Render At Maximum Quality at DVinfo.net

Steve Kalle · July 29th, 2010, 08:47 PM

In AME, it increases the scaling quality and adds significant rendering time. Because most people say that Premiere/AME do not scale very well, I am wondering if it actually increases quality. I am testing it tonight and the first encode took 90 mins with it checked and its currently encoding without RAMQ checked, which should take only 20 mins.

Anyone know if it actually provides noticeable improvement?

Mitch Hunt · July 29th, 2010, 09:59 PM

Yes it does. I use MRQ anytime scaling or deinterlacing is involved. When going from HD upper field first to SD Lower field first the difference is vast.

Steve Kalle · July 29th, 2010, 11:45 PM

I finished the encoding and there is a significant difference in quality going from XDCAM EX 1080p to 480p when using RAMQ.

Khalil Ismail · July 29th, 2010, 11:46 PM

I am going to try that.

Dan Burnap · July 30th, 2010, 05:46 AM

I always use MRQ, even though with an i7, 12GB RAM, RAID0 project disk and seperate drive to renders, it takes over 15 hours to render a 1080/30p AVCHD (HMC150), 2 hour, Encore project into a SD DVD.

Steve Kalle · July 30th, 2010, 11:04 AM

Dan, correct me if I'm wrong but it doesn't sound like you are using CS5 because your render speed would be faster.

I have a HP Z800 with dual 6 core X5660 (2.66GHz) and Hyperthreading turned OFF in addition to 24GB. Rendering a 30min XDCAM EX with slight CC and MRQ took 90 mins. With MRQ off, it took 25 mins.

Dan Burnap · July 31st, 2010, 04:15 AM

Steve, yep, I'm using CS4, however, I have CS5 installed also and the render times are more or less equal.

During the renders, task Manager reports all cores at 99%, judging by your results it would seem I have some sort of issue...

Ann Bens · July 31st, 2010, 06:05 AM

Quote:

Originally Posted by Mitch Hunt

Yes it does. I use MRQ anytime scaling or deinterlacing is involved. When going from HD upper field first to SD Lower field first the difference is vast.

If the SD is mpeg leave it to upper.
If the SD is dv avi it automatic converts to lower.

Steve Kalle · July 31st, 2010, 11:24 AM

Dan, are you using any effects? I did some testing with various graphics under CS5 using the hack and compared it to CS4. With 3 layers of AVCHD and 8 filters on each and 2 mins total length, CS4 was 51 mins and CS5 was 9-13 mins depending upon the video card. Furthermore, CS5's Dynamic Link got a speed boost due to better ram management and PPro and AME are now 64bit - Encore might be also.

Have you run the PPBM CS5 benchmark? This will give you an idea of where your performance bottlenecks lie. According to this benchmark, MPEG2 encoding really benefits from more cores.

This is a silly question but have you checked that Hyperthreading is ON? 8 'threads' in Task Manager is what you should see. I only ask because my custom PC's motherboard had HT OFF when I first booted it up.

Now that I think about it more - because you are using DL between AME, PPro & Encore, what does Task Manager show for ram/memory usage during the encode? It might be hitting a memory wall.

Dan Burnap · August 1st, 2010, 05:45 AM

Hi Steve, thanks for the input. Havent run the PPBM test yet, its on my list. Hyperthreading is ON, eight cores appear in task manager, all maxed out for the duration of the render. As the CPU is maxed out that doesnt seem to look like a disk bottleneck anyway.

The footage has a few effects dotted about but 90% just straight footage.

I should of added that in CS4 I render out a uncompressed AVI (original dimensions 1920x1080) and import this into Encore. I would love to do it all by dynamic link but in CS4 at least you get random stutters with AVCHD so its too hit and miss to get a decent result. When I move over to CS5 I will be using the dynamic link for sure.

I have GTX285 installed.

On my second PC (win 7, i7, 6GB of RAM) encode times in Encore are pretty similair. I dont know what's going on!

Steve Kalle · August 1st, 2010, 10:48 AM

Dan, what takes 15 hours to render - the render to uncompressed AVI, the encode in Encore or is it a total of both? If its a total of both, then I can somewhat understand why it takes so long. One thing I have noticed with uncompressed AVIs is that they use a significant amount of CPU power. With the PPBM CS5, one test is exporting to uncompressed AVI and it uses about 15% of my 24 'cores' (which run @2.66GHz) and that equates to roughly 45-50% on an i7 920.

Several months ago, I tested a possible speed booster in CS4 - if I am delivering SD but I am shooting 1080p, so, I created 1080p, 720p and 720x406 sequences in Premiere and scaled to frame size. The 720p render/encode was faster than the 1080p and the 720x406 was even faster. The reason for this, I believe, is there are fewer pixels for Premiere to render/calculate. I know for certain that this works in another program TMPGEnc. This is why I almost never edit in 1080p but usually 720p as most of my exports are SD MPEG2 & H264. However, this testing did not use MRQ.

July 29th, 2010, 08:47 PM	#1
Steve Kalle Trustee Join Date: Aug 2009 Location: Chicago, IL Posts: 1,554	Render At Maximum Quality In AME, it increases the scaling quality and adds significant rendering time. Because most people say that Premiere/AME do not scale very well, I am wondering if it actually increases quality. I am testing it tonight and the first encode took 90 mins with it checked and its currently encoding without RAMQ checked, which should take only 20 mins. Anyone know if it actually provides noticeable improvement?

July 29th, 2010, 11:46 PM	#4
Khalil Ismail Regular Crew Join Date: Jul 2010 Location: Palm Desert California Posts: 42	Thanks Mitch I am going to try that.

July 29th, 2010, 09:59 PM	#2
Mitch Hunt Regular Crew Join Date: Nov 2008 Location: Broken Bow, NE Posts: 145	Yes it does. I use MRQ anytime scaling or deinterlacing is involved. When going from HD upper field first to SD Lower field first the difference is vast.

July 29th, 2010, 11:45 PM	#3
Steve Kalle Trustee Join Date: Aug 2009 Location: Chicago, IL Posts: 1,554	I finished the encoding and there is a significant difference in quality going from XDCAM EX 1080p to 480p when using RAMQ.

July 30th, 2010, 05:46 AM	#5
Dan Burnap Major Player Join Date: Oct 2005 Location: Rio de Janeiro Posts: 335	I always use MRQ, even though with an i7, 12GB RAM, RAID0 project disk and seperate drive to renders, it takes over 15 hours to render a 1080/30p AVCHD (HMC150), 2 hour, Encore project into a SD DVD.

July 30th, 2010, 11:04 AM	#6
Steve Kalle Trustee Join Date: Aug 2009 Location: Chicago, IL Posts: 1,554	Dan, correct me if I'm wrong but it doesn't sound like you are using CS5 because your render speed would be faster. I have a HP Z800 with dual 6 core X5660 (2.66GHz) and Hyperthreading turned OFF in addition to 24GB. Rendering a 30min XDCAM EX with slight CC and MRQ took 90 mins. With MRQ off, it took 25 mins.

July 31st, 2010, 04:15 AM	#7
Dan Burnap Major Player Join Date: Oct 2005 Location: Rio de Janeiro Posts: 335	Steve, yep, I'm using CS4, however, I have CS5 installed also and the render times are more or less equal. During the renders, task Manager reports all cores at 99%, judging by your results it would seem I have some sort of issue...

July 31st, 2010, 11:24 AM	#9
Steve Kalle Trustee Join Date: Aug 2009 Location: Chicago, IL Posts: 1,554	Dan, are you using any effects? I did some testing with various graphics under CS5 using the hack and compared it to CS4. With 3 layers of AVCHD and 8 filters on each and 2 mins total length, CS4 was 51 mins and CS5 was 9-13 mins depending upon the video card. Furthermore, CS5's Dynamic Link got a speed boost due to better ram management and PPro and AME are now 64bit - Encore might be also. Have you run the PPBM CS5 benchmark? This will give you an idea of where your performance bottlenecks lie. According to this benchmark, MPEG2 encoding really benefits from more cores. This is a silly question but have you checked that Hyperthreading is ON? 8 'threads' in Task Manager is what you should see. I only ask because my custom PC's motherboard had HT OFF when I first booted it up. Now that I think about it more - because you are using DL between AME, PPro & Encore, what does Task Manager show for ram/memory usage during the encode? It might be hitting a memory wall.

August 1st, 2010, 05:45 AM	#10
Dan Burnap Major Player Join Date: Oct 2005 Location: Rio de Janeiro Posts: 335	Hi Steve, thanks for the input. Havent run the PPBM test yet, its on my list. Hyperthreading is ON, eight cores appear in task manager, all maxed out for the duration of the render. As the CPU is maxed out that doesnt seem to look like a disk bottleneck anyway. The footage has a few effects dotted about but 90% just straight footage. I should of added that in CS4 I render out a uncompressed AVI (original dimensions 1920x1080) and import this into Encore. I would love to do it all by dynamic link but in CS4 at least you get random stutters with AVCHD so its too hit and miss to get a decent result. When I move over to CS5 I will be using the dynamic link for sure. I have GTX285 installed. On my second PC (win 7, i7, 6GB of RAM) encode times in Encore are pretty similair. I dont know what's going on!

August 1st, 2010, 10:48 AM	#11
Steve Kalle Trustee Join Date: Aug 2009 Location: Chicago, IL Posts: 1,554	Dan, what takes 15 hours to render - the render to uncompressed AVI, the encode in Encore or is it a total of both? If its a total of both, then I can somewhat understand why it takes so long. One thing I have noticed with uncompressed AVIs is that they use a significant amount of CPU power. With the PPBM CS5, one test is exporting to uncompressed AVI and it uses about 15% of my 24 'cores' (which run @2.66GHz) and that equates to roughly 45-50% on an i7 920. Several months ago, I tested a possible speed booster in CS4 - if I am delivering SD but I am shooting 1080p, so, I created 1080p, 720p and 720x406 sequences in Premiere and scaled to frame size. The 720p render/encode was faster than the 1080p and the 720x406 was even faster. The reason for this, I believe, is there are fewer pixels for Premiere to render/calculate. I know for certain that this works in another program TMPGEnc. This is why I almost never edit in 1080p but usually 720p as most of my exports are SD MPEG2 & H264. However, this testing did not use MRQ.