[Nat Jencks]

Let me clarify the question. Forgetting the synology NAS for now, are you saying that if I connect two mac pros (either 2010 or 2013) to a netgear gs108t. And setup link aggregation on both machines and the switch I should or should not be able to do AFP and SMB file sharing between the two machine at ~180MBps? That’s what I’m looking for. Not really looking for full fledged shared storage, just faster file sharing between the machines. I have one system as the main system and the other as an assistant station.
Best-
-Nat

[Nat Jencks]

Yes its true that a single decopress/recompress cycle for prores4444 isn’t too bad, but if one does this process in FCP it is literally lossless since it doesn’t go through any decompression decompression. I’ll use DPX or do in joining in FCP.

Thanks folks!

-Nat

[Nat Jencks]

Rendering a new “Long Play” quicktime e.g. an assembly of the reels.

If this process is done in FCP and picture is untouched the software knows that there is no need to decompress/recompress and the new file is identical / lossless as it does not go through a decompress / recompress cycle.

best-
-Nat

[Nat Jencks]

I’m curiosu what others think here, my gut tells me that desktop drives with more redundancy is a better solution than enterprise drives with less. In other words, I’d go for a 8 drive RAID 6 with desktop drives and opposed to a 6 drive RAID 5 with enterprise discs.

given an identical setup, its always better to replace a desktop drive with an enterprise drive, but if you can provide redundancy via advanced levels of RAID, is the added reliability of the enterprise drive worth it or can you acheive more speed as well as reliability by simple adding more discs.

The above assumes that the physical size and power requirements of the discs are not an issue.

Thoughts?

best-
-Nat Jencks

[Nat Jencks]

Using noise reduction?
OFX Plugins?
Other GPU Memory hogs?

That would be the first thing I look at. Especially with 4 or 5K source some of these things are really memory intensive for the GPU, and resolve doesn’t handle running out of GPU memory during render very gracefully.

If thats not the issue, i’d look for a piece of problematic media. Mark in and out points at halfway marks and “divide and conquer” to isolate the problem.

Just some thoughts.

best-
-N

[Nat Jencks]

Umm.. “spring cleaning” that is, no print cleaning happening here 🙂

[Nat Jencks]

Thanks Sarah, I understand why implementing temporal Noise reduction on the track node might be difficult, but it would be very useful. That said I do understand that for it to work well it would need to look at the handles of each shot and not blend each shot with the next, which might be quite hard to implement,

Thanks for the response!

best-
-Nat

[Nat Jencks]

Not a matter of “fault” hopefully, mike I agree this could easily be something wonky with the newest red SDK, but hopefully this thread is about finding a solution to this very specific bug, not assigning vague “fault”. Bmd does a fantastic job of supporting red workflow and hardware and as discussed earlier in the thread the rocket card can certainly be moody and I’m sure it’s no joy to program for it, and a constantly moving target of the red SDK clearly makes things difficult too. That said it does seem clear that something changed in the recent 10.1.4 or earlier point release which broke r3d debayering via GPU or rocket. Whether this is a bug in resolve or in the red SDK is less important to me as an end user than that it get fixed, and the only way that will happen is with further troubleshooting and discussion and feedback to BMD… Which hopefully other folks seeing this problem can provide.

[Nat Jencks]

Sean, this is EXACTLY the behavior I was seeing when using the rocket, so I think its safe to say that this happens with either the rocket OR gpu debayer, but NOT with software debayer.

Since you have tried with CUDA 5.5.28 and the newer 6.0.37, it also seems that it is not directly related to the CUDA driver version.

My guess is that back rev-ing the version of resolve to 10.1.X prior to 10.1.3 would solve the problem, but obviously I’m not sure, and in your case you don’t have a rocket card that doesn’t help you.

Sean, perhaps you can send BMD support your logs… Would love to see this issue solved with a 10.1.5 release or sometime before my next red job! Fingers crossed!

good luck!
best-
-Nat

[Nat Jencks]

It does seem like its an especially moody card.

In this case though it seems to be related to the update in the way Resolve is handling R3D debarring, since it was working fine before (as of 10.1.1) and now doesn’t in 10.1.4.

Hmm. Luckily this isn’t urgent for me as I’m on another project (alexa) at the moment.

-N