[Joseph Owens]

I was under the impression that the recommended ATI card was the 5770.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

Have been dealing with this on a daily basis.

I think we’ll find its an RGB unscaled vs. Broadcast linear issue. It appears to be almost uncontrollable inProRes 4444 since you can’t flag it as RGB or Y’CbCr, but if Resolve sees a possibility to unscale the media to full 0-1023, it will. Remember there is a right-click available in all clips to toggle this.

You also have to do more than tick the scaling in Config. Even though there is an Auto- function for selecting scaling in the Render dialogue, and it should follow the project scaling, that is not necessarily the case. This was a bug from a couple of versions ago that I can’t say has been thoroughly exorcised. Try explicitly rendering a few clips with this setting selected to broadcast linear legal.

Also try bringing a few of the previously rendered media clips either into Resolve itself, or somewhere you can scope them. You will likely see that blacks are up around 8-10 IRE and whites never exceed about 90. This is not necessarily a gamma issue to start with, but because of the 0-1023 mismatch, the 2.2 scalar induces one, and makes a “re-correction” back in your NLE of choice very difficult — it’s not a challengeto re-establish a black/white point, but it’s fairly obvious that what’s in between is still screwed, and re-mapping graded media is not the objective here!

I ran straight into this issue on a DnxHD series’ last fall and this was the problem, no question because once the render scaling was fixed, we got cloned values on round trips. I’m seeing it again with ProRes 4444.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

[Mark Raudonis] ” There were some choices made during the
design process that just make me ask “What were you thinking?”.”

1. Trying to “build on the brand” out of one side of their mouths and calling it a “1.0” of something out of the other.

2. The main thing it does is cuts and dissolves. Now off the top of my head, how many applications can do that?

3. The marketing department lost track of what the difference is, to a buyer, between a “feature” and a “benefit”.

4. FCX looks more to me like “Esperanto” and the “Dvorak keyboard” every day.
Yes, the invented super-language and the astonishingly efficient letter layout. And how has that been working out? Or should I type (you wouldn’t believe me if I said I swapped out my keyboard to Dvorak to do this, but ) Kiel havas ke estita ellabor? Get over it, its the new pair-a-dimes. The two I have left to rub together.

5. True, FCP7 needed to be burned to the ground, but not with everyone still in the house.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

I’m going to add this comment for the record.

Admittedly there is a price difference, but unless a scope offers a YRGB parade, it is of limited value to color grade. RGB-only does not and cannot give any sense of the derived Luminance (Y’) value. If you cannot see and measure your absolute white/black point, this pretty much overrides any of the economies of non-real-time, sampled software scopes. Also, very, very few software scopes can offer accurate Y+C legal gamut evaluation.

I am surprised that Resolve does not have better internal numeric/waveform/vector display, since it was daVinci (the original manufacturer) who was militant about getting measurement right, and providing as much data to an operator as possible. Trainers like Bruce Graham starting hammering this into trainees on Day One if you were on one of his courses at Lauderdale-by-the-Sea (and that was 18 years ago this month, for me.)

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

[John Petaja] ” Red media (scarlett 4k)”

My impression was that Lite was limited to an upper resolution of 1920×1080. But you might be scaling the “4K’ down to HD, so maybe that’s allowable.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

[TaeJeong Kim] ” I will have 7 or 8 timelines for the project.”

I haven’t definitively decided yet whether some of the issues I have been seeing with “Running out of Memory” and flop-crashes to the desktop are project-size related or “something else”, but definitely smaller seems to be better. So my advice actually is to create several separate projects (especially since this is your first adventure with an overall project of this size) so that you aren’t having to cope with ALL of your media loaded at the same time, every time.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

A lot of questions.

I don’t own a REDRocket, but my impression is that it does fast de-Bayering and that’s all it does.
For about $7000 dollars.

Resolve doesn’t just “prefer” nVidia cards; until such time as CUDA starts appearing on other platforms, it is the “only” card. I read somewhere that nVidia has recently opened the door to open-sourcing that method of processing, so other manufacturers might become an eventuality. By then we all might likely have jumped the Mac ship anyway, given the reported speeds available under Windows7 and its nVidia support as opposed to Apple hostility, both w.r.t. drivers and physical hosting challenges.

I have a similar internal drive strategy, although I use drive 2 for cache/RTE renders, and RAID-zero 3/4 in a single stripe for double-spindle access to dense media. I also wedge another drive into the optical2 drive space for compressions and baked-movie output. Plus a port-multiplied eSATA slot because almost 100% of my work is with external projects on portable drives. I also build dual-enclosure 4TB RAID-0 external drives for backup, archive, and export.

You will also find that half- and quarter-res good proxies are your friend, even with multi-core MacPros. They’re just not that fast, really, when it comes to doing more than GarageBandX.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

Ugh is right. I converted a 1920×1080 2997 feature (shot in original 2997) to 25PAL last year. Cut it into 5 reels cinema-style and used multiple machines to do the FRC. It was acceptable coming out of Compressor with the “better” filter, but it still took a combined 85 hours. Seemed like if I wanted to use the “best” setting, the render estimate was “add a zero” onto the total time, and three weeks seemed excessive.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

Just a basic tip… don’t expect to handle the entire project in one timeline, and you should probably split the project into logical “reels” of somewhere around 20 minutes each. This is economical for the DaVinci browser, and is practical in terms of a 2,000 foot mag of 35mm, in case this is destined for a filmout.

jPo

You mean “Old Ben”? Ben Kenobi?

[Joseph Owens]

[andrew sableton] ” I thought the 4000 would be better”

Yeah, me too, but in practice it is not the case.

Even some benchmark testers still find the ATI 2600HD faster under some circumstances (non-CUDA).
Most users who have explored the flashed GTX series have the 4000 pegged as a very expensive lame duck.
jPo

You mean “Old Ben”? Ben Kenobi?