Holee Smokes!

Spot|DSE wrote on 6/8/2005, 2:36 PM

Just finished building an AMD dual 275 machine on a Tyan K8WE mobo, 3 Barracuda 250's, with a PATA boot drive. 4 gigs of RAM running WinXP 32.

This mama screams!
I captured a fairly lengthy stream of HDV, and dropped it on the timeline to test speed, and thought I'd made an error when the m2t file played back at a solid 29.97fps. Rendering out to 4:2:2 and DV simultaneously, this thing is moving at a fast clip. I'm rendering 9 hours of HDV at one shot, and started the render early this afternoon so it would be done by late tonight. It's moving along so fast, I suspect it will be done with both renders in a quarter the expected time.
The power of the new AMD multicores is pretty obvious at this point. This is the first AMD machine I've worked with deeply in over a year. The motherboard has some configuration issues, but after spending several hours with it and the Tyan tech support yesterday, it's been kicking up dust very nicely.
So, if you're looking into a new system for HDV....I'm bowled over.
We bought most of the parts from Colfax International, I'm not associated with them in any way, except they were recommended to us by a person at AMD. (oddly enough, Intel recommended them as well)

Back to post

Comments

GlennChan wrote on 6/13/2005, 8:52 PM

John, maybe your machine should be a little faster?

A 3.0E Prescott-core Pentium will do the original rendertest.veg / Vegas 5 in 89 seconds.
Prescott is about 6% faster than Canterwood Pentiums (Northwood), which are a few percent faster than Banias (533mhz FSB, no HT except for 3.06ghz model).
A 2.8ghz Canterwood should bench about 108 seconds. Your result is 123 seconds.
I have a 2.6ghz Canterwood which I forget the rendertest.veg result for, but I can double check my guestimate.

philfort wrote on 6/13/2005, 9:20 PM

On a 3 year old 1.7Mhz P4, I get a whopping 4:06 with the original rendertest.veg.

Looking forward to an upgrade...

rmack350 wrote on 6/13/2005, 10:17 PM

Canterwood is the name for the 875 chipset. Northwood is the name for that particular CPU core.

No P4s were Canterwoods.

Rob Mack

MRe wrote on 6/14/2005, 12:00 AM

Rmack350 wrote: "At that speed you could write each frame to a color printer and make a flip-book." LOL!!!

I tried that test also in my box with 3.0G P4 Northwood + 2 GB RAM and system configured so that OS, Vegas workspace, source material and final rendered AVI are on different physical SATA-disks.

It took approx. 3:30 to reach 90% limit and then I killed the process. Strange thing is that CPU usage was approx 50% during the render. It means that the bottleneck is somewhere else. It cannot be disk, though. Pagefile-usage was 1500 MB, which is also very strange.

"Something is rotten in the state of Denmark, er... Vegas".

StormCrow wrote on 6/14/2005, 7:22 AM

I just ran the old rendertest on my OC'ed P4 3.0 (running at 3.6GHz) with settings for NTSC DV, AVI, Good and came up with a time of :45 seconds.

MRe wrote on 6/14/2005, 10:29 AM

00:02:39 here. Std PAL Widescreen template.

GlennChan wrote on 6/14/2005, 2:39 PM

Canterwood is the name for the 875 chipset. Northwood is the name for that particular CPU core.
Looks like I probably got the Pentium naming system mixed up. I always thought the C in "2.6C" stood for Canterwood.

GlennChan wrote on 6/14/2005, 2:45 PM

StormCrow:
Most of the results posted for rendertest.veg are for "best" rendering quality. In the file --> export settings, you have to set it there because they override the project settings.

It took approx. 3:30 to reach 90% limit and then I killed the process. Strange thing is that CPU usage was approx 50% during the render. It means that the bottleneck is somewhere else. It cannot be disk, though. Pagefile-usage was 1500 MB, which is also very strange.
If you have a hyperthreading processor, Windows will see two processors. On particular renders it may be that one thread has a lot more work than the others, so you are always waiting on that thread and everything else is idle. So, only one processor in a HT CPU is working away. Windows sees 2 processors, hence 50%.

StormCrow wrote on 6/14/2005, 4:06 PM

1:15 time when set to best, NTSC DV, AVI.

rmack350 wrote on 6/14/2005, 9:32 PM

I spend most of my work day poring over this stuff to the point that I now forget more than I retain, but I think that's an indication of stepping. Seems to ring a bell and there was something significant about the C step northwoods.

Oh Yeah. Intel added HT and 800 MHz FSB at that point.

Good Times...

http://www.tomshardware.com/cpu/20041221/images/cpu_table_intel_big.gif

Rob Mack

rmack350 wrote on 6/15/2005, 11:01 PM

I ran the render test in two passes last night. For the first pass I pulled the motion blur down to zero and rendered to a new track as Sony YUV. Then I turned the Motion blur back up to 20 and rendered to DV AVI.

The motion blurred render only took about 20 minutes for me, rather than the infinite render when trying to do it all at once.

There were lots of other problems with this method but render time wasn't such a show stopper.

Rob Mack

TheRhino wrote on 6/25/2005, 12:10 PM

ATHLON X2 REPORT. . .

I just upgraded one of my Workstations to an Athlon X2 4400 on a DFI NF3 Ultra-D. This setup needs some tweaking and everything is at default/stock settings. [Windows XP Pro installed]

Vegas Video 6.0b Tests [fresh install, no tweaks, no preview window]
Original Rendertest = 0:40 in "best" mode AVI DV Codec
Original Rendertest = 0:25 in "good" mode AVI DV Codec
Original Rendertest = 0:41 conversion to mpeg2 "best" setting

GlennChan wrote on 6/25/2005, 12:25 PM

TheRhino:
Try re-installing your RAM into the color coded slots. That may be why you aren't running dual channel?

Spot:
Are you going to post your results for the original rendertest?

TheRhino wrote on 6/25/2005, 12:30 PM

I updated my post, I forgot to turn off the preview window.. . [duh]

Now we're screamin'! Also my dual channel is working and I updated the numbers to reflect.

GlennChan wrote on 6/25/2005, 12:41 PM

Your render time just dropped 36s.. I find that hard to believe!!

TheRhino wrote on 6/25/2005, 12:48 PM

I was trying to figure out why my results weren't better than a single processor system. . . Then I read that Vegas 6.0 doesn't use the second processor if you have the preview window on. I turned off the preview window, swiched out the memory so that the dual channel works, and now Vegas is taking full advantage of the system.

I knew something was wrong when I could get two instances of Vegas running simultaneously to perform almost as good as a single instance running by itself. . .

Motoxpress wrote on 10/19/2005, 12:22 PM

Just an FYI on this thread. I recently built an X@ 4200+ system for a client and ran the Vegas rendertest on it and I came up with :40 at standard speed and :36 OC'd 10%

This also has a WD Raptor as a system drive and I think it adds to the results. I am going to upgrade my own dual Opteron system to 265s or 270s in the next while and I will post those results as well.

-mx