Render Test HDV benchmark tests...weird flucuations

[Alex Thames]

I recently just put together a brand new desktop computer.

Specs:

CASE: Antec 900
MOTHERBOARD: Gigabyte GA-P35C-DS3R LGA 775 Intel P35 ATX Intel Motherboard (P35 mobo)
CPU: Intel Core 2 Duo Allendale E4300 1.8ghz 800mhz FSB 2mb L2 shared cache
MEMORY: Crucial Ballistix 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 1066 (PC2 8500) Dual Channel Kit Desktop Memory
GRAPHICS CARD: MSI Nvidia 8600 GTS (heatsink version)
AUDIO CARD: none (onboard sound)
POWER SUPPLY UNIT: Antec Earthwatts 500W
HARD DRIVE x1: Seagate Barracuda 7200.10 ST3500630AS 500GB 7200 RPM SATA 3.0Gb/s
FIREWIRE PCI CARD: SYBA IEEE 1394a FireWire PCI Card with Internal 9-pin Header Model SD-VIA-FW1E1H
AFTERMARKET COOLING: Scythe Ninja Plus B SCNJ-1100P 120mm Sleeve CPU Cooler (took out stock Scythe fan and replaced with Noctua NF-S12-800 120mm fan)
DVD BURNER: Lite-on 20x Lightscribe SATA DVD Burner

The Antec TriCool 120mm fans are all set on low, and the Antec 200mm fan is also set on low, while the Noctua fan is set at 12V at its fastest speed (800 RPM for the slow version).

So, I did my rendering to see how fast my computer could render in Vegas 7.0e. I closed all applications and only had background processes running (about 35-40 processes, which use 0% of the CPU processing power if I'm doing nothing and just letting the computer idle). No anti-virus, no firewall, no spyware killers, etc.

I opened up Vegas 7.0e and tried rendering the test render-HDV file someone had created. The fastest time I've seen was with someone's quad core Intel QX6700 2.66mhz, which rendered the file out in 2 minutes 0 seconds. He was able to repeat this a few times I think.

The first time I tried rendering the benchmark file, I ended up with 3 minutes 16 seconds, which I'm quite pleased with considering that I'm using the lowest end Core 2 Duo CPU. It was rendering at 61% of the quad core.

Note: my hard drive is partitioned into two drives. C Drive is 50 gb, D Drive is 415gb. I have around 50gb free total left because I copied a lot of files over from my external drive.

Then I installed antivirus, firewall, spybot, etc. I tried rendering with and without these programs on.

The following are the times I've gotten:
1st Trial: 3:16
(no antivirus, firewall, spyware, etc.) I wasn't paying too much attention, but I think it was at 100% CPU usage for the entire test.

2nd Trial: 3:20
(with antivirus, firewall on...this is acceptable, but I noticed that the CPU usage went down to 99% for a quick moment)

3rd Trial: 3:40
(with antivirus, firewall on...this was much longer than expected, and I suspected that CPU usage went down considerably)

4th Trial: 3:40 again
(with antivirus, firewall on...this time I watched the CPU usage and I noticed it would drop to around 50%, sometimes 60% and 70%ish for several seconds, then jump back up to 100%, then after awhile, go back down to the high 90s, then 100, then 50ish, etc.)

5th Trial: 3:18
(with antivirus, firewall on...this was a nice time, though not as fast as my 1st trial. I noticed the CPU usage was not at 100% all the time, but when it did drop, it was only for a few, not several, seconds and they only dropped to the mid to high 90s, never to the 50% level)

6th Trial: 3:39
(with antivirus, firewall on...again, horrible performance with the CPU dropping to the 50s a lot)

7th Trial: 3:39
(same thing as 6th Trial)

8th Trial: 3:39
(again...)

One thing I noticed on the 8th trial that I wasn't paying attention to on the previous trials was the graphic chart of the CPU usage history showed the two cores...the core on the left was displaying much lower CPU usage than the right core when there was a drop. For example, if the CPU usage dropped to 50%, the left core would show that it's usage was much lower than the right core.

Here is a screenshot of what I mean: http://img256.imageshack.us/my.php?i...dertestgb9.jpg
It shows the 50% level and that the left core drops more than the right core.

I'm not sure why? Anyone have ideas? My hypothesis, from what I've read online, is that the Scythe Ninja heatsink fan is so heavy and since my motherboard is mounted vertically, the heatsink base is not coming into full contact with the CPU cores. I did apply thermal paste. So, I think gravity is pulling the heatsink fan down a bit resulting in better contact with the bottom (right) core than the top (left) core. That means the top (left) core is not getting cooled as much as the bottom (right) core, which may or may not explain the difference in core usage.

Would different cooling performance on the two cores affect render speeds? Would the CPU temperature in general affect rendering speeds if it is able to run stably? Note that I did not overclock anything. No CPU, no memory, no mobo, no graphics card OCing.

But my MAIN question is, why does the CPU usage drop to the 50% level (or even the 60-70-80% levels) at all? It slows down render speed considerably, and Vegas being able to use multiple cores should be using the CPU at 100%. 93% and up for a few seconds might be acceptable, but when it drops to the 50% level for up to 10 seconds, it really kills the render time.

From the trials, you can see that difference from the fastest time (3:16) and the slowest time (3:40) is huge percentage wise. I can only imagine how much longer it will take rendering out my big projects, 2-3 hours of HDV 1440x1080 if the CPU usage drops to 50% this much.

Out of eight trials, only 3 times was the CPU able to reach optimal/near optimal peaks (mid 90-100%). This seems to suggest that more times than not, the CPU will drop to 50%, and more times than not, I will get a slower
render time.

How can I fix this? And why does it do this?

Thanks!

EDIT: Just did a 9th Trial: 3:41, the slowest time yet. It lingered in the 50% level several times, lasting anywhere from a few seconds to 11 seconds. I'd estimate the time it spent at the 50% level to be for about 30 seconds total during the entire 3 min 41 seconds.

And a 10th Trial: 3:41 again, the slowest. It dipped into the 50s four times, each time lasting around 10 seconds for a total of 40 seconds (estimate).

I don't know what to do to get it back to 100% consistently.

[Jon McGuffin]

Alex, are you running the test that was posted at the Sony Website Forums where a bunch of people chimed in?

I suspect your problem is probably related to only have one disk drive and the antivirus crap installed. Regardless of how many partitions you have, all NLE software likes to work with a physical disk that is seperate from the system drive. That of course doesn't explain why your first results were good and the rest were so much slower though. The only other answer might be the applications you installed after the fact.

[Alex Thames]

Yes, it is the one at the Sony Forum.

But, the thing is, even with the Antivirus, etc. the computer was still able to render near 3:16-3:20 a few times. Yet, all the other times it was closer to 3:40. So, that doesn't make sense to me.

Also, even though I only have one drive, the first test I did got an amazing 3:16...so, that seems to suggest it isn't being hindered by the fact that I only have one hard drive?

[Jon McGuffin]

Are you overclocking the CPU? Just for fun, try defragging the hard drive and then rebooting and run the test again...

Jon

[Jon McGuffin]

Alex, I can't seem to find the link to the Vegas file to run. Can you point me to where to download the file and I'll try it myself. I'm running a similiar configuration to yours except my processor is an E6600. I'd like to give it a whirl. Can you tell me step-by-step the procedure you do to render the file?

Jon

[Alex Thames]

I haven't overclocked anything. I do plan to later when I upgrade to quad core Q6600, and then when I give this CPU now to someone else, I'll want to OC it for them too.

But for now, everything is at stock speeds.

Sorry, I'm not very technical. How do I defrag? I've actually been wanting to do that.

The step-by-step instruction is open up the .veg file with HDV project settings. HDV-1080-60i template (1440x1080 29.970 fps). Then close all background applications (including antivirus, etc. for best render times, even though I didn't do that for some of my trials). Then render out as .wmv using the 8mbps template (the best format).

I'm not sure where I downloaded the .veg file, but it's small, only 18kb, so I can email it to you.

Please send me your email address.

Also, I just did an 11th trial: 3:30 (CPU did drop to 70%, better than 50% but still...also, I had all my messengers running and was even browsing the web during render this time, so...strange).

[Mike Kujbida]

Alex/Jon, my guess is that the veg file you want is in the first post in the NEW Rendertest-HDV.veg thread.
BTW, I came in at 1:59 (119 seconds) with my QX6700 :-)

[Jon McGuffin]

Originally Posted by Mike Kujbida

Alex/Jon, my guess is that the veg file you want is in the first post in the NEW Rendertest-HDV.veg thread.
BTW, I came in at 1:59 (119 seconds) with my QX6700 :-)

So much for Quad Core's being TWICE as fast! :-)

[Jon McGuffin]

Okay, I followed the directions and got 2:26. So how can my E6600 be so fast compared to the Quad Core at 1:59?

I rendered out to .wmv -> 8 Mbps HD 1080-30p Video as suggested...

[Alex Thames]

Like I said, I was quite pleased with the performance at 3:16 (or even 3:20) compared to the 2 minute (or 1:59) quad core speeds.

This proves that while quad core is indeed faster, it doesn't scale that efficiently. I'd estimate quad core to be around 30% faster render speeds for Vegas 7.0e than the same specs, but dual core (so same FSB, same clock speed, same cache, except of course no additional 2x).

But if the CPU usage drops to around 50% as it often does for me, then the render times slow down about 17%, which is significant.

Any ideas to fix this so that it will be at 100% CPU usage all the time while rendering?

[Jon McGuffin]

Sounds as though something is running the background that is screwing you up. I can't think of what else it could possibly be. I know that you ran the test after installing the spyware, antivirus stuff but maybe they installed something that is still going in the background.

If it's really bothering you, I'd probalby go ahead and uninstall all the other applications that you installed. Defragment the hard drive by going to My Computer, Right Clicking your hard drive, then clicking on TOOLS and then defragment from there. Then reboot the computer and try again.

I'd be really curious to see the kind of performance improvements you'd see from overclocking that CPU. Do you have plans to do this?

Jon

[Jon McGuffin]

I thought I had read in the forums that people were getting about a 90% increase in render speed by going with Quad Core. Somehow Vegas 7.0e was able to fully take advantage of the additional two cores. Perhaps I was wrong. At this juncture, $500 for a CPU that is only 10-15% faster is certainly not worth it considering I can probably overclock my CPU and get some of that performance back.

Frankly, rendering times really don't concern me that much. Playback performance on the Vegas timeline does. I work with HDV and I hate having to preview choppy video.

[Mike Kujbida]

Originally Posted by Jon McGuffin

Okay, I followed the directions and got 2:26. So how can my E6600 be so fast compared to the Quad Core at 1:59?

I rendered out to .wmv -> 8 Mbps HD 1080-30p Video as suggested...

From the thread I referenced earlier:
...it is intended to be rendered to an MPEG2 HDV file using the default "HDV 1080-60i" template.

The time I quoted was for that setting.
My WMV (8 Mbps HD 1080-30p) render time was 1:11 and AVI (HDV 1080-60i intermediate) was 2:09.

[Jon McGuffin]

Okay, I followed what Alex had posted above going to .wmv file. So this makes more sense.

Alex, what setting *are* you using to encode this file? I'd like to bench yours versus mine apples to apples....

[Alex Thames]

Sorry, I must have remembered the instructions wrong then. I was encoding as .wmv, but I'll try it as .mpeg2 as well.