HD For Indies

Did some testing, here's what this posting covers:

-mixing 8 & 10 bit, Apple & Blackmagic codecs on one timeline - what's RT and not, under what circumstances?
-Render times for color correction and transitions
-relative render times of 8 vs. 10 bit timelines & footage
-render disks - does speed make a difference in how fast renders take?
-----------

I'd heard of folks successfully mixing different codec footage on a single timeline for SD stuff.

So I decided to mess with it for HD stuff.

Some notes:

-if you make a 10 bit sequence (4:2:2) using the Apple Uncompressed 10 Bit Codec, and drop 8 bit footage onto it, the 8 bit footage gets a red render bar and doesn't play back at all (blue/black screen with "Unrendered" on it)

-interestingly, all of the new Dynamic RT settings are NOT available on a 10 bit sequence. Your only choice is Full Quality checkmarked on or...still checkmarked on.

Mixing 8/10 bit and Apple/Blackmagic codecs on the same timelines

I started doing some testing that got confusing, so I did this:

I took 2 clips of the same footage that I already had:
-BlackMagic 8 bit 2vuy (4:2:2)
-BlackMagic 10 bit (4:2:2)
Using those as starting points, I converted them using the following Compressor Advanced Format Conversions:
-HD Uncompressed 10 bit 1080p24
-HD Uncompressed 9-bit 1080p24

So I now had 4 clips.

I made 4 comps, two from presets and two "illegal" ones

Legal:

HD Uncompressed 10 bit
HD Uncompressed 8 bit

Illegal

Blackmagic codec 10 bit
Blackmagic codec 8 bit

Then I dropped all four clips onto all timelines.

Here's what I learned about which did or didn't play back (fine or with red line requiring rendering). All are 1920x1080, square pixels, 23.98 timebase:

"OK" means no red line, and a quick test showed it playing back full speed.
"red" means it showed a red "requires rendering" line and

Apple Uncompressed 8 bit timeline
(playback quality and frame rate both set to "High")

Apple 8 bit: OK *
Apple 10 bit OK *
BMD 8 bit: oddly, red
BMD 10 bit: OK *

* (I could apple 3way color corrector and do anything BUT adjust sliders under color wheels and NOT get a render bar. COULD adjust saturation though)

-turns out this is the most interesting new twist.

-in this sequence, applying a cross dissolve between two Apple Uncompressed clips, one 8 bit the other 10 bit, required a render, even with NO other effects added to those two clips. Applying a 1 second cross dissolve betwen these two clips (with no CC or other FX) took 5 1/2 seconds (stopwatched) on my dual 2.5 GHz G5. Same render status with all other combinations of uncompressed Apple/BMD 8/10 bit codecs - rendering required on an Apple Uncompressed 8 bit timeline.

-dropping the Playback Quality to Dynamic gave a green render bar - it will simply cut the resolution in half to render it on the fly in realtime


Apple Uncompressed 10 bit timeline:

Apple 8 bit: red
Apple 10 bit: OK
BMD 8 bit: red
BMD 10 bit: OK

-there are no Dynamic RT choices with 10 bit - the only RT option applies to tape output, not timeline playback.

-a 1 sec cross dissolve between two Apple Uncompressed 10 bit clips took 12 seconds on my dual 2.5 GHz G5 (two and a half times longer than 8 bit!)

BMD 8 bit timeline (Blackmagic 8 bit (2Vuy) codec):

Apple 8 bit: red
Apple 10 bit: red
BMD 8 bit: OK
BMD 10 bit: red

Again, no Dyamic RT options here.

BMD 10 bit timeline (Blackmagic 10 Bit (DV10) codec, not the one used in v4.5):

Apple 8 bit: red
Apple 10 bit: red
BMD 8 bit: red
BMD 10 bit: red

This is a bogus test - the older Blackmagic 10 bit codec is unavailable in FCP 5 with the v5.0b1 BMD drivers. BMD really expects you to use the Apple Uncompressed codec.

Lesson learned: Dynamic RT offers one advantage for uncompressed HD work: you can put Apple Uncompressed 10 bit footage on an Apple Uncompressed 8 bit timeline and get RT performance. But that's it - still no RT stuff in a natively 10 bit timeline, regardless of whose uncompressed 10 bit codec you use.

10 bit rendering takes LOTS longer than 8 bit rendering. Applying the exact same ugly color correction, where I moved ALL color wheels and sliders (for a craptastic look), took this long on two setups of the same footage, one 8 bit version one 10 bit version (23 sec 3 frame clip):

Apple 8 bit sequence, Apple 8 bit codec: 1 min, 10 secs (70 secs to render on a 23.125 sec long seqence, so a 3:1 rendering ratio)

Apple 10 bit sequence, Apple 10 bit codec: 2 min, 55 secs (70 secs to render on a 23.125 sec long seqence, so a 7.5:1 rendering ratio)

Yowch - so in FCP 5, 10 bit rendering takes 2 1/2 times longer than 8 bit rendering.

I don't know how much difference it makes, but I'm using a 4xSeagate 400 array on a Sonnet Tempo X 4+4 SATA card striped into a RAID 0 with SoftRAID 3.1.3.

BlackMagic Disk Speed Test run on this almost full array (170 GB free of 1.46 TB) gave 90 MB/sec reads, 162 MB/sec writes.

Render Disks - does speed make a difference?

For compressed SD render disk speed doesn't matter, it's CPU bound. For uncompressed HD, dunno...let's find out!

I reset the scratch disk for renders to my cheapie little 120 GB drive I bought at Fry's for an alternaboot volume - a Maxtor 6Y120MO 120 GB.

Apple 8 bit timeline, Apple 8 bit clip - 1 min, 16 secs - took 6 seconds longer, or 8.5% longer for the Maxtor vs. the 4 disk array.

Apple 10 bit timeline, Apple 10 bit clip - 3 min 1 secs - took 6 seconds longer again, or 3.4% longer for the Maxtor vs. the 4 disk array.

For comparison purposes, I ran the Blackmagic disk speed test on the solo 120 GB - it performed at 46 MB/sec reads, 44 MB/sec writes.

An empty 4 disk array, or an 8 disk array would, I suspect, widen the advantage, but not by much I'd guess. It seems fairly clear that with a 2x read advantage and a nearly 4x write advantage (and rendering involves readng and writing), that Video Render scratch disk doesn't make a difference in render speeds. It definitely makes a difference in how many simultaneous tracks you can play back simultaneously (even if that number of tracks is or is not only 1 track).

An interesting test - for SD and compressed HD work, would renders be as fast over a GigE network as locally? I've done some heavy computation After Effects rendering tests in the past and network (even reading from a single drive) was about as fast as local disk performance for the very computationally intensive stuff I was doing.

OK, enough for today. Time to go run in the 93 degree heat. Welcome to Austin! Crematoria is just down the block - "If I owned this place and Hell, I'd rent out this place and live in Hell." Well, I only feel that way during the summer, when I'm foolish enough to run at the hottest time of the day. By August I'll be used to it. Or dead.

Whee! :D

-mike, who wishes his conveyance would glow smoky red when he arrived at his scorched blacktop destination. Party Poppers would be cool, too, for sudden stops.

PS - (did anybody else notice that the Big Red Button for the retro jets was actually labelled "Party Poppers?" Somebody in props had fun) You'll get respect, and maybe a freebie question answered, if you know the name of that ship.