Vegas and VirtualDub Rendering Times - Why is Vegas so slow?

[harley2ride]

I'm using Vegas 5 on a Dell 8400 3.4ghz with 1mg ram and two SATA HD's. I'm getting pretty close on a 1 to 1 (actually a little better), converting captured DV-AVI to MPEG2. 1 1/2 hrs of avi are taking abour 1hr to encode and burn. I have no complaints at all about that... That includes custom into screens and ending rolling credit screens.

[edDV]

Just as a test, I encoded the VBR unfiltered file to a hard drive on a different EIDE channel (separate EIDE channel for tmp, source and destination). In this case the resurts were only better by 1 sec. 5:52 vs 5:53.

OS and source file were on the same EIDE controller but on separate HDDs.

[edDV]

I'm using Vegas 5 on a Dell 8400 3.4ghz with 1mg ram and two SATA HD's. I'm getting pretty close on a 1 to 1 (actually a little better), converting captured DV-AVI to MPEG2. 1 1/2 hrs of avi are taking abour 1hr to encode and burn. I have no complaints at all about that... That includes custom into screens and ending rolling credit screens.

If you have the time it would be interesting if you could repeat at least one of the tests above on a 5 min DV format source to say CBR 8000 224.

My P4 2.4 GHz. (without hyperthreading) is getting 5:58/5:00 or 1.193x*. It would be interesting to see what your P4 3.4 GHz (with Hyperthreading) gets as a direct comparison.

* (5+(58/60)) / 5 = 1.193

[VideoQuest]

Wow! This is great info for me.

edDV, our computer systems are simular as far as CPU speed. Yours is 2.4 and mine is 2.2. However; you have 1GB of memory while I have 640K. In addition I beleive my internal HD is only 5400rpm.

It would appear that in my case the biggest bottle necks are the MEMORY and the HD speed. Even so I am a little surprised that my render time for non-filtered is 3x yours and my filtered time is 2x to 3x yours. Just out of curiousity, on your system, when you are rendering in Vegas do know how much memory is AVAIBLABLE as stated in the task manager under PHYSICAL MEMORY?

When Vegas is open with no rendering I have available physical memory of 370k. When Vegas is rendering a 5 minute clip the available physical memory drops down to below 200k. In addition, my CPU is running a 100% the entire time.

More Info:

My 5 Minute clip contains about 5 events. Each event has my filters applied. When I add a in and out transition on 1 event my render time drastically increases. Now for a 5 Minute clip with filters and transitions it takes 30 mintues to render. This is 6x original clip time.

When I started rendering my available physical memory was at 200k. By end of rendering the 5 Minute clip my physical available memory dropped down to 170k. If your available memory is much more than mine that tells me that my bottle neck is memory.

This memory issue may mean nothing. I am just trying to understand why my rendering seems to be multiple times slower than other similar systems.

Also, if I take Vegas out of the picture and use VirtualDub and CCE instead the rendering times are much, much lower and the quality seems to be the same.

Thanks for everything.

[junkmalle]

Make sure your drives are running in DMA mode. From Device Manager select IDE ATA/ATAPI Controllers, Primary IDE Channel, Properties, Advanced Settings.

When you're not rendering start up Task Manager and look at CPU usage. It should be near zero.

If you are running antivirus software -- shut it down before rendering.

[edDV]

Wow! This is great info for me.

edDV, our computer systems are simular as far as CPU speed. Yours is 2.4 and mine is 2.2. However; you have 1GB of memory while I have 640K. In addition I beleive my internal HD is only 5400rpm.

It would appear that in my case the biggest bottle necks are the MEMORY and the HD speed. Even so I am a little surprised that my render time for non-filtered is 3x yours and my filtered time is 2x to 3x yours. Just out of curiousity, on your system, when you are rendering in Vegas do know how much memory is AVAIBLABLE as stated in the task manager under PHYSICAL MEMORY?

When Vegas is open with no rendering I have available physical memory of 370k. When Vegas is rendering a 5 minute clip the available physical memory drops down to below 200k. In addition, my CPU is running a 100% the entire time.

I don't see much evidence of Vegas using a great deal of memory during encoding. Windows seems to grab close to half of it during boot up.

Before opening Vegas


During render

[VideoQuest]

Thanks so much edDV.

I think I have come to the conclusion that I need a new system for my video editing. I have looked at all the charts in this thread and I am still a little confused about the benefits of dual core versus hyperthreading systems.

There is some evidence that shows that their is a rending benefit with dual core, while other forums state that it does not make much difference. I have seen the same conflicting opinions about hyperthreading.

Has anyone here actually rendered with Vegas with dual core and/or hyperthreading systems?

Has anyone experienced any direct benefits one way or another?


Thanks.

[edDV]

More Info:

My 5 Minute clip contains about 5 events. Each event has my filters applied. When I add a in and out transition on 1 event my render time drastically increases. Now for a 5 Minute clip with filters and transitions it takes 30 mintues to render. This is 6x original clip time.

When I started rendering my available physical memory was at 200k. By end of rendering the 5 Minute clip my physical available memory dropped down to 170k. If your available memory is much more than mine that tells me that my bottle neck is memory.

This memory issue may mean nothing. I am just trying to understand why my rendering seems to be multiple times slower than other similar systems.

I think you may be adding apples, pears and oranges. You have effects rendering, filtering and encoding going on. Render the 5min timeline first to a 5min DV file (Tools - Render to new track) and then encode just that to MPeg2. Then take that DV.avi file to the other encoders to compare times. There shouldn't be that much difference.

[VideoQuest]

I understand your point.

However; my bottom line is "Why is Vegas dramatically slow on my system".

Most people that use Vegas to not seem to have this problem with similar systems and configurations. Even now I do not understand why Vegas is slow for me. I have understood and implemented all suggestions and recomendations.

I suppose the only way for me to truely move on is to upgrade to another computer system. I am hopeful that this should resolve my problem.

So not I am trying to determine which environment to truely best for Vegas and MainConcept. Again I have read all info on this subject within this forum and others. There seems to be some difference of opinion when it comes to this matter.

I would like to know if anyone has first-hand experience with using Vegas in a dual-core or hyperthreaded system. I would like to collect more information on Vegas rendering times with dual core and hyperthreading. And, I really don't care about multi-tasking. I will have a dedicated video editing system.

edDV, you and others have been the greatest in helping me out here. I am very grateful.

Thanks.

[edDV]

Well try to encode just that DV file and let's see what you get with just the MPeg2 encoder.

Those of us that have been using these packages for awhile are used to the processing delays and compared to history, these editors seem fast enough to us.

With big projects, editing can be very fast using uncompressed SDI (Blackmagic hardware) and external video servers. Vegas allows attaching a "render farm" of PC machines to speed effects rendering and filtering. In the near future, this capability will move into multicore processors at low cost.

Even in a home system, DV format workflow is fairly fast. MPeg2 encoding is something you do overnight or in the background. Highest quality MPeg2 is done at the end of the project.

I have Premiere on one machine and Vegas on the other. When one is stuck rendering, I work on the other. These days I mostly use Premiere to preprocess elements and put it together in Vegas. Sometimes it goes the other way. I can run them both off the same drives.

[junkmalle]

According to this:

https://www.videohelp.com/tools?tool=78

Vegas is multithreaded so can take advantage of hyperthreaded and multi-core CPUs. I don't have the program but typically hyperthreading gets you 10 to 15 percent more performance on CPU intensive tasks. Dual core somewhere around 50 to 70.

Did you check your idle CPU usage, hard drive DMA, and antivirus settings as I suggested earlier?

[guns1inger]

Don't confuse dual-core with dual CPU. While dual core is at heart dual CPU on a chip, at present the two cores are comparatively slow (typically 2.2 - 2.4 Ghz each). A hyperthreaded 3.8 GHz P4E will give you pretty much the same speed form a single chip, and run single threaded application 100% faster or better (think games etc).

A dual CPU motherboard on the otherhand will allow you have two mych faster CPUs running together, which will give you substantial gains.

When dualcore runs cool enough to allow 3.4GHz or better per core, they will be worth looking at.

[junkmalle]

Intel's Pentium D processors run from 2.8 to 3.2 GHz. AMD's X2 line runs from 2.0 to 2.4 GHz (there's also the 1.8 GHz Opteron 165 which seems to have disappeared). But you can't directly compare clock speeds between AMD and Intel processors.

I haven't seen any direct comparisons of rendering speed with Vegas but here's one of a similar multithreaded MPEG2 encoder, Mainconcept:

http://www.tomshardware.com/cpu/20051121/the_mother_of_all_cpu_charts-33.html

In that benchmark the slowest Athlon 64 X2 processor (2.0 GHz 3800+, ~$320) is faster than the fastest single core P4 (3.8 GHz Prescott, ~$600) and almost as fast as the fastest Intel dual core processor (Pentium D Extreme Edition 840, ~$1000).

And if you want to go dual Xeon 3.8 GHz be prepared to spend $3000 for a pair of CPU's and $500 for a motherboard.

Of course, Vegas could be different...

[ronnylov]

I have a dual core AMD Athlon 64 X2 3800+ CPU. I have overclocked it from 2,0 GHz to 2,4 GHz (running at the same speed as Athlon 64 X2 4600+). I have 2 GB of RAM memory and also Vegas 6.0c installed.

But I have only access to PAL DV video. I think rendering times would be the same as NTSC because the higher resolution of PAL is compensated with lower framerate which make the amount of data to process about the same per minute.

Do you want me to benchmark vegas with PAL video?

[VideoQuest]

edDV, unfortunately I do not have Mainconept as standalone. Should I use Mainconcept in standalone to compare only encoding times or should I use CCE for example?

junkmalle, yes my idle cpu time varies between 5 to 7 percent.

ronnylov, yes a benchmark would be helpful, thankyou.

It seems to me that based on the feedback I am getting that a hyperthreading (instead of dual core) would be better for the Vegas environment.

Also, it seems that it may be better to go AMD instead of Intel.

Thanks.

[junkmalle]

AMD doesn't have hyperthreading, only dual core. Dual core is usually faster than hyperthreading (although the some of the dual core P4s are hyperthreaded too -- making the CPU look like four processors).

The P4 CPU can perform some operations in parallel. For example it can perform a floating point multiply and an integer multiply at the same time. Hyperthreading makes the single CPU look like two CPUs to take advantage of this. If one thread is waiting for a floating point operation to complete, the other can still run integer operations. But in most real-world applications there is only a small amount of code that can take advantage of this (not to mention that only some code is multithreaded in the first place). You usually only get 10 to 15 percent extra performance with hyperthreading.

Try running Task Manager while you're encoding. Look at the Processes tab and see if something other than Vegas is eating up CPU time. 5 to 7 percent, although not much in itself, is rather high for idle CPU usage. My computers (P4 2.8 HT, P4 2.8 non-HT, Athlon 64 X2 3800+) all run at 0 or 1 when idle.

[vhelp]

Hi everyone.

I have been following (here and there a little) and I couldn't help
but come up with an idea. Maybe it will flush-out another possible
cause or blame. Anyways.

@ edDV and VideoQuest --

But first, may I ask the question.. Have the *both* of you gone to this
thread and run the same test yet ??

--> VideoHelp.com Video Benchmarking

(please report your test results back here on this)

I think it might help you both. Else, you could try the test I roughly
laid out below.

Could you both try this test theory out.

Item List of tools:

** TMPGenc
** a set template
** 720 x 480 resolution
** same DV codec

Ok. If someone could agree on a specific template to use in TMPGenc,
and *both* of your guys run your DV (assuming you are using the same
DV Codec - make sure you are) and proceed to encode with CBR.


My theory is this ...

If you *both* get the same "timed" numbers (give or take a couple of minutes)
then both of the computers (edDV and VideoQuest) are setup properly.
This means that your system and software (TMPGenc) is setup properly
and that you both are equally/equivalently fine-tuned for MPEG encoding
in each system, at maximum, per system setup with TMPGenc. I guess you
would call this the "equalibrium" of both systems, and no (these two)
systems are faulty. In other words, you both passed.

(in this 1st instance of the test, we are determining to see if one sys
is improperly setup somewhere in the artchetecture of the pc, else it
could be software/BIOS etc that needs looking into)

However..

But, if one of you fail, by evidensing lower number, then it is safe
to assume that the slower system has some problems, and it could be
any number of things, such as: CPU; Mobo; Cache setup; RAM Type/Clock/Etc;
BIOS setup; Timing setup; Chipset setup; Vegas or Premier Sofware config/setup;
templates; configs; etc etc etc.

What do you say guys ?

-vhelp 3661

[edDV]

edDV, unfortunately I do not have Mainconept as standalone. Should I use Mainconcept in standalone to compare only encoding times or should I use CCE for example?

Thanks.

No, I suggested you render your 5 minute levels and filters into a DV-AVI file and then use Vegas (internal MPeg2 encoder) to encode the DV file to MPeg2. That way you are comparing apples to apples (encoding funtion to encoding function) with CCE.

My tests on the 5 min clip showed MPeg2 encoding at ~5:58 (apples only). When the levels filter, color correction and transitions were added the processing time went to 15+ minutes (oranges + pears + prunes + apples).

[edDV]

...
My computers (P4 2.8 HT, P4 2.8 non-HT, Athlon 64 X2 3800+) all run at 0 or 1 when idle.

Mine runs 0-1% as well when Anti-Virus is shut off. The idle Task Manager shot above spiked to 4% only during the screen capture (Gadwin PrintScreen was compressing a JPG).

[edDV]

Hi everyone.

I have been following (here and there a little) and I couldn't help
but come up with an idea. Maybe it will flush-out another possible
cause or blame. Anyways.

@ edDV and VideoQuest --

But first, may I ask the question.. Have the *both* of you gone to this
thread and run the same test yet ??

--> VideoHelp.com Video Benchmarking

(please report your test results back here on this)

My TMPGEnc MPeg2 demo ran out but Vegas rendered the file to MPeg2 in 26 sec. using VBR settings above. VideoQuest could try the same.

[VideoQuest]

AMD doesn't have hyperthreading, only dual core. Dual core is usually faster than hyperthreading (although the some of the dual core P4s are hyperthreaded too -- making the CPU look like four processors).

The P4 CPU can perform some operations in parallel. For example it can perform a floating point multiply and an integer multiply at the same time. Hyperthreading makes the single CPU look like two CPUs to take advantage of this. If one thread is waiting for a floating point operation to complete, the other can still run integer operations. But in most real-world applications there is only a small amount of code that can take advantage of this (not to mention that only some code is multithreaded in the first place). You usually only get 10 to 15 percent extra performance with hyperthreading.

Try running Task Manager while you're encoding. Look at the Processes tab and see if something other than Vegas is eating up CPU time. 5 to 7 percent, although not much in itself, is rather high for idle CPU usage. My computers (P4 2.8 HT, P4 2.8 non-HT, Athlon 64 X2 3800+) all run at 0 or 1 when idle.

I take it that you use Vegas and that you are a fan of hyperthreading instead of dual core. You seem to feel that dual core is not quite there yet or that Vegas would not benefit as much. Am I correct in this assumption?

However; I am still confused due to the fact that some of the performence charts show that the AMD's out perform the Intel's. I just can't seem to get conclusive results on one or the other.

I am looking into why my cpu idle time is higher than normal.

[edDV]

...

... fan of hyperthreading instead of dual core. You seem to feel that dual core is not quite there yet or that Vegas would not benefit as much. Am I correct in this assumption?

Hypertherading and dual core are separate issues and additive. Hyperthreading is similar to the old math processor chips, allowing hardware algorithms on the chip to substitute for CPU computation. Hyperthreading is targeted to boost image compositing and filtering algorithms.

Unless you have some project that needs to be completed in the next few months, your current processor is sufficient to learn Vegas and other programs. I'm waiting a few months before jumping into multi-core. Next April, new versions of both Vegas and Premiere Pro will be announced and those will be benchmarked against all the new dual core CPUs. A new version of the Mainconcept SDK is also rumored to be part of these updates. I assume Mainconcept is re-tuning for multi-core as well as multi-CPU.

Dual core AMD and Pentiums will be rapidly improving over the coming months.

[junkmalle]

I take it that you use Vegas

No, I don't use Vegas (I indicated as much earlier).

and that you are a fan of hyperthreading instead of dual core. You seem to feel that dual core is not quite there yet or that Vegas would not benefit as much. Am I correct in this assumption?

No! You must be confusing me with gunslinger who appears to prefer Intel's high clock speed hyperthreading CPUs. It's a matter of what software you run and how you use it. I'll write more about this later if you want (don't have time now).

However; I am still confused due to the fact that some of the performance charts show that the AMD's out perform the Intel's. I just can't seem to get conclusive results on one or the other.

If you look at the benchmark I referred to earlier (at Tom's hardware) you'll see that the slowest AMD X2 (dual core) processor is as fast as the fastest Intel processors (single or dual) -- using Mainconcept's MPEG2 encoder. I don't know about Sony's Vegas but my guess is you'll see similar results.

I am looking into why my cpu idle time is higher than normal.

Start by using Windows Task Manager (Alt+Ctrl+Del) to see what program is sucking up CPU time. If you can figure out what's wrong with your current system you'll get a 3x boost in speed for free!

[VideoQuest]

junkmalle, indeed, I am sorry for the confusion and thanks for your help.

[junkmalle]

I ran the TMPGEnc benchmark that vhelp suggested:

Athlon XP 1700+: 111 seconds
P4 2.8 GHz -HT: 83 seconds
P4 2.8 GHz +HT: 85 seconds
Athlon 64 X2 3800+: 35 seconds

Both P4 computers have 512 MB of PC3200 DRAM. The dual core A64 X2 has 1024 MB DRAM -- I don't think the additional DRAM had any effect on rendering time for the 7 second video. I used the latest version (free) of TMPGEnc from their web site. I verified that the TMPGEnc on the HT computer was running in HT mode. I would have expected it to do better, not worse! My numbers appear to be in line with what others have posted in the benchmark thread. Except those that mistakenly ran the benchmark in single pass CBR mode.

A side note: I noticed that the benchmark uses 2-pass rendering with TMPGEnc's highest motion precision. I usually use single pass (twice as fast as 2-pass) constant quality mode with motion precision set at "Low Quality (Fast)"(which gets you about 98 percent of the motion precision on most material). That took my A64 only 4 seconds. So the settings you use can make an enormous difference in encoding time. I don't know if Vegas gives you control over motion precision but that could make a large difference in encoding time.

I ran the benchmark on another computer in the house, and old Athlon XP 1700+, with 512 MB PC2700 DRAM. Added it to the list above.

[ronnylov]

junkmalle:
Do you have an AMD Athlon 64 X2 3800+ (dual core 2,0 GHz) or Athlon 64 3800+ (single core 2,4 GHz)? I guess you have dual core.
I got 34 seconds in the TMPGEnc benchmark with AMD Athlon 64 X2 3800+ when running at default speed. But I have 2 GB RAM.

[junkmalle]

Yes, my 3800+ is the dual core processor. Sorry, I should have pointed that out in the original message. I edited it to clarify.

Since TMPGEnc lets you disable its multithreading option I tried rendering with and without multithreading. With multithreading the conversion took 35 seconds as noted earlier. Without multithreading it took 68 seconds!

[ronnylov]

Now I have benchmarked my AMD X2 3800+ when running overclocked to 4600+ speed in Vegas. Dual core Athlon 64 running at 2400 MHz.

With a PAL DV video with length of 6:07 (previously rendered to new track) I encoded it to MPEG-2 with the DVD PAL separate streams template, setting CBR video 8000 kbit/s bitrate, max quality, best rendering quality and 224 kbit/s audio the encoding time was 4:01. So it is encoding at 152% speed with the mainconcept plugin in Sony Vegas. This is from a DV track with no filters applied.

So a one hour DVD may be encoded in 40 minutes. Still not very fast but quite fine with software encoding. The video quality was nice too.

EDIT: (4*60+1)/(6*60+7)=0.66x


Benchmark 2:

Added the filters Brightness and contrast, Color corrector (secondary) and Black Restore (not sure if this is same as black level, could not find the filter black level). Rendering the track to MPEG-2 as above.

Rendering time 4:59. It is still faster than real time (0.81x). But maybe the black restore filter is faster than black level? Or perhaps dual core works well in this case. Both CPU was loaded at 100 % during the rendering.

[VideoQuest]

Since I started this thread I decided to let everyone know where I am at.

Originally with my Dell Dimension 2400 I was experiencing rending plus encoding times on the order of 4x to 6x. This was very frustrating for me.

After attempting to optimize my system and run numerous test - with no results, I decided to purchase another system.

I purchase an Intel HT, 3.4cpu, 1 Gig Ram and two internal SATA Drives. Now using the same Vegas projects and clips my rendering and encoding times are about 1.3x to 2.5x. A vast improvment!

I have concluded that the most important factors for Vegas rendering and encoding is:

1. Raw CPU Speed
2. Memory
3. HD Speed (Write Speed Especially)
4. The Filter Chain Order

Anyway, all is right with the world again.

I really appreciate the help I have received from everyone here.

Thanks.