Welcome to our community

Be a part of something great, join today!

Register 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 :)

Rocket-X - Time tests

M

Meryem Ersoz

Well-known member
Joined
Apr 12, 2008
Messages
2,561
Reaction score
0
Points
36
Website
www.blackwingdigital.com
I looked around and haven't seen many time tests for the Rocket-X. So I did a bit of my own, to measure the Rocket-X performance.

These are by no means comprehensive tests but are a bit of a starting point.

It would be great if other people running time tests with different specs could add these to the thread, to centralize results.

I used a single 6K clip of 5:30 - my goal was to check on how feasible a 6K workflow is (or isn't) and to figure out best practices with what I have.

For all tests, my Rocket-X was housed in a Sonnet Chassis. I also tossed in an 480 gb SSD card which I plundered from a tower which I no longer own, to fill one of the lanes. OWC, the card manufacturer, now makes these same cards with 960 gb SSD. I'm tempted.

I found this card to be more invaluable to my workflow than expected. It's great for moving big data around. And copying off to the SSD first makes the transcodes scream....

My two computer systems are pretty maxed out, so these are fairly fast systems. Your mileage may vary.




Clip specs - 6K 2:1, 5:1 compression, 23.98fps, 180 degree shutter, Runtime: 5:30, file size 53gb.

Sonnet Echo Express III-D tested with Rocket-X card and with Mercury Accelsior 480gb SSD PCi-E card

TRANSCODES
MacPro, fully loaded 8 core

No Rocket-X over Firewire 800 - 17:28
No Rocket-X (from desktop SSD) - 17:05
No Rocket-X over eSata using Thunderbolt hub - 17:07


Rocket-X over Firewire - 5:05
Rocket-X over eSata using Thunderbolt hub - 3:55
Rocket-X over Mercury SSD - 2:57

Laptop 2.6ghz Intel i7 Quad

Rocket-X over Firewire - 19:03
Rocket-X over eSata using Thunderbolt hub - 4:26
Rocket-X over Mercury SSD - 3:10

PLAYBACK

Playback over MacPro with Rocket-X in Sonnet box w/SSD maxes at 6K 1/2 res

Playback over MacPro without Rocket-X (eSata via Thunderbolt hub) maxes at 6K 1/4 res
Playback over MacPro without Rocket-X via Firewire 800 - no 6K playback! at any resolution without hiccups - even with loading the clip, cursor caught up halfway - around 2:45 mark

Playback over laptop with Rocket-X in Sonnet box w/SSD maxes at 6K at 1/2

Playback over laptop without Rocket-X (eSata via Thunderbolt) maxes at 6K at 1/8
Playback over laptop without Rocket-X via Firewire 800k - no 6K Playback without stuttering
 
So over thunderbolt 2 to a New Mac Pro we don't get more than 1/2 res realtime playback now? Even though it appears we get faster than realtime transcodes? Any thoughts on why this would be?

I wonder how it would stack up to running on a PC on a full 16x slot.
 
What are you transcoding to? I somehow can't see the specs for the rendered file resolution. The reason why this is so important is that Thunderbolt 2 is very limited for bandwidth. Playback of 6K Full frame at 100% resolution requires a 6K uncompressed image to be sent from RRX to the computer, whereas a 1080p image for rendering is much smaller (about 1/10th the size).

If you did some more tests in a computer with a Pcie Gen 3 X16 slot, you'd find that while the card is capable of 6K full frame playback, TB2 is not without some scaling on the RRX.
 
Thanks for sharing the test results! PS I have to agree that naming the codec and aspect ratio your encoding to (and whether its 6k HD or not) is relative.

On old rocket prores full hd was fairly fast for me, h264 a bit slower, dpx a pain. plus is the source 16/9 or not...stuff like that becomes very relative variables... I dont mean to nitpick and authentically appreciate the test. I posted some results of my own last night as well...6k to 1080 prores 422LT was faster then real time via TB1 on a Mac Air...

Not bad at all! Cant wait until they tweak it more!
 
good point. I was transcoding to ProRes 422 HQ.

Again, these aren't intended to be definitive tests but rather starting point for discussions - and they are based on what I'm most likely to need, for the jobs I'm most likely to be hired for - when there is obviously a much broader range of needs out there. I'd love it if people chimed in with additional benchmarks.

It took me a full day of testing and timing just to shoot a good clip with audio (important to gauge clean audio playback, based on what I've heard about mixed audio results with the Sonnet) and organize these results. I think a thread where a range of folks chime in would help us all.
 
Meryem Ersoz said:
good point. I was transcoding to ProRes 422 HQ.

Again, these aren't intended to be definitive tests but rather starting point for discussions - and they are based on what I'm most likely to need, for the jobs I'm most likely to be hired for - when there is obviously a much broader range of needs out there. I'd love it if people chimed in with additional benchmarks.

It took me a full day of testing and timing just to shoot a good clip with audio (important to gauge clean audio playback, based on what I've heard about mixed audio results with the Sonnet) and organize these results. I think a thread where a range of folks chime in would help us all.
Click to expand...

Thanks Meryem. I assume 1080P resolution transcodes?
 
The playback , on my MBr, maxed 2013, is marginally better, but the transcode, as I have posted many times and , by PM, to many of you, is quite snappy, from 6K to 422 HQ, for example. The Playback acceleration, not worth the card and the display port still on the way, but the transcode speed works a treat. The card is at 2/3 of it's transcode power, as I remember Jared telling me at NAB. The new Cinex really kicked up the speed as well, YMMV.
 
Steve Freebairn said:
If you did some more tests in a computer with a Pcie Gen 3 X16 slot, you'd find that while the card is capable of 6K full frame playback, TB2 is not without some scaling on the RRX.
Click to expand...

I think this is an important point - I certainly wasn't trying to say that we could not get full playback at 6K with Rocket-X, but rather letting people know that my fully loaded 8-core trashcan isn't quite up to it. A lot of people are studying the parameters of what these little machines can and can't do - I had not seen much study of the Rocket in a box coupled with the trashcan, which is what prompted me to run and print the tests. I think this will be a very popular pairing.

But this is really a computer-specific and workflow-specific test. I'd love to see other benchmarks posted....

I'd be interested in testing on a tower versus the trashcan. I think full playback is quite possible with Rocket-X maximized in a x16 slot, as Steve notes. I love my trashcan for its portability. It suits my purposes - also I'm fine with 1/2 res playback in 6K.

I think it's quite remarkable that the little traschcan and a Rocket box can do this!

It's not going to be ideal for everyone, but I'm pretty excited. The trashcan has its limits, but I don't think there is a more portable system as powerful as this one. For the type of work we do in the small Colorado market, where we do a lot of demanding location shooting, it's a very good option. Not necessarily perfect but very, very good. I imagine that, in my market, even though 6K is an option, we will still be doing quite a bit of shooting in 4K and 5K with Dragon. I feel comfortable that the trashcan and Rocket-X will be able to fly through these jobs. I was more interested in seeing how hard I could push it, just to have that in my knowledge base.

The big surprise was the SSD card and the amount of speed enhancement it brought to my 6K workflow. Adding it to my Sonnet chassis was an afterthought. Blair had rightfully advised me to wait and see what innovations emerge in the coming months, as these multi-slot boxes gain traction, instead of spending a bunch of money to fill the lanes. So I just dropped in something that I already owned.

But now I think it is a huge timesaver....
 
Thank you for these tests, it's exactly the info I was looking for. We have the new 12 core Mac Pros as well and ordered Red Rocket-X cards, and I was curious as to the performance increase we could expect. We know that using a PC with a full PCI slot will give the best performance with the Rocket-X, but we weren't willing to move to a Windows environment. The familiarity of working in MacOS on a day to day basis was more important.

I have to say, the 12 core without any Rocket-X with the new Premiere and DaVinci Resolve 11 Beta is working really well. The Rocket-X is just going to make it even better, especially by handling the Red debayer and leaving the GPUs for effects processing and Resolve nodes.
 
System:
Windows 8.1
2x Xeon E5-2690v2
256 GB 1866Mhz ECC RAM
nVidia Quadro K6000
Rocket-X
RedCine-X 30.1.x

footage source and target iofx 1.65TB PCI-E

Source footage: 6K Dragon 9:1 23.98 fps 1:17 runtime

Transcoding to 1080p 422 (HQ) ProRes Mirazon

Rocket-X 2:25
CUDA: 2:29 but footage has black frames flashing in
OPENCL: 2:23 no flashies

Going to try some other codecs to see if that is limiting it.
 
David Kudell said:
I have to say, the 12 core without any Rocket-X with the new Premiere and DaVinci Resolve 11 Beta is working really well. The Rocket-X is just going to make it even better, especially by handling the Red debayer and leaving the GPUs for effects processing and Resolve nodes.
Click to expand...

This is actually really good news - the new trashcan had me contemplating a shift in workflow, in the direction of FCP-X, but if the trashcan is playing well with Premiere and Resolve, then that's probably the best option. I had these on my tower but wasn't sure if they would play well on the trashcan, which I'm still sorting out.
 
System, OS Mac/Win7
Z87, 4770k (quad-core), GTX 770, RAM 16 GB, Rocket-X (x8 PCI-e slot)

6k HD Dragon test clip of 32 seconds
RedcineX = 11 second transcode
Resolve = 45 second transcode

4k HD Red One, 1 hour timeline
Resolve 10-Node Color Grade, avg. 40 fps transcode speeds
Resolve with no CC, avg. 50+ fps transcode speeds

Resolve was suffering from a severe lack of RAM, really needs 32 GB.
Dragon 6k HD would play back at full resolution in Redcine-X at 24 fps and between 20-24 fps in Davinci Resolve, again suffering from lack of RAM.
 
Last edited:
I finally got in my Red Rocket-X!

Here are some test results transcoding to ProRes 4444.

Dragon 6kHD 1 Minute clip.

With RRX and 2x Titan Blacks
Transcode to 4k HD:
1:26

Transcode to HD:
0:30

With RRX only:
Transcode to 4k HD:
1:36

Transcode to HD:
0:36

NO ROCKET, 2x Titans:
Transcode to HD:
1:24

Playback in RCX at full-res is real time, no hiccups.
Playback in DaVinci full-res is real time.

System Specs:
2x E5-2670
64 GB RAM (RCX allowed 30GB)
2x Titan Black
Rocket-X in a 8x Slot
Reading and Writing to Samsung 840 Evo 1TB SSD
 
You must log in or register to reply here.
Back
Top