Comments

GlennChan wrote on 2/23/2004, 11:49 AM
>>>Did you download the rendertest_2 I uploaded to the sundance site? I got the approval e-mail thismorning saying it was submitted. It has the CC, chroma, and film effects.<<<
I can't reach it right now.

>>>I don't think HT would realy show use unless you had audio to process, which the 2nd thread normaly uses. I don't think Vegas uses HT the way you're thinking (to process lots of effects).<<<
The second thread is used for: encoding DV, processing audio. I think it's used for decoding your media files too (that would make sense). In nearly every render, you are going to need to decode and encode DV. As renders get more intensive, the 1st processor has more work to do but the 2nd processor has the same amount of work. So it ends up that the 2nd processor isn't working most of the time. When that happens you don't see hyperthreading kick in.

I just did a quick test on rendertest.veg (the original). My computer is a pentium 2.6C, 800FSB with 1X512MB RAM (video setups should be using 2X256 or 2X512MB for dual channel).
normal render time: 111, 111, 111 (seconds)
turning multiprocessor off within Vegas: 112, 112, 112
Setting affinities in Windows to one processor: 113, 113, 113
Gonna reboot and disabled Hyperthreading within BIOS, so it's not quasi-disabling hyperthreading from within Vegas or with WIndows.

EDIT: I rebooted and got a whole bunch of different results, because some of the programs that start up slow down XP and because file fragmentation started becoming a factor (rendered to a 10GB partition on my hard drive, which is quite full :/ ). But anyways, enabling hyperthreading seems to make a 1 or 2 second difference in rendertest.veg. In my own tests, hyperthreading boosts things by 3 to 9% and can boost rendering speed by 20% or so on a ridiculously render-unintensive project (JPEGs color corrected and outputted to DV).

>>>Also, have you checked out the stuff used in the render test? a gaussian blur was added, 2 layer masks (the bottom 2 and upper 2 layers), and a track motion were applied. I think these are more real world then excessive CC & chroma key (if the video is shot "correctly" lots of CC isn't necessary, just maybe a little).<<<
Maybe you find that combination of filters to be real world, but I don't. A render test IMO should be using DV footage, not generated media. I don't find masks or gaussian blur too useful, but I suppose some people might. I think everyone will find color correction useful (hey, maybe people don't find it useful).
TheHappyFriar wrote on 2/23/2004, 1:05 PM
I mostly do slice and dice video work with a couple transitions thrown in. :)

i'm a simple man
MarkWWW wrote on 2/24/2004, 4:03 PM
I've just tried rendering your rendertest_2 on my Athlon64 FX51 and it takes 3 minutes 23 seconds. (By comparison the original rendertest takes 1 minute 24 seconds on this machine.)

Interesting that my FX51 takes proportionally longer on the second half of the render than your XP1800.

Mark
RichMacDonald wrote on 2/24/2004, 7:03 PM
My memory is poor, but some months ago there was a claim/test that the Intel chips are sensitive to different FX, whereas the AMDs crank along, no matter what. IOW, the Intels did best on the simple stuff but poorly on the complex stuff, while the Athlon did less well on the simple stuff but better on the complex stuff.

I think the post was commenting on a render test that was performed by one of the video mags.

There was no followup, so the subject died. Perhaps it would be valuable to develop more than one render test
jamcas wrote on 2/25/2004, 4:04 AM
P4 3.2 (overclocked to 3.8) 1 gig 800fsb

Rendertest 1 1min 18sec (78 seconds)

Rendertst 2 3min 6 sec (186 seconds)

faster than the opteron results quoted further above. What is the fastest render time todate and what kind of machine was it on ?

jc