What's the system you have for the Win2k3 32-bit? I'm wondering because I'd like to get a duel Opteron but i'm not going to if it's not a lot faster then a comparable Xeon (which i'd prefer AMD).
:)

"Until Vegas is optimized for AMD64 - probably not too much."

Do you mean it has to be optimized for that specific processor, or for 64-bit proccesors in general? These are the kinds of the things that bring out my ignorance.

(addendum)
And I just noticed that it requires Windows XP 64, too. Missed that first time around.

J--

optimized for that specific processor, or for 64-bit proccesors in general --both , though 64 bit is the more crucial one there at this point (the new xeon 64bit will be compatable with amd64) .

>Do you mean it has to be optimized for that specific processor, or for 64-bit proccesors in general?

Right now the only relevant (for desktop NLE) 64 bit processor on the market IS AMD64. Since Microsoft has decided to support AMD64 with XP and Win2k3 - Intel has little choice but to make a processor that is compatible with AMD64. So a 64 bit Vegas (IMHO) is inevitable - just a matter of time. Maybe Vegas 6.

Just to make it clear from the start: An AMD64 processor will run 32 bit apps and operating systems without any real speed compromise. The 32 bit part of the processor is NOT some skimpy add on - it is truly designed to be BOTH a 32 bit and a 64 bit processor.

So you can happily run the 32 bit version of XP and Vegas and everything else just like a normal PC. But much of the 64 bit areas will sit dormant until you get a 64 bit OS like the new XP (still in beta?), Win2k3(definitely in beta), and Linux of course.

Right now the biggest barrier to us Vegas users for moving to 64 bit Windows would be lack of hardware drivers and .NET. Otherwise I would be running in 64 bit now.

See, 32 bit software runs fine in 64 bit windows, except for the low level stuff. Virus scanners, Firewalls, and hardware drivers all have to be recompiled for the new 64 bit Windows kernel. So far there are lots of beta drivers out - much of it is very stable and very fast. NVidia was right on top of it, ATI was not. I was lucky that most of my hardware is solid in 64 bit already.

I've done capturing and editing in Vegas without any problems in 64 bit Windows. Everything was very fast. It's just the lack of .NET Framework for scripting (is the 64 bit version out in beta yet?) and I never got the DirectX audio plugins working.

Hope this isn't redundant info!

I have a lot of experience running CAD workstations on both 32 bit M$ machines, as well as on 64 bit UNIX machines. From my experience, I can tell you that making the transition from 32 bit to 64 bit computing buys you practically NOTHING in terms of rendering speed. The whole reason for moving to 64 bit computing is because of the increase numerical depth available with 64 bit. What this means in lay terms is that color images can have much greater bit depth than they presently have with 32 bit. IOW, the 4:2:2 bit depth can go to 4: 4:4, or whatever, necessary for High Def rendering. The net result with 64 bit computing is the SAME RENDER speed but at much greater bit depth. For some reason, non computing types think this means faster computing with 4:2:2 bit depths. I have found this to not be true. I suspect a lot of the mis-information is from AMD marketing hype as AMD tries to capture more market share.

BillRavens is right. Think of 64 bit as "wider" not necessarily faster, although memory intensive stuff like database applications run a lot faster. I believe this is because memory operations can be executed a lot more efficiently in a 64 bit environment. An operation that requires two reads from memory in 32 bit could be done with one read in 64 bit.

The other benefit is a little further out for the desktop market, but 64 bit allows you to use a lot more total RAM than you can with a 32 bit PC. Most of us aren't approaching the limit at 32 bit though.

Thanks, guys for the clarification. That helped a great deal. Much appreciated!

So I'm presuming that the processor's speed is still the means to faster render times? Not that it's an issue with me. I'm happy with what I've got, but when faster is available, I'll take it. ;o)

J--

It just means you can move around (and perform operations on) twice as much data at the same speed. Technically it has nothing to do with higher bit depth in images, other than to say that if you did double the bit depth in an image, you could render at the same speed like you said (theoretically), or if you kept the same bit depth, render at twice the speed (*theoretically*). Of course there are so many other factors, like what kind of processing you're doing on the image, whether or not that has to be done at the pixel level, or can be applied to multiple pixels at once, etc....

Ahh... the old "give people washing machines to save time and they find other work to do" situation. :)

Or, get a computer that renders what you do faster so you ad more "stuff" and the render time goes back to what it was. :)

I don't think it''s mis-info from AMD. I think it's misunderstand of most people (not ones here). Example:
several years ago rendering basic 3-D on a computer took forever (simple polygon objects, flat shades, low res, etc). Now, it goes faster, BUT people bring it to a higher res, more effects, etc. So, it still takes as long because people add more stuff (take a look at the Pixar ball/lamp animation. Took a long time to render when they did it. id software did it in real time on a mac with a Geforce 3 back on '01).

So, if Vegas supported 64-bit, theroiticaly (i say that because there's no Vegas 64 yet) the same stuff we do now (DV) should go faster. But, by then we will probely be using HD, which will slow things down to the same speed they are now. :)

It's the circle of life.

Here are the results of my test. No big surprises. It looks like we'll have to wait for Vegas to be compiled for AMD64 before we'll see any real benefits, but even now, in its 32 bit form on an early beta of Win2k3 64 (with beta drivers for everything) it runs a bit faster.

Notice that the best efficiency (highest frames per/sec) is reached when rendering not two, but four rendertests at once. This result is consistent across both the 32 bit and 64 bit platforms.

I used this rendertest.veg
http://www.sundancemediagroup.com/help/kb/kb_download.asp?id=8

Win2k3 32 bit

Single render: 89s   1.68 fps

Two at once:   89s   3.37 fps

Four at once: 176s   3.41 fps

Five at once: 221s   3.39 fps

Six at once:  266s   3.38 fps



64 bit

Single render: 84s    1.78 fps

Two at once:   88s    3.41 fps

Four at once:  173s   3.46 fps

Five at once:  217s   3.45 fps

Six at once:   263s   3.42 fps

Also notice that even when rendering 6 projects at once, the frame rate is only about 1% off peak. I configured this system for multitasking so I would expect it to be very good at it, but I have nothing to compare it to. :P Anyone?

By the way, I was operating with the System Performance optimized for background processes. That helps to make sure all the projects get to 100% at about the same time. I wanted them to all be running at the same time for as long as possible.

For the test I started all the projects within a couple seconds of each other - then waited for the last one to finish and used its elapsed time to calculate the frame rate.

re: What's the system you have

This is the system I have except with DDR 400 RAM:
http://www.digitalvideoediting.com/cgi-bin/getframeletter.cgi?/2004/04_apr/reviews/5bhm8ltr.htm

I got it in February. At the time it was the fastest (overall) dual configuration I could find. Unless Intel came out with a new Xeon since then with a higher fsb speed... it's probably still the case. You can certainly get faster single-proc pc's. Probably the P4 EE is the fastest but I haven't been keeping track of that stuff.

Regards,
- Ben

Hey Ben/Cheesehole,
your tests turn up some surprising results that don't fit in with the theory I know about how computers work. This leads me to be skeptical and question your testing methodology. First off, why I'm skeptical.
A- Your test shows that four rendertests is faster than two. However, your two AMD64 processors can only render one project at a time each so four should be slightly slower than two. Vegas also happens to be multithreaded so you can have two processors working on one render at a time (although the division of labor is very very unequal with rendertest.veg). The most logical explanation for this to me is that your testing methodology is faulty in some way. The way you start the tests and measure might somehow favor four tests at once.
B- Your other surprising result is that vegas in 64-bit windows is slightly faster than 32-bit windows (XP? 2k3?). Now the benefits for 64-bit computing are that you can use bigger numbers (64bits instead of 32) and that AMD64 processors get to use a larger number of registers (not sure about this last point). The first shouldn't help Vegas until it is written to take advantage of it. This may not be likely since 32-bit floating points are currently the biggest numbers in use (32bit audio processing; FCP4 uses 32-bit floating point numbers for high definition rendering; DV is 8 bit, uncompressed can be 10bit).

It's hard to tell whether Vegas is slightly faster because of the jump to 64-bit or because Windows is now faster.

potential testing pitfalls:
A- rendertest.veg could be more indicative of real world projects. As it currently stands, the design is very synthetic. One definite flaw is that it is very stacked against hyperthreading and dual processors. Vegas seems to be multithreaded such that the first thread handles video FX and the second all the audio plus DV decoding/encoding. Because rendertest.veg is so 'intensive' the second thread does not have much work to do. In real world renders the second thread will usually have more to do.
B- From my own testing, I've found that renders tend to vary up to 2 seconds. You can minimize this by shutting down every other program and service, moving the mouse as little as possible once the render starts, deleting your rendered file every time (hard drive fragmentation issues), and by starting the test by hitting enter on your keyboard (so you don't move the mouse at all).

Also because of your testing methodology (rendering multiple projects), variation might be much much greater.

C- Background processes and programs can slow down your renders by a few seconds. It might be that the two versions of windows are different in the programs that start up.
D- Your tests are awfully close. The differences cannot be teased out from statistical variation. There are no significant differences between some of the results.
E- Also if you start comparing some of the times, the differences are not consistent. Comparing to single renders to two, in the first (32bit windows) there is a 100% bump but in 64bit there is a ~90% speed bump.
F- (EDIT: nevermind)
G- Usefulness of your testing: It doesn't help to determine whether 64-bit is faster if you accpet the theory that programs have to be written for it. It does point out that 64-bit windows 2k3 might be a faster version of windows... but you have to prove that you controlled for crap that accumulates in 32-bit windows (which I presume is the OS you normally use and have installed programs on it and such, which can slow things down).

EDIT: I might have missed the point of your testing entirely.

He was just to testing to see (basicly) if it's worth it to get the AMD 64-3200 over the XP 3200 (or other same speed AMD 64/XP chips).

But, I've noticed the same thing he has on renders (I've only used Vegas on AMD's so I don't know if it's the same as Intel. I assume it is).

I just took 1 minute of footage I had. I cut it in 4 places, overlaped, then added transitions. I rendered that by itself. 2:07 seconds total (just over 53 second long).

Next, I opened up that project in another instance of Vegas. I rendered both Vegases. I got a total time of 4:03. 11 seconds less then if I rendered then at seperate times.

When I setup Vegas on an AMD xp 2500 last year, I could render 3 things at once (all were 30 seconds commercials). The total render time was about 1 minute (per 3 renders). Took 30 seconds for each one if I did them by themselves.

You can dublicate this on your computer. And, the reason this doesn't work with the was you think computers with is because (no offence) you've got it a little wrong. No program is 100% efficient. Even if it's operating at 100% cpu, it still bounces down. With 2+ copies of Vegas running it's a little more efficient. If a program used 100% of the CPU then your computer would freeze every time it went to process something intense. Same theory that's behind car pools (you use X gallons of gas to get to work. If you add 2 more people you don't use 3X gallons, you use 1.2x). :)

Hey I'm glad someone is paying attention! :D

The test was just a quickie. It's true the 32 vs. 64 may not be that valid, but the results of the renders were very consistent - they never varied by even a second. But yes we're only talking about rendertest.veg, which may not reflect the way you use Vegas.

I stand by the results that show increased efficiency when rendering more than 2 projects at once though. HappyFriar is right... Vegas doesn't really use 100% of the processor during a render. I should have tried 3 renders at once. That might actually be more efficient than 4.

It's hard to tell whether Vegas is slightly faster because of the jump to 64-bit or because Windows is now faster.

Right and we're not really flying faster - we just have a strong tail wind! ;) Either way you get where you are going earlier.

Some of the steps I took to improve the accuracy in a less-than-ideal testing environment were:

1 - Ran the test a few times for each
2 -Shut down every process possible before running the renders.
3 - Deleted the files between each render (not that hard drive speed / fragmentation would have had much effect on this test since it is so CPU intensive)
4 - A few times I closed all the Vegas instances and started over. I did that every once in a while to see if it would make a difference - it never changed the render results by even a second so I didn't bother to do it every time. Also restarting the PC had no effect.

Things I could have done:
- Rendered 3 projects at once (just didn't have time - but 3 may actually be more efficient than 4)
- Built the system from scratch with a blank drive (although I do have most of the same software installed on both OS's)

If I have time to do a different VEG I'll post those results. But I don't see a compelling reason to pursue it too much until we get final versions of Win2k3 for AMD64 / finalized drivers / Vegas64, etc...