Love Vegas Hate Render time

Ralph Bowman · May 15th, 2007, 09:01 AM

So after all this discussion of ram and raid should Vegas offer a render board
like the Matrox or Grass Valley? Or is this too limiting?

Ralph Bowman

Glenn Chan · May 15th, 2007, 11:00 AM

Specifically-designed hardware acceleration cards will make little sense on the low and mid-end now that GPU acceleration is available. GPU acceleration runs on commodity video cards (you don't even need the workstation cards) and is flexible (whereas hardware cards can only do a limited # of effects or whatever is programmed into them).

The downside to hardware acceleration right now is that ATI and Nvidia cards operate differently. Code that works on one card may not work properly on the other. Also, not everyone has a good graphics card. So only a small portion of an editing software's users will have GPU acceleration work for them. To support both CPU rendering and GPU rendering can get tricky.

Matthew Chaboud · May 15th, 2007, 04:07 PM

Current cards also report features and accuracy that they don't meet.
Tough game.

As far as what you should task your RAID with, this is almost always reading. Think about it like this:

- 4 HDV clips going into a semi-fancy composite, 100 mbit in.
- Let's say you have sufficient RAM (please try to avoid swapping), 0 mbit through.
- Let's be generous and say that you're going to HDV, 25 mbit out.

This is not uncommon. Even if you have only two pieces of media cross-fading, you're better off reading from the RAID. It's also nicer for playback. If you get swap-file usage during rendering, turn down the Video RAM preview pref. Swapping is insanely slow.

Ron Evans · May 16th, 2007, 07:07 AM

Tomshardware ( http://www.tomshardware.com/2007/05/...d_drive_guide/ )
has this report on some of the latest hard drives that may be of interest to this thread. All the drives have performance for multiple streams of video. Access time for WD Raptor is twice as fast as the slowest drive but throughputs have negligible difference. Simplistically the WD Raptor would be the best for lots of small file accesses in a project, OS swap file, rendered and temp files, but would have negligible difference to the cheapest drive for continuous read of the longer captured files. So if the project is a short video with lots of short clip use and graphics, like a commerial, then the Raptor would make a difference( access time more important then continuous read). If the project is a long form multicamera shoot with, essentially switching cameras in post then the cheapest drive will probably be just fine.

Ron Evans

Danny Fye · May 18th, 2007, 03:28 AM

Ok, before leaving anymore replies I decided to do some tests.

My hard drives are as follows:

C: = Western Digital WDC WD400 2mb 40 gig cache PATA
S: = Western Digital WDC WD1200 8 mb 120 gig cache SATA
T: = Hitachi Deskstar HDS722516VLAT80 160 gig 7.8mb cache PATA
V: = Western Digital WDC WD2500KS 16MB cache 250 gig SATA

According to SiSoft Sandra tests that I have made, the speed of the drives from slowest to fastest are:

C:
T:
S:
V:

The V: drive is suppose to be somewhat faster than the C: drive.

I did a render test using a 5 minute region from a recent Church service. Before I did this test, I did tests with the project having the Sony filter for limiting the IRE level so that it doesn't exceed 100% IRE and the processor was not overclocked. The render time was somewhat slower with the filter and the CPU not being overclocked. I don't have the specific times on those tests. Most of the speed difference was in the filter. The overclock is very mild so it doesn't affect render time very much. The render times after all filters were removed and after the processor was overclocked are as follows:

C: = 2:02
T: = 2:11
S: = 2:05
V: = 2:11 as Source 2:13 Not as Source

The C: drive rendered the fastest even though it is the slowest, oldest and a PATA drive. The V: drive is the fastest drive and yet it was slower than the C: drive. Note: When testing Drives c:, T: and S: drive V: has the source files. I did a test to it to see how it does when redering to the source drive. I moved the source files to the s: drive and did another test render to drive V: and the time was a little bit slower.

I think the differences are mostly due to what part of the drive(s) are being written to. The SiSoft Sandra tests show the rated speed of the drives to be different depending on what part of the drives the data is written to.

No matter the drive type or interface, the render times are very close on all drives.

With this in mind, if one can make changes to the system and/or project to reduce render time then wouldn't it be safe to assume that the hard drives are not really a factor in render times because they apparently are not the bottleneck?

Seems that if the drives are indeed the bottleneck then doing other things such as removing all filters and overclocking the processor wouldn't make any difference in render speed because the drives would then be limiting the render speed? And if the drives are not the bottleneck then would creating a raid-0 be a waste of time in trying to reduce render times?

So when DSE says, "Having a SATA RAID really speeds things up." How is this so? I assume from my tests that this would not the case even though I didn't use SATA RAID. Am I correct or do I need to go sit in a corner somewhere? LOL! No offense intended DSE.

Anyway, that is the best I can come-up with for the moment. All comments are welcome.

Danny Fye
www.dannyfye.com
www.vidmus.com/scolvs

Ron Evans · May 18th, 2007, 05:41 AM

I too do not believe that drives are the bottleneck. Modern drives are large and fast but seek times have not improved as much so a fragmented drive will slow down considerably. I think the key is to think out where the data is coming from and going to and arrange your files to minimize the seek operations. Using uncompressed files with resulting high data rates will change the picture a lot and RAID system will likely help in this circumstance but only if the CPU can manage the rate too. USing a RAID makes organising a little easier, just load everything to the one big hard drive. I don't think there is a performance gain in real life and an exposure to failure that is unneccessary.

Ron Evans

Chris Rieman · May 18th, 2007, 11:55 AM

I would assume in rendering you are writing not reading. Fast drives help in reads, but do not offer much performance mileage in writes.

Ron Evans · May 18th, 2007, 12:28 PM

For a video source the render first reads the source information from hard drive into RAM, processes it as requested ( filters etc ) and renders to the output format then writes file to hard drive. The processing may also involve writes and reads to and from the temp and/or swap files where these are located on some hard drive. So what appears to be a render from the timeline may not be that straight forward. The more RAM the less swap file use etc. Of course if all this information is on one drive then all the reads, writes swap file, temp file seeks , reads and writes will be from this drive. I just checked my drives with the Canopus Raptest and none of them is less than 50MBs reads or writes, that is almost 15 times more than is needed for a single DV/HDV stream. In my mind software design and CPU performance are much bigger issues than performance of modern hard drives.

Ron Evans

Danny Fye · May 18th, 2007, 02:42 PM

Well I have searched and searched the internet and have found that to do a true hardware RAID-0 one would need to spend more than $300.00 to get a card for it. That is more than what my budget will allow for at this time.

So with Windows XP Home, how does one setup a RAID-0? I have seen articles about Windows XP Pro but nothing about Windows XP Home.

Is there a software that one can buy that will make it possible? Why have the ability to create a RAID-0 in the bios like what my system has if Windows XP Home can't see it as one drive instead of two?

I have seen a refference to a hack method from Tom's hardware but I do not want to use such a method. Creates future problems to do so.

Anyone have any answers or is RAID something that Windows XP Home users can only daydream about?

Danny Fye
www.dannyfye.com
www.vidmus.com/scolvs

Marcus Marchesseault · May 18th, 2007, 03:23 PM

"So with Windows XP Home, how does one setup a RAID-0?"

Don't even dream about that one. A software RAID uses CPU cycles. The CPU is always the limiting factor in rendering so anything that adds to it's work is bad. This is why I don't recommend RAID systems for most people. Without the proper hardware and setup, they can actually decrease performance.

Danny Fye · May 18th, 2007, 04:19 PM

Quote:

Originally Posted by Marcus Marchesseault

"So with Windows XP Home, how does one setup a RAID-0?"

Don't even dream about that one. A software RAID uses CPU cycles. The CPU is always the limiting factor in rendering so anything that adds to it's work is bad. This is why I don't recommend RAID systems for most people. Without the proper hardware and setup, they can actually decrease performance.

Does a motherboard with the Intel ICH8R have a true hardware raid? My current motherboard does not have the Intel ICH8R.

If not then I will settle for raid-none for a while. :(

Note: The reason why I am now asking about RAID is a little bit OT with this thread in that whether or not RAID-0 helps with rendering it does help with other things NLE in various ways.

Maybe we need a RAID book for Vegas? What works and what does not work? Tips on how to setup RAID and even a system to best work with Vegas? Not just the highest price stuff either.

Simply saying that one should have a properly configured system for Vegas is one thing but what one needs is some details and even in depth information about what a properly configured system consists of.

What we don't need is posts about how l-o-n-g it takes for Vegas to render because systems are not configured properly or because of editing errors and so on.

My goal is to make my system as productive as possible and to have maximum efficiency in editing. Vegas itself is a huge help in this but even Vegas can be limited with a poorly configured system and/or poor editing habits.

A more productive system leads to much reduced stress which is another one of my goals.

Danny Fye
www.dannyfye.com
www.vidmus.com/scolvs

Ron Evans · May 18th, 2007, 05:06 PM

Optimizing systems needs some effort and planning and VideoGuys http://www.videoguys.com/WinXP.html is a good starting point for the sorts of things you may need to do to your system. For my editing system, first drive " C " WinXP Home seperate hard drive stores programs and OS only, drive 2 " D " Temp drive for editing apps temp files and rendered files, image for OS hard drive backup and nothing else, three drives ( E,F,G)for video files storage, two external USB drives for backups as needed. Most of my projects are multi camera shoots and the camera files get captured to seperate drives. Most services not needed are turned off, no firewall or virus checking etc even CD/DVD auto seek is turned off. Keep temp drive defraged and normally clean off video drives after project so that captures are usually contiguous files.

Ron Evans

Marcus Marchesseault · May 18th, 2007, 06:21 PM

Ron said pretty much everything you need to know in the simplest way possible. Hard drive speed used to be a serious issue, but the incredible data density is effectively giving us RAID performance on a single disk. If you double the data density, twice as much data passes by the heads for every rotation of the disk. This is, in effect, the same as making two drives read/write the same file in a RAID. With up to 250GB on a single platter, drives are more than ten times as dense as they were several years ago but the heads still move the same speed.

I did my first video editing on a 200MHz machine with an 8Gig multi-platter drive. The data rate of HDV is similar to what I was using on that same system. On the one hand, hard drives transfer almost ten times as much data per second. Unfortunately, the access time is exactly the same as it was in 1997 with my 7200rpm SCSI drive.

I'm not saying that a RAID is bad or that you are stupid to want to use one, but rather that it adds complexity and doesn't solve the basic problem of hard drives or even have a significant impact on render times. With long GOP compression in HDV, the processor is what is getting slammed during a render. A good hard drive system is definitely a great idea, but it is my opinion that simplicity is more important than eeking out a tiny bit of performance. Complexity is the breeding ground of problems that are difficult to diagnose.

I know it has been said before, but other than a RAID there are steps you can take that keep things simple yet have a real impact on your hard drives:

Try to keep them defragmented. Fragmented files will bring out the seek-time weakness in all hard drives.

Separate your media amongst different drives. Drives are very good at doing one thing at a time but start to choke if everything is all on one drive. Have separate system, source video, and render drives.

Don't completely fill your drives. Hard drives read/write from the outside towards the inner tracks. The inner tracks are slower than the outer tracks due to circles of smaller diameter having a smaller circumference. Also, a full drive will inevitably have more fragmentation than an empty drive. I've noticed on benchmarks that drives really fall off in speed past about 80% full.

There is no substitue for processing power. The CPU is the workhorse and it will be the bottleneck as long as you don't have a shortcoming in your RAM or hard drive. It's time to stop using that 8Gig hard drive then upgrade to 2Gig of RAM.

Marcus Marchesseault · May 18th, 2007, 06:24 PM

Here is something that isn't often mentioned about Vegas render times and I have no answer at all if it would be effective:

Do people have any significant performance improvements using a networked second machine to help render? With gigabit ethernet and a couple of multi-core machines, it seems like today's computers could really tackle the job of being a small render farm for Vegas. I've never seen it done.

Jeff Harper · May 18th, 2007, 07:17 PM

I cannot back this up, but I'm sure I've read on this forum or somewhere that networked rendering is a real disappointment with Vegas and does not speed rendering times as one would hope.

May 15th, 2007, 09:01 AM	#46
Ralph Bowman Regular Crew Join Date: Sep 2003 Location: Grants Pass, Oregon Posts: 59	Vegas Render Board? So after all this discussion of ram and raid should Vegas offer a render board like the Matrox or Grass Valley? Or is this too limiting? Ralph Bowman

May 18th, 2007, 03:28 AM	#50
Danny Fye Major Player Join Date: Dec 2003 Location: Independence MO. Posts: 318	Hard Drive Tests for Render Speed Ok, before leaving anymore replies I decided to do some tests. My hard drives are as follows: C: = Western Digital WDC WD400 2mb 40 gig cache PATA S: = Western Digital WDC WD1200 8 mb 120 gig cache SATA T: = Hitachi Deskstar HDS722516VLAT80 160 gig 7.8mb cache PATA V: = Western Digital WDC WD2500KS 16MB cache 250 gig SATA According to SiSoft Sandra tests that I have made, the speed of the drives from slowest to fastest are: C: T: S: V: The V: drive is suppose to be somewhat faster than the C: drive. I did a render test using a 5 minute region from a recent Church service. Before I did this test, I did tests with the project having the Sony filter for limiting the IRE level so that it doesn't exceed 100% IRE and the processor was not overclocked. The render time was somewhat slower with the filter and the CPU not being overclocked. I don't have the specific times on those tests. Most of the speed difference was in the filter. The overclock is very mild so it doesn't affect render time very much. The render times after all filters were removed and after the processor was overclocked are as follows: C: = 2:02 T: = 2:11 S: = 2:05 V: = 2:11 as Source 2:13 Not as Source The C: drive rendered the fastest even though it is the slowest, oldest and a PATA drive. The V: drive is the fastest drive and yet it was slower than the C: drive. Note: When testing Drives c:, T: and S: drive V: has the source files. I did a test to it to see how it does when redering to the source drive. I moved the source files to the s: drive and did another test render to drive V: and the time was a little bit slower. I think the differences are mostly due to what part of the drive(s) are being written to. The SiSoft Sandra tests show the rated speed of the drives to be different depending on what part of the drives the data is written to. No matter the drive type or interface, the render times are very close on all drives. With this in mind, if one can make changes to the system and/or project to reduce render time then wouldn't it be safe to assume that the hard drives are not really a factor in render times because they apparently are not the bottleneck? Seems that if the drives are indeed the bottleneck then doing other things such as removing all filters and overclocking the processor wouldn't make any difference in render speed because the drives would then be limiting the render speed? And if the drives are not the bottleneck then would creating a raid-0 be a waste of time in trying to reduce render times? So when DSE says, "Having a SATA RAID really speeds things up." How is this so? I assume from my tests that this would not the case even though I didn't use SATA RAID. Am I correct or do I need to go sit in a corner somewhere? LOL! No offense intended DSE. Anyway, that is the best I can come-up with for the moment. All comments are welcome. Danny Fye www.dannyfye.com www.vidmus.com/scolvs __________________ www.dannyfye.com

May 18th, 2007, 11:55 AM	#52
Chris Rieman Regular Crew Join Date: Jan 2004 Location: Dayton OH Posts: 100	I would assume in rendering you are writing not reading. Fast drives help in reads, but do not offer much performance mileage in writes. __________________ Unsurpassed Dayton Flyer Sports Coverage -------------------------- UD Pride http://www.udpride.com

May 18th, 2007, 02:42 PM	#54
Danny Fye Major Player Join Date: Dec 2003 Location: Independence MO. Posts: 318	Raid-0 How??? Well I have searched and searched the internet and have found that to do a true hardware RAID-0 one would need to spend more than $300.00 to get a card for it. That is more than what my budget will allow for at this time. So with Windows XP Home, how does one setup a RAID-0? I have seen articles about Windows XP Pro but nothing about Windows XP Home. Is there a software that one can buy that will make it possible? Why have the ability to create a RAID-0 in the bios like what my system has if Windows XP Home can't see it as one drive instead of two? I have seen a refference to a hack method from Tom's hardware but I do not want to use such a method. Creates future problems to do so. Anyone have any answers or is RAID something that Windows XP Home users can only daydream about? Danny Fye www.dannyfye.com www.vidmus.com/scolvs __________________ www.dannyfye.com

May 15th, 2007, 11:00 AM	#47
Glenn Chan Inner Circle Join Date: Jun 2003 Location: Toronto, Canada Posts: 4,750	Specifically-designed hardware acceleration cards will make little sense on the low and mid-end now that GPU acceleration is available. GPU acceleration runs on commodity video cards (you don't even need the workstation cards) and is flexible (whereas hardware cards can only do a limited # of effects or whatever is programmed into them). The downside to hardware acceleration right now is that ATI and Nvidia cards operate differently. Code that works on one card may not work properly on the other. Also, not everyone has a good graphics card. So only a small portion of an editing software's users will have GPU acceleration work for them. To support both CPU rendering and GPU rendering can get tricky.

May 15th, 2007, 04:07 PM	#48
Matthew Chaboud Regular Crew Join Date: Sep 2006 Location: Chicago, IL, USA Posts: 93	Current cards also report features and accuracy that they don't meet. Tough game. As far as what you should task your RAID with, this is almost always reading. Think about it like this: - 4 HDV clips going into a semi-fancy composite, 100 mbit in. - Let's say you have sufficient RAM (please try to avoid swapping), 0 mbit through. - Let's be generous and say that you're going to HDV, 25 mbit out. This is not uncommon. Even if you have only two pieces of media cross-fading, you're better off reading from the RAID. It's also nicer for playback. If you get swap-file usage during rendering, turn down the Video RAM preview pref. Swapping is insanely slow.

May 16th, 2007, 07:07 AM	#49
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	Tomshardware ( http://www.tomshardware.com/2007/05/...d_drive_guide/ ) has this report on some of the latest hard drives that may be of interest to this thread. All the drives have performance for multiple streams of video. Access time for WD Raptor is twice as fast as the slowest drive but throughputs have negligible difference. Simplistically the WD Raptor would be the best for lots of small file accesses in a project, OS swap file, rendered and temp files, but would have negligible difference to the cheapest drive for continuous read of the longer captured files. So if the project is a short video with lots of short clip use and graphics, like a commerial, then the Raptor would make a difference( access time more important then continuous read). If the project is a long form multicamera shoot with, essentially switching cameras in post then the cheapest drive will probably be just fine. Ron Evans

May 18th, 2007, 05:41 AM	#51
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	I too do not believe that drives are the bottleneck. Modern drives are large and fast but seek times have not improved as much so a fragmented drive will slow down considerably. I think the key is to think out where the data is coming from and going to and arrange your files to minimize the seek operations. Using uncompressed files with resulting high data rates will change the picture a lot and RAID system will likely help in this circumstance but only if the CPU can manage the rate too. USing a RAID makes organising a little easier, just load everything to the one big hard drive. I don't think there is a performance gain in real life and an exposure to failure that is unneccessary. Ron Evans

May 18th, 2007, 12:28 PM	#53
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	For a video source the render first reads the source information from hard drive into RAM, processes it as requested ( filters etc ) and renders to the output format then writes file to hard drive. The processing may also involve writes and reads to and from the temp and/or swap files where these are located on some hard drive. So what appears to be a render from the timeline may not be that straight forward. The more RAM the less swap file use etc. Of course if all this information is on one drive then all the reads, writes swap file, temp file seeks , reads and writes will be from this drive. I just checked my drives with the Canopus Raptest and none of them is less than 50MBs reads or writes, that is almost 15 times more than is needed for a single DV/HDV stream. In my mind software design and CPU performance are much bigger issues than performance of modern hard drives. Ron Evans

May 18th, 2007, 03:23 PM	#55
Marcus Marchesseault Trustee Join Date: Nov 2005 Location: Honolulu, HI Posts: 1,961	"So with Windows XP Home, how does one setup a RAID-0?" Don't even dream about that one. A software RAID uses CPU cycles. The CPU is always the limiting factor in rendering so anything that adds to it's work is bad. This is why I don't recommend RAID systems for most people. Without the proper hardware and setup, they can actually decrease performance.

May 18th, 2007, 05:06 PM	#57
Ron Evans Inner Circle Join Date: Mar 2003 Location: Ottawa, Ontario, Canada Posts: 4,220	Optimizing systems needs some effort and planning and VideoGuys http://www.videoguys.com/WinXP.html is a good starting point for the sorts of things you may need to do to your system. For my editing system, first drive " C " WinXP Home seperate hard drive stores programs and OS only, drive 2 " D " Temp drive for editing apps temp files and rendered files, image for OS hard drive backup and nothing else, three drives ( E,F,G)for video files storage, two external USB drives for backups as needed. Most of my projects are multi camera shoots and the camera files get captured to seperate drives. Most services not needed are turned off, no firewall or virus checking etc even CD/DVD auto seek is turned off. Keep temp drive defraged and normally clean off video drives after project so that captures are usually contiguous files. Ron Evans

May 18th, 2007, 06:21 PM	#58
Marcus Marchesseault Trustee Join Date: Nov 2005 Location: Honolulu, HI Posts: 1,961	Ron said pretty much everything you need to know in the simplest way possible. Hard drive speed used to be a serious issue, but the incredible data density is effectively giving us RAID performance on a single disk. If you double the data density, twice as much data passes by the heads for every rotation of the disk. This is, in effect, the same as making two drives read/write the same file in a RAID. With up to 250GB on a single platter, drives are more than ten times as dense as they were several years ago but the heads still move the same speed. I did my first video editing on a 200MHz machine with an 8Gig multi-platter drive. The data rate of HDV is similar to what I was using on that same system. On the one hand, hard drives transfer almost ten times as much data per second. Unfortunately, the access time is exactly the same as it was in 1997 with my 7200rpm SCSI drive. I'm not saying that a RAID is bad or that you are stupid to want to use one, but rather that it adds complexity and doesn't solve the basic problem of hard drives or even have a significant impact on render times. With long GOP compression in HDV, the processor is what is getting slammed during a render. A good hard drive system is definitely a great idea, but it is my opinion that simplicity is more important than eeking out a tiny bit of performance. Complexity is the breeding ground of problems that are difficult to diagnose. I know it has been said before, but other than a RAID there are steps you can take that keep things simple yet have a real impact on your hard drives: Try to keep them defragmented. Fragmented files will bring out the seek-time weakness in all hard drives. Separate your media amongst different drives. Drives are very good at doing one thing at a time but start to choke if everything is all on one drive. Have separate system, source video, and render drives. Don't completely fill your drives. Hard drives read/write from the outside towards the inner tracks. The inner tracks are slower than the outer tracks due to circles of smaller diameter having a smaller circumference. Also, a full drive will inevitably have more fragmentation than an empty drive. I've noticed on benchmarks that drives really fall off in speed past about 80% full. There is no substitue for processing power. The CPU is the workhorse and it will be the bottleneck as long as you don't have a shortcoming in your RAM or hard drive. It's time to stop using that 8Gig hard drive then upgrade to 2Gig of RAM.

May 18th, 2007, 06:24 PM	#59
Marcus Marchesseault Trustee Join Date: Nov 2005 Location: Honolulu, HI Posts: 1,961	Here is something that isn't often mentioned about Vegas render times and I have no answer at all if it would be effective: Do people have any significant performance improvements using a networked second machine to help render? With gigabit ethernet and a couple of multi-core machines, it seems like today's computers could really tackle the job of being a small render farm for Vegas. I've never seen it done.

May 18th, 2007, 07:17 PM	#60
Jeff Harper Inner Circle Join Date: Jun 2005 Location: Cincinnati, OH Posts: 8,425	I cannot back this up, but I'm sure I've read on this forum or somewhere that networked rendering is a real disappointment with Vegas and does not speed rendering times as one would hope.