CS5 observations and intermediate codec usage
 

Overall CS5 is a very nice upgrade. I missed out on CS4 and went from 3 to 5.

At the moment i have an ATI FireGL card and so i will need to get an nVidia one to take full advantage but comparing software to CS3 i found that cineform under CS3 still produced better, smoother playback.

However i was careful under CS3 to ensure that everything was cineform encoded.

Now i don't have cineform under CS5 (save for the playback codec) and am using source footage - XDCAM and H264 from canons. These are both more CPU heavy codecs than cineform and just because we *can* use them i don't know whether we should still be transcoding.

What is everyone else doing?

And when Premiere is rendering, what format does it render into for previews when there is mixed footage on the timeline? (The same format as specified in the sequence settings?)

Is DNxHD a worthy intermediate codec these days?

I'd love to hear some other approaches, i think i'm being very inefficient at the moment...

cheers
paul

I have been a Cineform user since the early days of HDV and PPro v. 1.5.
I am currently using CF Neo4K v. 5 and find that it works very well with CS5.
Conversion from HDCam and AVCHD is fast, and once in the CF codec, editing is a breeze without any worry about what the original acquisition formats were. Real time previewing is very high quality
Plus there's the fact that editing in a 12 bit, 4:4:4, near lossless codec will maintain excellent visual fidelity throughout the post workflow, all the way to final delivery formats.
Additionally, the included Cineform First Light software allows you to do metadata based non destructive color correction, as well as graphic overlays, that do not require rendering to preview.
CS5 seems to have made great strides towards smooth editing in native codecs, but I remain very happy with Cineform.

Robert has offered a great overview of the Cineform environment.

The only thing I would add is the fact that Cineform makes a very nice archival format for masters.

For the client, all of mine are set up with the free Neo player so I can send them very high quality previews, or in some instances, they can even use them directly in their presentation work.

Many acquisition formats don't lend themselves to editing or output. Cineform fills that need very nicely, and as Robert notes.... First Light is a very thick layer of icing on that cake.

I too have used cineform for years but haven't upgraded since v3ish i think. It worked well on CS3 for what it did. I suspect the upgrade to v5 for a couple of seats is going to be quite expensive and that has to be balanced against nvidia hardware costs.

I happen to like the idea of staying in original source footage wherever possible and this discussion is really about whether that is truly possible in CS5 as claimed. Perhaps it's not and i should reexamine the cineform route.

It's also about what gets accelerated with cuda, both now and in the future. In fact it's actually about understanding premieres pipeline a bit better in order to see how to be most efficient i guess.

For example i believe all codec decoding is done by the host CPU. Some effects are accelerated but i don't know which ones. Certainly magic bullet looks doesn't seem to be (at least to real time viewing)

Are there any whitepapers on the mercury engine?

Update: Actually just checked cineform and the upgrade wasn't anywhere near what i thought it was, so have just bought it and will see how that goes. However i am very interested in whether getting an nvidia card is going to help me that much?

cheers
paul

If you will be editing with CS5 and Cineform v.5, it is probably not very important to have a CUDA card on board. The Cineform RT engine does most of the heavy lifting, and the CFHD.avi does not require a lot of resources to decode/ preview.
I believe that the benefits of CUDA and MPE relate more to editing in the native formats (like AVCHD).

Correct. The major benefit from using the MPE is that it offloads the calculations associated with selected effects and transitions from the CPU to the GPU. Those effects and transitions are explicitly labeled as "MPE capable" in Premiere Pro. If a codec doesn't use CUDA (such as CFHD), there is little additional benefit from using CUDA for decoding and encoding. But the whole point of the MPE is to balance the computational load between the CPU and the GPU, giving you the smoothest previewing.

Robert,

It seems the cineform blog confirms your thoughts. Cineform isn't cuda accelerated however any appropriate effects on top that could be accelerated are.

At the moment though i use magic bullet looks a lot and so that isn't cuda accelerated either.

So i shall forgo the nvidia card for a while, although if i can find a way to speed up looks that would be useful...

so it seems i'm back to cineform...

cheers
paul

Is there a list somewhere of effects that are compatible (adobe and 3rd party?)

(i'm nowhere near my premiere machine at the moment otherwise i'd look through them...)

cheers
paul

When I use my older magic bullet editors software with Premiere Pro CS3 the effects are near realtime when I check the box that says GPU accelleration. However, now that i am in CS5 I had to upgrade to Magic Bullet Looks 64bit and it has a completely different interface (which I don't care for just yet) and it doesn't appear to be accelerated at all by the video card. So 3 steps forward for Adobe accelerated effects and 2 steps back for Magic Bullet as it takes forever to render....again. :(

It's been a while since i used editors as i was running looks on CS3 too. I don't think there's that much difference between looks on CS3 and CS5 and i believe it has to use the video card otherwise it simply doesn't work.

You may find that looks is doing more than editors

But a valid question is how to speed looks up - there's something very pleasing about the way that it deals with corrections, despite the painful interface and a few functional issues.

There isn't a magic bullet forum on here is there?

cheers
paul