Welcome to our community

Be a part of something great, join today!

Log in
  • 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 :)

Do I really need a RED ROCKET X for DRAGON ?

M

Mert Icgoren

Member
Joined
Mar 5, 2014
Messages
18
Reaction score
0
Points
0
I've been going back and forth on this..

Is it better to spend all that money on a new MAC PRO(aka trashcan) or the Rocket X card ?

Im mostly shooting commercials and/or short content.

How has your experience been with the Rocket X ? what would tou suggest ?

thanks in advance !
 
If you can scrape by with half debayers and longer encode times go for a trash can. But my rMBP in resolve on epic 4khd footage was transcoding full debayer at 100fps from a single drive to a single drive, no raids. That is damn impressive behavior for a laptop, plus given the long history of the former rocket I thought the rocket x would be a good choice since it will hopefully be a viable rental unit for 5 years or so. At the discounted rate rocket x seems like a decent solution versus the fully kitted trash can, plus on my non red jobs the laptops mobility is very handy i think
 
We are using new MacPros to edit our dragon footage natively in Premiere Pro CC without any problems. If we edit in a 1080 timeline, there are zero delays and it plays back full res/real time. Even in higher-res timelines it's still just about perfect. No different than editing ProRES in FCP 7. We have 8-core NMPs with dual D700s.
 
Im using a tb2 enclosure from Magna, its pretty sweet. The trash can is more rounded in that it can do okay with any format, the rocket x though with a laptop hauls plenty fast for me
 
Mert,

It really depends on how fast you need the footage turned around for transcoding. If you're getting the 12 core version of the Mac Pro with dual D700's, you'll likely get good turnaround time on it. It won't be realtime, but it won't be bad either. Someone else that has one handy may be able to tell you more precisely how fast it can cut through footage.

If you need at least realtime transcoding though, then go for the Rocket-X.
 
I have been running a few tests testing the Rocket X on different machines and thought it might interest this thread. It basically shows that the Mac Pro without a Red Rocket X card transcodes at less than half the speed of a Macbook Pro running with the Red Rocket X. These times scale considerably with larger files and when you introduce other components besides the GPU, such alternating the source and destination. Even when you transfer to and from the SSD of the Mac Pro, its still slower at roughly 1/5 the time of transferring to and from the SSD in the Macbook Pro while using the Rocket X. Even more interesting is that even when you use the Rocket X with the Mac Pro, the times are not much faster than the same type of transcode on the Macbook Pro Retina with the Rocket X as well. Obviously this is comparing relatively stock machines and not anything heavily optimized with customized hardware. As such, its just supposed to give an overall idea of Red Rocket X transcode speeds. Dragon footage would obviously have comparable results based on these machines, even though this is all Epic 5K.

Explanation of times breakdown by differing variables

*external to external, times are from and to a 5400rpm HDD through USB 3.0
*external to internal, time are from a 5400rpm HDD to stock Internal drive of computer (stock drives are the at very bottom under specs by model)
*internal to internal, times are from stock internal to stock internal of computer

*In addition, times are varied with and without Rocket for certain models to give further comparison


Red Rocket X Transcode Test
All footage transcoded to ProRes 422 HQ 23.98
Footage shot on Red EPIC

1) Original File: 8.54GB, 5K FF, 23.98, Duration 00:1:26 (HH:MM:SS)

Time to Transcode with different variables:

Macbook Pro 17:
w/Rocket-X (00:02:17 *external to internal)

Mac Pro:
w/o Rocket (00:04:25 *external to internal)
w/ Rocket (00:00:54 *external to internal)

Macbook Pro 15.4 Retina:
w/Rocket (00:02:24 *external to external)
(00:01:15 *external to internal)
(00:00:44 *internal to internal)


2) Original File: 38.7GB, 5K Full Frame, 23.98, Duration 00:06:34 (HH:MM:SS)

Time to Transcode with different variables:

Macbook Pro 17:
w/Rocket (00:10:37 *external to external)

Mac Pro
w/o Rocket (00:23:00 *external to external)
(00:16:23 *internal to internal)
w/Rocket (00:10:10 *external to external)
(00:05:31 *external to internal)
(00:02:15 *internal to internal)

Macbook Pro 15.4 Retina
w/Rocket (00:10:47 *external to external)
(00:6:01 *external to internal)
(00:03:19 *internal to internal)

*Specs for compared models:

2014 Macbook Pro 15.4 Retina
(2.8GHz Quad-core Intel Core i7, 16GB 1600MHz DDR3L SDRAM, 1TB PCIe-based Flash Storage, Intel Iris Pro Graphics and NVIDIA GeForce GT 750M with 2GB of GDDR5 memory)

Late 2013 Mac Pro
(3.5GHz 6-core with 12MB of L3 cache, 16GB (4x4GB) of 1866MHz DDR3 ECC, 256GB PCIe-based flash storage, Dual AMD FirePro D500 GPUs with 3GB of GDDR5 VRAM each)

Late 2011 Macbook Pro 17
(2.4GHz quad-core Intel Core i7 processor with 6MB shared L3 cache, 16GB 1600MHz DDR3L SDRAM, 750GB 5400rpm Serial ATA HDD, AMD Radeon HD 6770M graphics processor with 1GB of GDDR5 memory)

-------------------------
Yair Alon Leventhal
DIT
yleventhal@gmail.com
 
Last edited:
Latency is probably the biggest factor in the transcode times listed above. Most 7200 rpm drives have a latency of 8 to 12 MS while most SSD's have a latency of .1 to .4 MS, and newer TB2 controllers also seem to have less latency than older TB1 controllers.

Spinning media is probably the main bottleneck of any high end computer system in use these days, next after that is probably RAM and/or the GPU.
 
Shiblon Wixom said:
Latency is probably the biggest factor in the transcode times listed above. Most 7200 rpm drives have a latency of 8 to 12 MS while most SSD's have a latency of .1 to .4 MS, and newer TB2 controllers also seem to have less latency than older TB1 controllers.

Spinning media is probably the main bottleneck of any high end computer system in use these days, next after that is probably RAM and/or the GPU.
Click to expand...

Yes agree, to utilize the rocket X you kind of need fast read write and normal solo HDD's will for sure be whats holding it back. Personally I think we do more than fine transcoding over normal graphic boards. Realtime is great, but getting about 20fps is, for me atleast, no big difference when creating proxy files for all the shot source files. Normally I put it to render late at night after a days shoot and before I wake up it's all done. If it was done 20 minutes faster with a rocket X it would not really make much difference.
 
Shiblon Wixom said:
Latency is probably the biggest factor in the transcode times listed above. Most 7200 rpm drives have a latency of 8 to 12 MS while most SSD's have a latency of .1 to .4 MS, and newer TB2 controllers also seem to have less latency than older TB1 controllers.

Spinning media is probably the main bottleneck of any high end computer system in use these days, next after that is probably RAM and/or the GPU.
Click to expand...

Agreed, thats why the transfer times from an internal SSD to the internal SSD show considerable speed increases. However, this kind of transfer is still considerably slower than using a Rocket X and transferring to a slow 5400 rpm external spinning drive. As such, there is a sizable speed difference to having a Rocket X, even more so if you consider SSD in the mix.

Björn Benckert said:
Yes agree, to utilize the rocket X you kind of need fast read write and normal solo HDD's will for sure be whats holding it back. Personally I think we do more than fine transcoding over normal graphic boards. Realtime is great, but getting about 20fps is, for me atleast, no big difference when creating proxy files for all the shot source files. Normally I put it to render late at night after a days shoot and before I wake up it's all done. If it was done 20 minutes faster with a rocket X it would not really make much difference.
Click to expand...

Another thing to consider is that these times are for a single 38GB file. A typical commercial I work on tends to average 500GB for a single cam shoot at which point ten minutes more scales to hours quicker with the Rocket X. On a movie, with less turnover times between dailies receipt, and more footage shot the time increase becomes more crucial. In general, I work onset and need to have the transcoded files created close to real time and ready to go on wrap for post. As such, theres no question that 10 minutes or an hour, that speed increase is a major necessity for my workflow. However, if you have time for night transcodes then you might be able to get away without a Rocket X and just a good workhorse machine.

-------------------------
Yair Alon Leventhal
DIT
yleventhal@gmail.com
 
Yair Leventhal said:
Agreed, thats why the transfer times from an internal SSD to the internal SSD show considerable speed increases. However, this kind of transfer is still considerably slower than using a Rocket X and transferring to a slow 5400 rpm external spinning drive. As such, there is a sizable speed difference to having a Rocket X, even more so if you consider SSD in the mix.



Another thing to consider is that these times are for a single 38GB file. A typical commercial I work on tends to average 500GB for a single cam shoot at which point ten minutes more scales to hours quicker with the Rocket X. On a movie, with less turnover times between dailies receipt, and more footage shot the time increase becomes more crucial. In general, I work onset and need to have the transcoded files created close to real time and ready to go on wrap for post. As such, theres no question that 10 minutes or an hour, that speed increase is a major necessity for my workflow. However, if you have time for night transcodes then you might be able to get away without a Rocket X and just a good workhorse machine.

-------------------------
Yair Alon Leventhal
DIT
yleventhal@gmail.com
Click to expand...

Having a rocket is better than not having one, that is for sure. same thing goes for super fast raids, CPU's and everything else that builds a faster in the pipeline. Simply put there is no graphic card to date (that I know of) that chew r3d's faster than the REDROCKET X so if maximum possible speed is the only concern and the graphic board is the bottle neck, yes then it's simple by a few rocket X cards. But again, if your client wants big fat proxy files delivered on duplicates of single spinning discs then most likely the graphic cars is not the bottle neck the disc speed will be.

Question what kind of proxies do you normally deliver? only low debayer/high compression for offline editing or full proress 444 full debayer? As we work low debayer proress LT is more then fine for editing and then we send r3d trims to grading. But yes I guess you can make such files even faster than realtime with a rocket X, still a normal graphics card does it pretty fast as well.
 
My Dragon decoding times:

6K 5:1, 1 minute duration clip, decoding to 1080p half good debater (for editorial)

On a 2011 Mac Pro 2 x 2.4 GHz 32 GB ram, 2 Titans Black on Cubix: 2'20"
On a Hackintosh i7 3.5 GHz 16 GB ram, 1 Titan Black: 3'15"

Curious about mac vs pc decoding performance and SuperMicro / Z820 configurations. After 5 rockets cards and only 1 working, I'm reluctant to a Rocket-X...

Greetings
 
You must log in or register to reply here.
Back
Top