Matrox - H.264 accelerator board (PCIe 1x)

Nigel Barker · May 23rd, 2009, 12:01 PM

Here is a short clip from my Canon 5D Mk II (41Mbps & 58.6MB in size) converted, firstly with Compressor & secondly with the Elgato device. The file converted in Compressor is PondQuickTime8Mbps.mov while the one converted with the Elgato Turbo.264 is PondElgato8Mbps.mp4.mpg (I had to add the .mpg suffix so that I could attach it to this posting.). The original file was (41Mbps & 58.6MB in size).

William Hohauser · May 23rd, 2009, 07:28 PM

Both files look great. What were the encoding times?

Nigel Barker · May 24th, 2009, 01:22 AM

Quote:

Originally Posted by William Hohauser

Both files look great. What were the encoding times?

I really can't say that the picture quality of one file is better than the other. I didn't time the conversion times but recall that the Elgato gizmo took less than a minute whereas Compressor took several minutes.

Nigel Barker · May 24th, 2009, 01:31 AM

Quote:

Originally Posted by Christopher Drews

CONS:
1) I believe this device is "Single Pass" - see attached screen shot to verify this. There is no setting to choose different like standard H.264. Sizable quality difference between single pass and double pass (of course).

In theory at least but I don't see any difference in quality between video produced by the single-pass Elgato & Compressor multi-pass.

Quote:

Originally Posted by Christopher Drews

2) Flimsy USB - It just feels like a cheap consumer memory key. It tends to rock in the USB slot.

A USB key doesn't waste one of the PCIe slots which are in short supply on the Mac Pro. A couple of graphics cards, an eSATA/RAID card or two & you are out of slots. More importantly it also means that the device can be used on an iMac or MacBook too.

Cheers

Nigel

Andy Mees · May 24th, 2009, 08:39 AM

Those encodes look excellent Nigel, any chance you could post the original clip too for comparison? Also any chance you could post a second or two of high motion footage with similar settings?

Nigel Barker · May 24th, 2009, 10:57 AM

Quote:

Originally Posted by Andy Mees

Those encodes look excellent Nigel, any chance you could post the original clip too for comparison? Also any chance you could post a second or two of high motion footage with similar settings?

I tried but failed to attach it to a posting here so I have uploaded it to Vimeo. You will need register if you haven't already done so & then login to be able to click on the download link on the lower right hand side of the page By the pond on Vimeo

Christopher Drews · May 24th, 2009, 11:29 AM

Here is my test done on my MBP.
Encode times were astoundingly different. I intentionally chose two pass H.264 vs Turbo's single pass. Data rate was set to 1250.

Original was DVCPRO HD, 1080 23.98.

Astounding Time Difference.
Standard took 18:41:991 (min)
Turbo took 22:773 (sec)

The quality difference is evident though.
-C

Andy Mees · May 24th, 2009, 06:21 PM

Chris
I think some of the quality loss you're seeing might be due to some inadvertent distortion? Looking at the Turbo encode there seems to be a significant stretch and subsequent cropping of the image. Also whereas the Compressor encode produced a 459 frame (24fps) movie file, the Turbo encode has given a 573 frame (30fps) movie. Might want to double check your settings.
Andy

Andy Mees · May 24th, 2009, 08:55 PM

Hi Nigel
So I had a good long look at the 3 clips: Elgato, Compressor & Vimeo. The Elgato 8 Mbps encode is really nice, and its very hard visually to tell any difference between it and the Vimeo 40 Mbps encode. Quite a feat. However I will say that both the Elgato and Vimeo encodes seem much more contrasty than the Compressor encode which shows much more detail in the shadows and highlights. Take a look at the very blown out highlights on the grass stems in the foreground at left on the first frame of the Elgato/Vimeo encodes versus the Compressor encode; similarly note the loss of detail in the shadows of the grass in the background at right of the last frame. Admittedly, this higher contrast suits the sample pond image really very well but of course this is something one would really only want to add by choice rather than as a byproduct of the encoding. Would still like to compare against the camera native original for reference though, if we can find a way of sharing that, as I'm thinking the lower contrast evidenced in the Compressor encode might actually be more to do with the Quicktime H.264 gamma shift issue than a better encode.
Thanks again
Andy

Nigel Barker · May 24th, 2009, 10:46 PM

The 40Mbps download file on Vimeo is the original file straight off the camera.

Andy Mees · May 25th, 2009, 01:10 AM

Ah, I didn't realize the raw was also AVC/H264 ... just out of interest, what camear are you shooting with?

But I digress ... an edit and update on my earlier post:

To compare the 3 clips I originally transfered each clip to our video server so that I could properly pass the image to our broadcast and technical monitoring equipment but in doing so each clip was passed through an additional transcode to a 75 Mbps MPEG2 HD intraframe codec. The tested results were as formerly posted, with the original (Vimeo) and the Elgato encode being very hard to tell apart, and with the Compressor encode evidencing less contrast and more detail in highlight and shadows.

Well I've since been comparing the original encodes directly within QuickTime (ie without transcode) and that's skewed the results significantly ... a complete reversal in fact.

In Quicktime we actually see it's the Compressor encode rather than the Elgato that is truer to the original Vimeo clip in terms of quality, contrast and detail; the Elgato's encode still evidences the higher contrast, slightly blown out highlights and crushed blacks, and although thats a good look on this footage I'd still rather see such adjustments only when they are intentional. Hmmm .. is it possible to compensate for this in the Elgato's encoding settings?

Thanks again for posting your clips Nigel, interesting stuff. I'll try to get my hands on some CompressHD/Matrox MAX encodes as and when possible for comparison too, but I'm definitely liking the look of the Turbo 264 HD.

Best
Andy

PS if you do get the chance to rerun your tests on this or other clips, it would be great if you could jot down the encoding times. If nothing else they would be good to ogle at! :-)

Christopher Drews · May 27th, 2009, 07:04 PM

This week at work I've been putting the ElGato through the ranks.
Just thought I'd mention that it is NOT faster than realtime when compressing to 1080p.
See attached screenshot to verify this.
-C

Nigel Barker · May 27th, 2009, 11:27 PM

Quote:

Originally Posted by Christopher Drews

Just thought I'd mention that it is NOT faster than realtime when compressing to 1080p.

It's still pretty close at 23fps though. Software methods can be 5 or 10 times real time.

Ivan Snoeckx · May 30th, 2009, 11:06 AM

The Matrox CompressHD is now shipping! :-)

H.264 encoding accelerator - Matrox CompressHD now shipping

Guy Cochran · June 8th, 2009, 02:14 PM

I was able to spend a little bit of time with the CompressHD card in an old "2008" MacPro. We also happened to have a "2009 Newhalem" MacPro in so I loaded the same clip on both and put the CompressHD in the slower MacPro. Surprisingly enough here are the results of a multipass encode with the "Matrox YouTube 1280x720P" setting in Compressor:

Source Clip 1:31 duration Format: ProRes 1080i60

2008 MacPro Encode Time 1:45 with CompressHD
2009 MacPro Encode Time 2:02 without CompressHD

So with an old Mac and $495 you can still encode faster than a brand new $6599 machine.

I screen captured the encode over a VGA mixer switcher and chromakeyed me in. Here's the
Video

May 23rd, 2009, 07:28 PM	#17
William Hohauser Inner Circle Join Date: Dec 2005 Location: New York City Posts: 2,650	Both files look great. What were the encoding times? __________________ William Hohauser - New York City Producer/Edit/Camera/Animation

May 24th, 2009, 06:21 PM	#23
Andy Mees Major Player Join Date: Mar 2006 Location: Kuala Lumpur, Malaysia Posts: 645	Chris I think some of the quality loss you're seeing might be due to some inadvertent distortion? Looking at the Turbo encode there seems to be a significant stretch and subsequent cropping of the image. Also whereas the Compressor encode produced a 459 frame (24fps) movie file, the Turbo encode has given a 573 frame (30fps) movie. Might want to double check your settings. Andy Last edited by Andy Mees; May 24th, 2009 at 10:13 PM.

May 24th, 2009, 08:55 PM	#24
Andy Mees Major Player Join Date: Mar 2006 Location: Kuala Lumpur, Malaysia Posts: 645	Hi Nigel So I had a good long look at the 3 clips: Elgato, Compressor & Vimeo. The Elgato 8 Mbps encode is really nice, and its very hard visually to tell any difference between it and the Vimeo 40 Mbps encode. Quite a feat. However I will say that both the Elgato and Vimeo encodes seem much more contrasty than the Compressor encode which shows much more detail in the shadows and highlights. Take a look at the very blown out highlights on the grass stems in the foreground at left on the first frame of the Elgato/Vimeo encodes versus the Compressor encode; similarly note the loss of detail in the shadows of the grass in the background at right of the last frame. Admittedly, this higher contrast suits the sample pond image really very well but of course this is something one would really only want to add by choice rather than as a byproduct of the encoding. Would still like to compare against the camera native original for reference though, if we can find a way of sharing that, as I'm thinking the lower contrast evidenced in the Compressor encode might actually be more to do with the Quicktime H.264 gamma shift issue than a better encode. Thanks again Andy Last edited by Andy Mees; May 24th, 2009 at 10:04 PM.

May 27th, 2009, 07:04 PM	#27
Christopher Drews Major Player Join Date: Jul 2007 Location: Los Angeles, CA Posts: 628	Not Faster Than Realtime*** This week at work I've been putting the ElGato through the ranks. Just thought I'd mention that it is NOT faster than realtime when compressing to 1080p. See attached screenshot to verify this. -C Attached Thumbnails

June 8th, 2009, 02:14 PM	#30
Guy Cochran DVCreators.Net Join Date: Oct 2001 Location: Seattle, WA Posts: 892	I was able to spend a little bit of time with the CompressHD card in an old "2008" MacPro. We also happened to have a "2009 Newhalem" MacPro in so I loaded the same clip on both and put the CompressHD in the slower MacPro. Surprisingly enough here are the results of a multipass encode with the "Matrox YouTube 1280x720P" setting in Compressor: Source Clip 1:31 duration Format: ProRes 1080i60 2008 MacPro Encode Time 1:45 with CompressHD 2009 MacPro Encode Time 2:02 without CompressHD So with an old Mac and $495 you can still encode faster than a brand new $6599 machine. I screen captured the encode over a VGA mixer switcher and chromakeyed me in. Here's the Video __________________ Guy Cochran DVinfo Sponsor, Cool Gear - DVeStore!

May 24th, 2009, 08:39 AM	#20
Andy Mees Major Player Join Date: Mar 2006 Location: Kuala Lumpur, Malaysia Posts: 645	Those encodes look excellent Nigel, any chance you could post the original clip too for comparison? Also any chance you could post a second or two of high motion footage with similar settings?

May 24th, 2009, 10:46 PM	#25
Nigel Barker Inner Circle Join Date: Feb 2009 Location: Norwich, Norfolk, UK Posts: 3,531	The 40Mbps download file on Vimeo is the original file straight off the camera.

May 25th, 2009, 01:10 AM	#26
Andy Mees Major Player Join Date: Mar 2006 Location: Kuala Lumpur, Malaysia Posts: 645	Ah, I didn't realize the raw was also AVC/H264 ... just out of interest, what camear are you shooting with? But I digress ... an edit and update on my earlier post: To compare the 3 clips I originally transfered each clip to our video server so that I could properly pass the image to our broadcast and technical monitoring equipment but in doing so each clip was passed through an additional transcode to a 75 Mbps MPEG2 HD intraframe codec. The tested results were as formerly posted, with the original (Vimeo) and the Elgato encode being very hard to tell apart, and with the Compressor encode evidencing less contrast and more detail in highlight and shadows. Well I've since been comparing the original encodes directly within QuickTime (ie without transcode) and that's skewed the results significantly ... a complete reversal in fact. In Quicktime we actually see it's the Compressor encode rather than the Elgato that is truer to the original Vimeo clip in terms of quality, contrast and detail; the Elgato's encode still evidences the higher contrast, slightly blown out highlights and crushed blacks, and although thats a good look on this footage I'd still rather see such adjustments only when they are intentional. Hmmm .. is it possible to compensate for this in the Elgato's encoding settings? Thanks again for posting your clips Nigel, interesting stuff. I'll try to get my hands on some CompressHD/Matrox MAX encodes as and when possible for comparison too, but I'm definitely liking the look of the Turbo 264 HD. Best Andy PS if you do get the chance to rerun your tests on this or other clips, it would be great if you could jot down the encoding times. If nothing else they would be good to ogle at! :-)

May 30th, 2009, 11:06 AM	#29
Ivan Snoeckx Major Player Join Date: Nov 2006 Location: Belgium \| Europe Posts: 441	The Matrox CompressHD is now shipping! :-) H.264 encoding accelerator - Matrox CompressHD now shipping