 when panning. What pushed the thing over was auto-guessing brightness/exposure. Fading up and down when walking into the sun/shade wen over 4000 kbps (500K/s).

* Also, I was able to sync the dropped frames with the bit rate. Every time it dropped, bitrate was trying to push 4000. As a result, I tried to lower the resolution to fine 4:3 (640x480) and while I have better performance it is also choking on 4000+ bitrate.

We may have a software limitation in place?

Here are the performance tests on both storage systems: