Pro Tools - AVE performance test results

[marsian]

Hello fellows,

I did a lot of tests with various PT versions and video hardware.
I tested on a 12core 5,1 MacPro with 24GB and a Radeon 5770.
PT hardware: HDN PCIe, OMNI, SYNC HD.
Videohardware: BM Decklink Extreme 3D+, AJA Kona LHi
Syncmeasurement: CatchinSync
The Testsession was a regular stereo 30 min. TV-doc.

Best combinations were:

AJA Kona LHi V.12.5
OS 10.10.5
PT 11.3.2
Video-Offset: +5qf
Sync H.264: +/- 0qf
Sync DNxHD: +/- 0qf
CPU usage: ca.30%

BM Decklink Extreme 3D+ V.10.2.3
OS 10.10.5
PT 11.3.2
Video-Offset: +6qf
Sync H.264: +/- 3qf
Sync DNxHD: +/- 0qf
CPU usage: ca.20%

BM Decklink Extreme 3D+ V.10.2.3
OS 10.10.5
PT 12.7 with AVE 12.4
Video-Offset: +6qf
Sync H.264: +/- 3qf
Sync DNxHD: +/- 0qf
CPU usage: ca.25-30%

BM Decklink Extreme 3D+ V.10.6.8
OS 10.12.2
PT 12.7
Video-Offset: +2qf/+6qf
Sync H.264: +/- 1qf
Sync DNxHD: +/- 2qf
CPU usage: ca.25-40%

In general, audio and video on different drives showed ca. 5-10% lower CPU usage.
A SSD audio/video-drive showed a 5-10% lower CPU usage to a spinning drive.
When the videotrack in PT was set to a different resolution than the original videoresolution, CPU usage was 2-times more than with the right resolution.

BM Decklink performed as well as the AJA Kona, as long as I used a specific driver. All newer drivers were really bad.

I hope this helps when choosing systems. Please let me know your experiences.


Cheers
Maik

[Frank Kruse]

Hi,

was this with or without genlock?

Frank.

[MIKEROPHONICS]

No disrespect, but not sure I would be trusting video measurement to Catchin sync. It is too blunt an instrument, and can not measure drift only static start stop play passes.
Buy a SyncOne2 and measure to millisecond accuracy. It is less than the cost of the time you spent to do these tests I suspect.
If you are genlocking your system the system can only move video by whole frames.


Sent from my iPhone using Tapatalk

[marsian]

Hi Mike,

no problem. I'm aware than Catchinsync ist not the most acurate tool and that you can only really trust sync when genlocked. But my intension was to see relative numbers between different hardware and software without genlock. Especially to see the differences in (CPU) performance.
It was a surprise, that with the right drivers, BM is not far from AJA, or performance wise, even better (to be fair, I only tested the latest AJA version).
This tests are no exact science, but I think it helps a bit to find the right combination.


Cheers
Maik

[lode]

Thanx!
What excactly is the "+/- 3qf" mean?
Does that mean imprecise playback? That the playback would vary up to "+/- 3q"?

[Vedat]

(don't mean to hijack the thread)

Originally Posted by MIKEROPHONICS

No disrespect, but not sure I would be trusting video measurement to Catchin sync. It is too blunt an instrument, and can not measure drift only static start stop play passes.
Buy a SyncOne2 and measure to millisecond accuracy.

Have you used it? With iPhone 6 and later, it records at 240fps. That's a frame about every 4ms. If you're competent with your fingers, you can get well within 1/4f accuracy.

This is more than enough to get the job done and move on.

Besides, with Catchin Sync, you don't need test videos for every environment/codec you are in, just use the countdown or 2 pop on the video you have, or even an event in the action (like a slap, ball hit etc.)
You can even put a pop at a frame boundry within the video if there is no other reference available for a quick sanity check.

You can't do that with the hardware device.

[MIKEROPHONICS]

But you are only measuring half of the story and not any perturbation - which the BM drivers suffer badly
3 qf is in all likeliness the error or measurement as the measuring methodology is no where near robust enough. When measuring scientifically your accuracy of measurement must be orders of magnitude better than what you are trying to measure, or the measurements are just error or noise.
Sorry but these results are to all intents and purposes little better than speculative and irrelevant, despite noble intentions.
Video cards ought to be able to be accurate to the frame on each play start pass (genlockable), with little perturbation (as long as you are not using a silly codec like h264 (which is akin to driving on ice).



Sent from my iPhone using Tapatalk

[simonchase]

I found the BM solid in terms of drift using the syncheck hardware. Again assuming the correct firmware.

But in terms of errors and crashes the AJA performs significantly better.

[MIKEROPHONICS]

I found the BM deck link studio would have a greater span than the AJA, and be less responsive too. But it was a while ago since doing my tests for PTE. Using the the AJA IOXT I was hooked on the response after three presses of the space bar, that combined with the fact that support at AJA is actually available and helpful and that there wasn't Russian roulette with each release of drivers.


Sent from my iPad using Tapatalk

[marsian]

Mike,

I wonder why you insist so heavy on your opinion. Did you test BM with any of the drivers/OS combinations that I meant? It makes all the difference. As simonchase told us, sync doesn't drift. The question is: simonchase, what was your combination?
Also, when the BM is genlocked, it doesn't drift for shure, right?

@lode: it means that sync varied about 3 quarter frames.

I was about to change our systems to AJA, because it didn't feel snappy enugh and we got to much errors. But with the right combo it's as good as an AJA. Thats the point.

Cheers
Maik