dual processor Videostudio 9

[blegate]

Hi all- I've been a Videostudio fan since V6. I recently upgraded by system from a AMD 2400+ to a Pentium D 805 (dual processor) system. I'm not seeing any stellar reduction in time with rendering videos from AVI to MPEG. I know I have dual processor enabled on my W2K box.

Videostudio 9 supports dual processor doesn't it? Are there any special registry entries, etc. I have to change?

I suppose it may be a tad bit faster but I expected "wow" results.

Just curious to know.
b

[Ron P.]

Hi Blegate, and welcome to the forums..

You would think that upgrading to a dual core cpu, the performance would be "wow". However others here have reported a moderate increase in performance with dual cores and Hyperthread singles. From what I read, you might see about a 30% increase.

You are correct that the programs need to be written for dual core to utilize it, and AFAIK VS9 was not. VS9 was released prior to dualies.

However the latest version VS10+ does support dual core and HT.

Have a look here:

http://www.ulead.com/vs/sysreq.htm

Ron P.

[blegate]

Are you sure about this? I upgraded to version 9 a while back because the specs back then mentioned support for dual processor- meaning the code was written to take advantage of it.

BTW dual processors have been around a very long time... It's only until recently that dual cores have been developed to fit in the same size/format has a regular processor. Correct me if I'm wrong.
b

[Ron P.]

Yes, Dual Processors have been around for several years. However the Dual Core processors have not. I tried looking for the specs on VS9, however Ulead no longer has that page available. I can't fnd anything in the manual on it.

Ron P.

[THoff]

blegate, I recently did some tests for someone on another forum to help him predict the performance difference of going from his current system (a 2.1GHz Celeron) to a P4 640 3.2GHz processor with hyperthreading.

Here is my post, verbatim:

I did some benchmarks on my system using Videostudio 9 for you.

The processor in my video editing system is a P4D 840EE, a dual-core, hyperthreading-enabled processor running at 3.2GHz with an 800MHz FSB. So aside from the fact that it appears as four processors in Task Manager, it is comparable to your new system.

I took a Type 2 DV AVI file from my GS400, marked a six minute section (exactly) of it, and transcoded it to a VCD MPEG1 file using the default settings in Videostudio.

I performed the conversion four times, once using using just Processor #0 (Single Processor), once using Processor #0 and Processor #2 (Single Processor + HT), once using Processor #0 and Processor #1 (Two Processors), and once using Processors #0 through Processor #3 (Two Processors + HT).

The processor selection was made by changing the Processor Affinity Mask in Task Manager. The two physical cores appear as Processor #0 and Processor #1 in the Windows XP ACPI HAL, so by changing the PAM to allow execution only on Processor #0 and Processor #2, I can simulate a single-core hyperthreaded processor.

Anyway, without further ado, here are the results for transcoding a 6:00.00 DV AVI file to VCD MPEG1 using Videostudio 9:

Single Processor: 6m03s
Single Processor + HT: 5m52s
Two Processors: 5m49s
Two Processors + HT: 5m16s

This is essentially real-time, with slight improvements as you add processors (physical or logical). The only surprise is the performance jump in the final run, I did not expect as much of a jump in throughput.

I also ran the last test using Videostudio 9's bigger brother, Mediastudio 8 Pro:

Two Processors + HT: 3m52s

As you can see, HT made only a slight difference (11 seconds), as did having two cores (14 seconds). I also didn't see the processor utilization approach 50% at any time, so it isn't surprising that additional processors didn't help a great deal.

I have seen the utilization hit above 50% when transcoding to WMV, and I've also seen higher utilization when doing 2-pass VBR encoding, and I do feel that getting a dual-core processor is worthwhile and a better choice than a HT processor.

[maddrummer3301]

.

[THoff]

It depends on the algorithm and how the code is written.

Encoding DV video can easily be written to take advantage of multiple CPUs, because DV uses intra-frame compression, i.e. each frame is compressed independently of the previous one. Thus, a programmer can spin off one thread for each of the CPUs in a system and have it encode a frame.

For MPEG, it's a different story, because it uses inter-frame compression where the current frame generally only encodes the differences from the previous one. This continues for a while until the current GOP (Group of Pictures) is closed, and a new keyframe/independent frame is encoded, which describes an entire frame of the video.

Because most MPEG frames depend on the previous one, the encoding is harder to parallelize, but it isn't impossible. Detecting the differences between one frame and the next could be one thread, and other thread(s) could then perform the actual encoding, and another thread would handle I/O.

In my experience, encoding WMV files will use 50 or more percent of the CPU cycles on my system, which has four logical processors. That indicates to me that Microsoft's WMV encoding engine will utilize at least two processors fully during encoding.

I noticed in my tests that Mediastudio Pro 8 was significantly faster encoding the VCD file, and I'll probably do some more tests to see how it scales. I ran the VCD test with Videostudio because that is what the person I was testing this for had asked about, and I ran the same test with MSP 8 once just for kicks.

[GuyL]

I did similar tests on VS9. I had an AMD 3000+ and updated it to a AMD 3800+ X2. I rendered the same file before and after the upgrade and saw little difference. I believe the difference was the increase in processor and nothing to do with the dual cores.

I've recently updated to VS10 and there is no difference between it and VS9 encoding the same file. So, Ulead can claim all it wants that it is optimized or supports dual cores. My measurements conclude that there is no difference.