[PLUG-AV] other system options

[Jason Nicholls]

Hey James,

> I called Dell; they CAN'T take Windows off this! They can't swap the HDD for
> an SSD. There's basically no customisation of any parts on the vostro
> series. Quote attached. I'm sure we could grab an SSD for the video capture
> hosts if we want - small is fine - HT has a 16 GB SATA Kingston SSDNow for
> $66, probably better prices around.

With an eSATA port we can get any external enclosure with SSD (eSATA
just because that's the best performing interface for it). Regardless, SSD
isn't a requirement, it's just nice to have.


> For the DVswitch host - looks like the fastest clock speed CPU is around 2
> GHz with speed boost to 2.9 Ghz. DVSwitch wiki (a little out of date)
> suggests "The DVSwitch GUI does require at least a 1 Ghz CPU, but the GUI
> will flicker until about 2Ghz." Storage is going to be limited to one disk -
> so would have to be a traditional HDD here for storage of video in a
> combined dvswitch + encoding platform. Perhaps we still get an Optiplex with
> a 3.4 Ghz core i7 as the encoding host (with the gobs of storage) and have
> the DVswitch host swapped to an SSD too?

Summary of what I've written below: the CPU in the Vostro 3750 is more than
fast enough.

The above is not a useful comparison as cpu frequency between generations is
meaningless so posting a frequency needs to be coupled with the actual CPU
generation/architecture... To compare that P4 at 3Ghz we have is equiv to a low
end Atom @ 1.6Ghz. The Core i7 2635QM @ 2Ghz is (much) faster than just
the previous generation Core 2 Duo quad-core @ 3Ghz. It's also much faster
than that quad-core Athlon upgrade for the existing box we have. Doing some
division by the number of cores, each core on the i7 2635 is approximately 2x
faster than each core in my Core 2 Duo @ 2.53Ghz (so equiv to ~5Ghz!).

For reference using ffmpeg2theora (without any capturing or post-processing
beyond de-interlace) I can encode at close to running time on my Core 2 Duo
system with the options for video quality = 4 (out of 10) and audio =
0 (out of 10).

A CPU that's twice as fast, and potentially faster with turbo-boost,
then we should
have no trouble encoding in real-time with headroom to increase
quality (although
with theora that comes at a cost of a larger file/stream).

If you see all my previous posts the bigger issue I have is encoding with x264
with post processing to achieve excellent quality video with as small
a filesize as
possible. This is way too slow for streaming but doesn't matter cause it can be
done afterwards. Theora can't match the quality for size, and I haven't tried
WebM yet.


I'm arguing because it's silly to carry around more stuff than necessary and a
waste to purchase more than we need, even if it's "free" thanks to the grant.


Jason Nicholls
jason at mindsocket.com.au
0430 314 857