RAM upgrade produces incredible results at DVinfo.net

Jason Donaldson · June 26th, 2008, 11:15 AM

I couldn't believe this. I am running a Quad Core Q6600. I had 2 1gb sticks of DDR2 533 RAM. I used vegas to capture, edit and render a video file from my Sony HVR-A1U in HDV format. Rendering of the finished product including transitions and color correction etc took 13 hours and 14 minutes. This was on Tuesday. I recently changed my RAM to a 2GB (1 stick) of DDR2 667. I re-rendered the exact same project, and this time it only took 8 hours 43 minutes. The RAM was the only thing that was changed on my PC. I used the same HD to render to and I didn't defrag or anything else between the two renderings.

Is it normal to have such a difference in performance between the two RAM configurations? I am definately not complaing at all.

Paul Kellett · June 26th, 2008, 11:50 AM

8 hours 43 mins !!
How long was the project ?
What format did you render to ?
My quad core q6600 with 4gb ram renders my HD footage into SD for dvd in about 2.5 times.

Paul.

Jason Donaldson · June 26th, 2008, 12:24 PM

Do you mean 2.5 times longer than 8 hours and 43 minutes?
Anyway, my project is 85 minutes in length, and I rendered out to WMV "8Mbps HD 1080-30p Video" with the render quality set to "best".

Jon Fairhurst · June 26th, 2008, 12:46 PM

RAM includes some metadata that tells the motherboard how fast to run it. It's likely that the new RAM allowed access with fewer wait states, speeding up all access to the memory. And yeah, that will speed up renders and anything else that exceeds the cache (on chip RAM) and uses the RAM in the motherboard slots.

The funny thing is that the previous RAM might have been *capable* of running as fast as the new RAM, but to find that out would take testing and optimization. Overclockers adjust things, test, adjust, test and so on to optimize not just CPU speed, but RAM access times, ignoring the metadata.

On the other side of the coin, if you ever build a machine that crashes, try slowing the memory access. That can turn a crash dummy PC into a rock solid performer.

Paul Kellett · June 26th, 2008, 01:24 PM

Jason, i mean 2.5 times the length of the project, a 1 hour project will be rendered to mpeg-2 in 2.5 hours.

So you had an 85 mins of hdv project and it took 8hrs and 43 mins to render to WMV !! That's very slow.
I wouldn't even wait that long.

Paul.

Jason Donaldson · June 26th, 2008, 02:01 PM

Yes, WMV HD. You're rendering to SD, not HD. There is a ton of transitions and color corection within the project, and 2 nested .veg files as well as 5.1 sound (6 tracks audio). I don't think your PC could do any better considering the contents of the project file.

Jeff Harper · June 26th, 2008, 02:15 PM

Hey Jason, this is off topic, but I notice you are using the extremely slow "best" setting.

Unless I'm mistaken, (wouldn't be the first time) rendering using the "best" setting is only needed for using lots of photos, as I understand it.

Someone correct me if I'm mistaken. I do know it slows rendering to a crawl, and from having tried it a few times I truly saw no difference in output quality, except for photos, where I have see a slight improvment.

Anyone else?

Incidentally, if you rendered with Best setting the first time and then good the second (not knowing you were doing so) that would account for a drastic speed increase as well.

Kim Olsson · June 26th, 2008, 02:32 PM

Render time is not all about time length of your film...

If you use 3d track motion, FX (filters, color corrections and so on), transitions and more the rendering time will be affected... These stuff is heavy things for Sony Vegas...

I recently made a 80 seconds long film, which took 4 hours to render (M2T, 1280x720, 50fps).
Its all about the content in your project.

Edward Troxel · June 26th, 2008, 02:33 PM

"BEST" should be used when resizing is done. This is almost always when using images. But it can apply even when using video. For example, HD sources to an SD render could benefit from "BEST". Otherwise, "GOOD" typically gives results that are just as good as "BEST". I would recommend doing a small test both ways and see whether or not it will benefit your situation.

Garrett Low · June 26th, 2008, 04:02 PM

Wanna see your render times go crazy, try rendering a 32-bit floating point project. I had one that was about 2.5 hours in length with lots of color correction, transitions and some inserted picture in picture type portions. On a Q6600 based machine with 4 GB of ram it took 8 days to render!

Jeff Harper · June 26th, 2008, 04:31 PM

Yes Garrett, it is amazingly slow. I have done it, and admit color was a bit better, but not better enough to justify it's use for general video use.

Jason Donaldson · June 26th, 2008, 04:37 PM

Quote:

Originally Posted by Edward Troxel

"BEST" should be used when resizing is done. This is almost always when using images. But it can apply even when using video. For example, HD sources to an SD render could benefit from "BEST". Otherwise, "GOOD" typically gives results that are just as good as "BEST". I would recommend doing a small test both ways and see whether or not it will benefit your situation.

Thanks for the info Edward. I just assumed that if good was good, then best was best so to speak.

Jason Donaldson · June 26th, 2008, 04:38 PM

Quote:

Originally Posted by Jeff Harper

Incidentally, if you rendered with Best setting the first time and then good the second (not knowing you were doing so) that would account for a drastic speed increase as well.

I have saved a preset using the same settings so best was used for both...good thinking though

Jeff Harper · June 26th, 2008, 04:40 PM

Well, I'm impressed. I had no idea memory speed could have such a drastic impact...thanks for letting us know...

John Miller · June 26th, 2008, 04:45 PM

Quote:

Originally Posted by Garrett Low

Wanna see your render times go crazy, try rendering a 32-bit floating point project. I had one that was about 2.5 hours in length with lots of color correction, transitions and some inserted picture in picture type portions. On a Q6600 based machine with 4 GB of ram it took 8 days to render!

See this post for an explanation as to why floating point is s-o-o slow:

http://www.dvinfo.net/conf/showpost....1&postcount=37

June 26th, 2008, 11:15 AM	#1
Jason Donaldson Regular Crew Join Date: Apr 2007 Location: Ottawa, Ontario Canada Posts: 180	RAM upgrade produces incredible results I couldn't believe this. I am running a Quad Core Q6600. I had 2 1gb sticks of DDR2 533 RAM. I used vegas to capture, edit and render a video file from my Sony HVR-A1U in HDV format. Rendering of the finished product including transitions and color correction etc took 13 hours and 14 minutes. This was on Tuesday. I recently changed my RAM to a 2GB (1 stick) of DDR2 667. I re-rendered the exact same project, and this time it only took 8 hours 43 minutes. The RAM was the only thing that was changed on my PC. I used the same HD to render to and I didn't defrag or anything else between the two renderings. Is it normal to have such a difference in performance between the two RAM configurations? I am definately not complaing at all. __________________ HVR-A1u,WCS-999,HVL-20DW2 w/diff,2xAudio-Technica Freeway 600,Flycam 3000,Vegas 7E

June 26th, 2008, 11:50 AM	#2
Paul Kellett Trustee Join Date: Nov 2007 Location: Bristol UK Posts: 1,273	8 hours 43 mins !! How long was the project ? What format did you render to ? My quad core q6600 with 4gb ram renders my HD footage into SD for dvd in about 2.5 times. Paul. __________________ Round 2 GH5,FZ2000

June 26th, 2008, 12:24 PM	#3
Jason Donaldson Regular Crew Join Date: Apr 2007 Location: Ottawa, Ontario Canada Posts: 180	Do you mean 2.5 times longer than 8 hours and 43 minutes? Anyway, my project is 85 minutes in length, and I rendered out to WMV "8Mbps HD 1080-30p Video" with the render quality set to "best". __________________ HVR-A1u,WCS-999,HVL-20DW2 w/diff,2xAudio-Technica Freeway 600,Flycam 3000,Vegas 7E

June 26th, 2008, 12:46 PM	#4
Jon Fairhurst Inner Circle Join Date: May 2006 Location: Camas, WA, USA Posts: 5,513	RAM includes some metadata that tells the motherboard how fast to run it. It's likely that the new RAM allowed access with fewer wait states, speeding up all access to the memory. And yeah, that will speed up renders and anything else that exceeds the cache (on chip RAM) and uses the RAM in the motherboard slots. The funny thing is that the previous RAM might have been capable of running as fast as the new RAM, but to find that out would take testing and optimization. Overclockers adjust things, test, adjust, test and so on to optimize not just CPU speed, but RAM access times, ignoring the metadata. On the other side of the coin, if you ever build a machine that crashes, try slowing the memory access. That can turn a crash dummy PC into a rock solid performer. __________________ Jon Fairhurst

June 26th, 2008, 01:24 PM	#5
Paul Kellett Trustee Join Date: Nov 2007 Location: Bristol UK Posts: 1,273	Jason, i mean 2.5 times the length of the project, a 1 hour project will be rendered to mpeg-2 in 2.5 hours. So you had an 85 mins of hdv project and it took 8hrs and 43 mins to render to WMV !! That's very slow. I wouldn't even wait that long. Paul. __________________ Round 2 GH5,FZ2000

June 26th, 2008, 02:01 PM	#6
Jason Donaldson Regular Crew Join Date: Apr 2007 Location: Ottawa, Ontario Canada Posts: 180	Yes, WMV HD. You're rendering to SD, not HD. There is a ton of transitions and color corection within the project, and 2 nested .veg files as well as 5.1 sound (6 tracks audio). I don't think your PC could do any better considering the contents of the project file. __________________ HVR-A1u,WCS-999,HVL-20DW2 w/diff,2xAudio-Technica Freeway 600,Flycam 3000,Vegas 7E

June 26th, 2008, 02:15 PM	#7
Jeff Harper Inner Circle Join Date: Jun 2005 Location: Cincinnati, OH Posts: 8,425	Hey Jason, this is off topic, but I notice you are using the extremely slow "best" setting. Unless I'm mistaken, (wouldn't be the first time) rendering using the "best" setting is only needed for using lots of photos, as I understand it. Someone correct me if I'm mistaken. I do know it slows rendering to a crawl, and from having tried it a few times I truly saw no difference in output quality, except for photos, where I have see a slight improvment. Anyone else? Incidentally, if you rendered with Best setting the first time and then good the second (not knowing you were doing so) that would account for a drastic speed increase as well.

June 26th, 2008, 02:32 PM	#8
Kim Olsson Major Player Join Date: Oct 2007 Location: Sweden Posts: 263	Render time is not all about time length of your film... If you use 3d track motion, FX (filters, color corrections and so on), transitions and more the rendering time will be affected... These stuff is heavy things for Sony Vegas... I recently made a 80 seconds long film, which took 4 hours to render (M2T, 1280x720, 50fps). Its all about the content in your project.

June 26th, 2008, 02:33 PM	#9
Edward Troxel Sponsor: JET DV Join Date: Dec 2001 Location: Southern Illinois Posts: 7,953	"BEST" should be used when resizing is done. This is almost always when using images. But it can apply even when using video. For example, HD sources to an SD render could benefit from "BEST". Otherwise, "GOOD" typically gives results that are just as good as "BEST". I would recommend doing a small test both ways and see whether or not it will benefit your situation. __________________ Edward Troxel [SCVU] JETDV Scripts/Scripting Tutorials/Excalibur/Montage Magic/Newsletters

June 26th, 2008, 04:02 PM	#10
Garrett Low Trustee Join Date: Jul 2007 Location: Novato, CA Posts: 1,774	Wanna see your render times go crazy, try rendering a 32-bit floating point project. I had one that was about 2.5 hours in length with lots of color correction, transitions and some inserted picture in picture type portions. On a Q6600 based machine with 4 GB of ram it took 8 days to render!

June 26th, 2008, 04:31 PM	#11
Jeff Harper Inner Circle Join Date: Jun 2005 Location: Cincinnati, OH Posts: 8,425	Yes Garrett, it is amazingly slow. I have done it, and admit color was a bit better, but not better enough to justify it's use for general video use.

June 26th, 2008, 04:40 PM	#14
Jeff Harper Inner Circle Join Date: Jun 2005 Location: Cincinnati, OH Posts: 8,425	Well, I'm impressed. I had no idea memory speed could have such a drastic impact...thanks for letting us know...