Good / Bad time on render test?

kirkdickinson wrote on 2/22/2003, 6:32 PM
I am setting up my new system and downloaded the render test from Sundance to test it. My old system (dual 1GHZ P-III) runs it in just a little over 5 minutes.

Just got my new system going, but it isn't finished yet. I only have one HD in it right now. Here are the specs:

Supermicro X4DA8
2 - Xeon 2.4 Gig
1 - Gig DDR
1 - 20 Gig Ultra-160 Drive (I need more, but this is my OS drive)
XP-Pro with all the patches
Hyperthreading turned on.

My new system runs the render test in 1:58. This seems slow to me for the system that I have. I know that once I get another HD, that it should help.

There is a system on the rendertest results that is only a dual P-III 1.2 that gets it done in 2:02. Why doesn't this dual Xeon whip through that text in 60-70 seconds?

Kirk

I didn't monitor the CPU activity, so I don't know if it was using them both.

Comments

riredale wrote on 2/22/2003, 7:14 PM
I recall that when the rendertest first came out, no one could break the 2 minute barrier, so count yourself among the Chosen Few. I would say anything approaching 2 minutes is very fast, anything around 3 minutes is fast, anything around 4 is good, and anything around 10 is usable but slow.

I don't think VV takes advantage of dual processors, but could be wrong. Do a search on this board.
jsteehl wrote on 2/24/2003, 1:06 PM
If you are using the Mainconcept codec installed by VV then you may see better utilization of HT if you try the 1.3 version. Unfortunetly VV does not have an upgrade to use it (yet) but you can download the demo and try the render test with that (fully functional but with a watermark).

Vegas is not written to support SMPT (Dualies) but even if it were the issue is with the render engine, in this case MainConcept.

-Jason, S.
kosins wrote on 2/24/2003, 1:58 PM
Interesting.
I'm new here, and I'd not seen the render test before.
My recent system finished it in 1:28 minutes.

P4,3.06GHZ W/533MHZ
1GB DDR SDRAM AT 333MHZ
128MB DDR ATI RADEON 9700 PRO W/TV-OUT
60GB 7200RPM ULTRA

It's a Dell system.
I considered the dual Xeon processor setup but couldn't get find any information confirming that it would be worth the extra expense.
bjtap wrote on 2/24/2003, 2:07 PM
kosins,
Exactly which model Dell do you have? Can you give info on the mobo, and os? Thanks, Barry
kosins wrote on 2/24/2003, 2:19 PM
Barry,
This is "copy and paste" of the system info:

"OS Name Microsoft Windows XP Professional
Version 5.1.2600 Service Pack 1 Build 2600
OS Manufacturer Microsoft Corporation
System Name XXXX
System Manufacturer Dell Computer Corporation
System Model Dimension 4550
System Type X86-based PC
Processor x86 Family 15 Model 2 Stepping 7 GenuineIntel ~3056 Mhz
Processor x86 Family 15 Model 2 Stepping 7 GenuineIntel ~3056 Mhz
BIOS Version/Date Dell Computer Corporation A04, 1/20/2003
SMBIOS Version 2.3
Windows Directory C:\WINDOWS
System Directory C:\WINDOWS\System32
Boot Device \Device\HarddiskVolume2
Locale United States
Hardware Abstraction Layer Version = "5.1.2600.1106 (xpsp1.020828-1920)"
User Name XXXXXXXXXXX
Time Zone Eastern Standard Time
Total Physical Memory 1,024.00 MB
Available Physical Memory 714.19 MB
Total Virtual Memory 3.40 GB
Available Virtual Memory 2.59 GB
Page File Space 2.40 GB
Page File C:\pagefile.sys

I'm looking for specific motherboard specs. I'll post 'em when I find 'em.

I got this system about two weeks ago. Pretty happy with it.

John


kirkdickinson wrote on 2/24/2003, 6:28 PM



Doesn't look like the the dual Xeon is worth the extra expense if the render test is the deciding factor.

My 2.4 dual took 30 seconds longer than your 3.06 single.

If I figure right, an increase in clock speed of .66 Gig = 27.5% increase
and a savings of 30 seconds in render time = 25% faster.

Looks like it is scaling almost exactly baised on processor speed. I couldn't afford the 2.6 or 2.8 Xeons.

Well, with my duallie, I can probably render a project full blast in one instance of VV and still edit in a second. :-)

Kirk
kirkdickinson wrote on 2/24/2003, 11:12 PM
Interesting, when I turn hyperthreading off, I get a one second improvement in render time.

One strange thing, with the HT on and the virtual 4 CPU's it looks like they are all sharing nearly equal loads with the entire CPU load averaging 26%. With the HT off, the load seems to get shifted mostly to CPU 2 and the total CPU load is reported as 56%.

Maybe I will run two instances of VV and two rendertests at the same time and see what happens.

Kirk
kirkdickinson wrote on 2/24/2003, 11:36 PM
OK, the results for two at a time.

HT off. CPU usage 100% 2:06 and 2:05
HT on. CPU usage 54% 2:16 and 2:15

Seems like the HT is eating up extra cycles trying to manage extra virtual cpus.

Is is surprising that it only takes less than 10 seconds extra to do two at a time verses one with the HT off.

Kirk
riredale wrote on 2/25/2003, 1:00 AM
One way that dual processors might really shine would be if you ran multiple instances of Vegas. I did that, last fall, when facing a deadline. I ran two instances at idle priority doing rendering in the background, while running a third doing editing in the foreground. I have never used a dual-processor setup, but assume that it would be smart enough to utilize both chips in that instance.
Ritchie wrote on 2/25/2003, 6:59 AM
For the rendertest what is the intended render target? It doesn't really specify. Couldn't you expect different times if you were rendering to PAL vs NTSC vs Uncompressed vs MPEG2? Maybe I missed where it states what you are rendering to so all things are equal.

I tried it on my Athlon XP 2000+ and it was around 3 minutes rendering to MPEG2 NTSC.
slacy wrote on 2/25/2003, 9:28 AM
When you say "idle" priority, is that the same as "low" priority on W2K?
riredale wrote on 2/25/2003, 11:13 AM
Don't know; probably. Here in W98SE-land I use a freeware utility called "Multitask Monitor" to go into a particular VV instance and change its process priority to "idle" from "normal." In that way, I still have full use of my foreground application, yet the work gets done in the background when the foreground task is just waiting for me to do something.

It would be terrific if one could just pull down a menu within VV to adjust this priority level. Maybe in VV5.

Ritchie: I think the whole point of the RenderTest is to render to just plain DV avi, for the sake of consistency. I don't know whether it would make much of a difference doing PAL or NTSC, but I doubt it, since both formats require about the same amount of pixel processing.