Welcome to our community

Be a part of something great, join today!

  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

New Mac Pro

Your first clue is in your post, if you read back through the thread the consensus is the sweet spot is 8 if you're doing lots of AE/resolve and premiere. For some tasks the 12 core is actually slower due to clock speed.

Again, Jeff did a really awesome job explaining a lot of this.
 
Last edited:
I have some further observations concerning the New Mac Pro and playing back RedCode.

I have had the 6 Core, (32 gigs of Ram and D700s) for about two weeks now and these are my conclusions.

It can certainly play back Epic X 5K under "better performance" without any obvious strain. It wont play back at full resolution without significant stuttering so forget about that.

The bottle neck playing back full resolution is clearly the CPU.

There is one more important conclusion. You cannot address this bottleneck by joining the RAID array madness that seems to have broken out recently and getting a RAID 5 etc. Ordinary bog standard 7200 RPM drives are fast enough to play redcode (certainly for what I do) quite easily. Apart from simply working out the numbers using redcode data rates, you can establish this by putting some red footage on the internal hard drive which reads at about 1000 MB/sec. You get exactly the same kind of editing performance playing back original red footage as you do reading straight off a hard drive with sequential read of about 120 MB/sec.

If this kind of system is not speedy enough for what you do and there are certainly people on this board who would fit this category, you really need to invest in a proprietary card sold by red.

Just my observations.

Best

Rob Castiglione
 
Apart from simply working out the numbers using redcode data rates, you can establish this by putting some red footage on the internal hard drive which reads at about 1000 MB/sec. You get exactly the same kind of editing performance playing back original red footage as you do reading straight off a hard drive with sequential read of about 120 MB/sec.

Yeah but wouldn't a lot of the bandwidth of the internal drive be used by running the OS and whatever program you're using to play the footage back?
 
Rob, I'm confused, you're saying it plays back without obvious strain, then you say it won't play back without stuttering. Care to elaborate?

I can do full 1:1 playback off my RAID no problem. I do have to give it 2-3 seconds to start buffering but it's not having an issue with playback.
 
In addition to my previous post, i will indeed curious to see a head to head comparaison with a hackinotsh dual xeon running a Titan Nvideo GTX card next to the most powerfull macpro. For pure rendering , i know the dual xeon will crash to the macpro, wondering what will be the difference now with a single Titan Vs Dual Fire D700....?
 
Watched the Video. Not sure it was aimed around this industry.

You guys should also have a look at this one:
http://www.youtube.com/watch?v=wfz_KAFPml8

Rob : Better performance is aimed at making (hopefully minimal) video quality compromises to hopefully make playing a video more fluid.
If you need to edit a video that reads 2 x streams, and writes 1 x stream - that 120MB/s could create an additional bottleneck to your workflow.

The use of RAID 5 is actually to guard against data loss by spreading N drives of data across N+1 drives.
Without resilience - when a drive fails, you will loose data.

AJ
 
The bottle neck playing back full resolution is clearly the CPU.

Provided you don't have any other bottlenecks in the chain, yes. When helping others solve playback issues, I still see more storage bottlenecks than anything else. We also have to factor in that there is no fully-optimized R3D playback solution out there right now. REDCINE-X is the only R3D playback system that utilizes GPU decoding right now. And while it is utilizing GPUs, it pretty much does little to use the multiple CPUs/cores available in a system -- it's multithreading is not great by any means. Premiere and Resolve do a far better job on the CPU usage front, but don't use GPU acceleration for the decoding of R3D data, only for the acceleration of image processing tasks.

Ordinary bog standard 7200 RPM drives are fast enough to play redcode (certainly for what I do) quite easily.

I disagree. Few, if any, 7200rpm HDDs out there can actually sustain 120+MB/s across the entire drive volume. Most tend to slow down well under 80MB/s once you pass the half-way point. Solid state storage doesn't suffer from this, but has its own share of issues. Fast storage is paramount and the R3D I/O system implemented by RED likes to do multiple reads on the data as far as I can tell, for doing 5KFF clips, I would say the minimum is 150MB/s sustained for proper playback for each clip or stream you're working with. If you're working with EPIC footage and trying to seriously edit or composite with it, you're essentially jerking off if you don't have storage that can sustain 500+MB/s. That's entry-level, IMO.

If this kind of system is not speedy enough for what you do and there are certainly people on this board who would fit this category, you really need to invest in a proprietary card sold by red.

Yep. And there are other solutions as well -- like a more mainstream PC that can hold additional processors/ cores, more RAM, etc.. I'm really liking the new Mac Pro, though. I've played with a 6-core model (with D500 GPUs) and it just didn't quite have enough oomph to really drive it's way through EPIC 5K+ footage in a way that I would like. It could work fine for many though. People who are comfortable working on a Macbook Pro retina or iMac will think it's a huge step up in computing power, and it is. OTOH, there are times my 8-core feels a bit lacking. The 12-core would be better when I'm in Resolve, possibly for Premiere, depending on what I'm doing. OTOH, the 12-core takes a 10% reduction in CPU speed and 90% of the time we don't engage all 12 cores. I like my HP Z820 16-core and my dual 10-core Xeon for heavy lifting, but they have their own issues too -- big, loud, cost more money to operate, run Windows (hehe)...
 
this is why its so interesting to watch this play out. lots to take account for no matter what you buy as far as trying to optimize our workflows for our various apps. one thing i like is the fact tb2 and usb3 unites the rMBP line accessories with the nMP accessories... i think once this 4k support thing is sorted out i'm gonna pull the trigger on something..

Provided you don't have any other bottlenecks in the chain, yes. When helping others solve playback issues, I still see more storage bottlenecks than anything else. We also have to factor in that there is no fully-optimized R3D playback solution out there right now. REDCINE-X is the only R3D playback system that utilizes GPU decoding right now. And while it is utilizing GPUs, it pretty much does little to use the multiple CPUs/cores available in a system -- it's multithreading is not great by any means. Premiere and Resolve do a far better job on the CPU usage front, but don't use GPU acceleration for the decoding of R3D data, only for the acceleration of image processing tasks.



I disagree. Few, if any, 7200rpm HDDs out there can actually sustain 120+MB/s across the entire drive volume. Most tend to slow down well under 80MB/s once you pass the half-way point. Solid state storage doesn't suffer from this, but has its own share of issues. Fast storage is paramount and the R3D I/O system implemented by RED likes to do multiple reads on the data as far as I can tell, for doing 5KFF clips, I would say the minimum is 150MB/s sustained for proper playback for each clip or stream you're working with. If you're working with EPIC footage and trying to seriously edit or composite with it, you're essentially jerking off if you don't have storage that can sustain 500+MB/s. That's entry-level, IMO.



Yep. And there are other solutions as well -- like a more mainstream PC that can hold additional processors/ cores, more RAM, etc.. I'm really liking the new Mac Pro, though. I've played with a 6-core model (with D500 GPUs) and it just didn't quite have enough oomph to really drive it's way through EPIC 5K+ footage in a way that I would like. It could work fine for many though. People who are comfortable working on a Macbook Pro retina or iMac will think it's a huge step up in computing power, and it is. OTOH, there are times my 8-core feels a bit lacking. The 12-core would be better when I'm in Resolve, possibly for Premiere, depending on what I'm doing. OTOH, the 12-core takes a 10% reduction in CPU speed and 90% of the time we don't engage all 12 cores. I like my HP Z820 16-core and my dual 10-core Xeon for heavy lifting, but they have their own issues too -- big, loud, cost more money to operate, run Windows (hehe)...
 
Just took delivery of three of six ordered.
We have the D500 6-Cores with 32GB RAM first.
I will do some tests with a ROCKET in a Magma TB chassis tomorrow.
Ill do the basics with REDCine and BMDs Disk Speed app.
Anyone have any tips on what to test for?

D700/12-Core/64GB will be in soon but Ive read enough tests on those online.
Kinda wished we ordered the 8-Cores now.
 
What's "enough",and what's overkill?

What's "enough",and what's overkill?

I am of the opinion that the D700 GPUs offer more per dollar than any other upgrade from the base models. While I look forward to Eric's findings, I'm much more keen to see a comparison of:

8 core/32G RAM/D500
6 core/32G RAM/D700
8 core/32G RAM/D700

These are the 3 configs that I project will be the best bang for the buck for running apps RedUsers typically use frequently like RCX, Resolve, FCPX, AdobeCC, Scratch, etc with R3Ds, ProRes, DNxHD, DPX stacks, TIFFs, etc. Speciality use cases excepted.

Sure, a loaded 12 core with D700's/64GB RAM will kick ass and a quad core with D300s/16GB RAM will bog down if you throw too much at it. I don't think you need a pocket protector to figure that out. I'm more interested in comparing configs at a more attractive price point than the max BTO, but still beefy enough to hang in our world.

Cheers - #19
 
Well, I can only relay my own personal actual experience Jeff. I am working just fine editing with no RAID with 4 and 5 K files. It seems fluid and I am getting sustained data rates even for drives which are about 75% full of over a 100 MB/sec - plenty for most editing tasks with Redcode. I think that a lot of people are investing heavily and unnecessarily in devices that really are only needed for specialist purposes. It is a kind of "maximiser" instinct that quite a few have.
 
Well, I can only relay my own personal actual experience Jeff. I am working just fine editing with no RAID with 4 and 5 K files. It seems fluid and I am getting sustained data rates even for drives which are about 75% full of over a 100 MB/sec - plenty for most editing tasks with Redcode. I think that a lot of people are investing heavily and unnecessarily in devices that really are only needed for specialist purposes. It is a kind of "maximiser" instinct that quite a few have.

That's great if it's working for you. Likewise I'm just reiterating my own experiences as well. The problem with single drives is they're really only good for single 5K clips, possibly two 4K clips if the compression is high enough. 5KFF EPIC-MX clips at 5:1 compression need 130MB/s+ throughput, regardless of which decode level or quality you play them back at. You want to do a real-time dissolve or fade between two of those, it's not going to happen reading them both off the same 7200rpm single HDD. Not unless they're small and already ponied up for access and tagged for cacheing. Or you face taking a breather while the system renders.

I agree that a lot of people probably do over-invest and have a maximizer instinct or attitude. I can also attest that more or bigger is not always better. But I also can't work off of individual 7200rpm hard drives, not even the fastest latest and greatest ones out there. And I'm only doing corporate and commercial work, moderate graphics, etc.. These days a 4 to 6 drive RAID is reasonably cheap and provides many benefits; primarily speed, capacity and enhanced data integrity. Not only that, but the ability to read and keep the drive caches full and have enough speed to flush that cache and load it with other data fast enough to keep ahead of what's going on in the software is a great asset. All things you can't do with individual HDDs.
 
Cheers Jeff. I certainly have the "maximiser" instinct in spades. I was just about to invest in an Areca array but really just thought - let me test this proposition through empirical experience. For my work flow with FCP X, I was somewhat astonished myself to find that the drives I had were more than sufficient. I was really looking forward to getting that Areca ! This is not to say that I will not need a RAID for specific jobs. However, for the bulk of the work I just dont.
 
For your work, if that's working for you, that's awesome. :) Really. Any time you don't have to spend the extra money on certain hardware that will eventually fail or become obsolete, you're money-ahead on that.
 
Your first clue is in your post, if you read back through the thread the consensus is the sweet spot is 8 if you're doing lots of AE/resolve and premiere. For some tasks the 12 core is actually slower due to clock speed.

Again, Jeff did a really awesome job explaining a lot of this.

Premiere ad AE perform better with Red on the 12 Core versus the 8 Core. I have tested this time and again. If a 6 Core at 4.5GHz cant come close to the 12 Core at 2.7/3GHz then there is no way the 8 Core can outperform because GHz. The 8 Core is just cheaper so it may be the better investment for many. However the 12 Core still has a far higher realtime Performance ceiling. That is why you have to specify when the pricing of the 8 Core out ways the performance of the 12 Core. Keep in mind the aspects that the higher clock speed do perform better would also perform far better on a 6 Core at 4.5GHz than the 8 Core.
 
As you yourself said in an earlier post it provides a higher ceiling, not necessarily practical speed. As it depends on the workflow and application. Again which is why a lot of folks have settled on the 8 as the trade off. Once we get that tie in between more CPUs and GPUS things will start getting crazy good.

I have a feeling it'll be a little while before we see that optimization though.
 
D700/12-Core/64GB will be in soon but Ive read enough tests on those online.
Kinda wished we ordered the 8-Cores now.

Just got the 12-Core (1 of 3) but no time to play.


The D500/6-Core ($4800 CAD each)rocks in REDCine at 1/2 but with files off the internal 500GB (BMD Disk Speed Test 907/891MBs).
That's with full 4K. On FULL it plays but needs to pause every three to five seconds.
I get the same results with a SanDisk SSD 480GB in a Sonnet Tempo SSD PCI card housed in a Magma ExpressBox 3T.
However under BMD Disk Speed Test I only get 492/219MBs.
This workstation would handle nicely on set if needed.
Its deemed for photography/graphics use.
Should be on the loaded nMP soon for testing :)

Currently using an Apple Thunderbolt Display....sigh....
 
Back
Top