Trasncode times with MetaFuze v1.3

[MichaelP]

OK... not an official test yet, I did get a little bit of time on an HP Z800 workstation. I installed MetaFuze v1.3 and transcoded a single 01:30:00 R3D file to DNxHD. The R3D was shot 2K/16:9. Storage subsystem was a raided 16bay SAS. The file had no audio.

Full debayer: 04:30:00
1/2 debayer: 01:11:00 (faster than real time!)

Benefits also include native MXF wrapped DNxHD so editing can start right away and not have the additional "fast import" step. Consider this a sort of "off the record" test, but things look quite promising. It was great to see all processors slammed at near 100%. Kudos to the Avid DS engineers!

Michael

 

[Igor Ridanovic]

That's roughly 3X faster than what I was projecting would be attainable in MF 1.3 on an 8 core XW8600 workstation. Excellent speed. Was this "1/2 good"?

 

[Seán_T]

Now were talking!

I didnt know Metafuze was the DS team, I suppose it makes sence now.

 

[Chris Parker]

whoa baby.

 

[filip kovcin]

to Michael P.

to Michael P.

Michael,
just two questions:

- can you confirm that this version (1.3) has batch support also?
- did you check this version also via bootcamp, just to compare non official transcoding results?

thank you,


filip

 

[MichaelP]

The 1/2 debayer was "1/2 good" - I was testing for quick dailies for offline work. I will be testing the same files on a comparable Mac system under bootcamp and VMWare.

Michael

 

[MichaelP]

Also.. I will be testing on my 8600 at home for a more common middle of the road system on the same file.

Michael

 

[Jason Diamond]

good to know michael...

 

[MichaelP]

Again, a very quick test, but transcoding the same 01:30:00 file on my "older" HP 8600 workstation with 4GB RAM off a G-Tech eSATA RAID, the "1/2 Debayer Good" took 02:30:00 to transcode - 1.6x longer than real time.

Michael

 

[MichaelP]

- can you confirm that this version (1.3) has batch support also?
- did you check this version also via bootcamp, just to compare non official transcoding results?

Yes, v1.3 has batch support. I have not tried bootcamp yet, those tests are in the plan.

Michael

 

[Chris Parker]

if possible, can you use a 4K source clip as the basis for these tests? 2K at 1/2 good may not be fully representative of the 'standard' needs of metafuze users.....

thanks michael. keep up the good work. we all look forward to 1.3 release soon.....

 

[MichaelP]

I agree - it was what I had on my drive at the time. I will do real tests with 4K media. Thus the "unofficial" status of these tests.

Michael

 

[Chris Swartz]

Is the render completely CPU dependent? Or does the GPU play a part as well?

In order of importance what speeds up the process more?

Chris

 

[MichaelP]

All CPU at this time - drive speed and RAM are also important.

Michael

 

[Chris Parker]

does this mean that metafuze won't take advantage of the red rocket when it comes out??

 

[MichaelP]

Unknown at this time - we haven't seen the SDK that talks to the hardware. We all found out about Red Rocket at the same time. No reason it can't - just more development.

Michael

 

[Igor Ridanovic]

Drive speed is of great importance with any RED workflow. MetaFuze or not. I get much better performance debayering when source material is on a SAS raid than firewire 800.