Sunday, August 28, 2011

PowerDirector 9 Ultra64 video performance on Intel Q6600 vs AMD Phenom II X6 1055T

I recently decided to review some of my older video and audio footage. I have been using several programs in the past for video editing including multiple versions of PowerDirector and Pinnacle Studio .

This year, Cyberlink released a native 64-bit upgrade to PowerDirector called PowerDirector 9 Ultra64. Earlier this year, I also built a new machine based on a new AMD Phenom II X6 1055T CPU. This is a 6-core machine. It had been mostly sitting idle in a corner even though it was hooked up to my KVM switch, as I was still mainly using my older machine based on an Intel Core II Quad. With prices of RAM and hard drives plunging, I splurged for some more RAM for the new 6-core box and upgraded it from 4GB to 16GB. I also added a 2TB SATA drive. And I bought the upgrade for PowerDirect 9 Ultra64. I proceeded to install the Win7 64-bit Ultimate OS on the new AMD box.

Here are the full specs for the two machines :

Old (2007) New (2011)

Motherboard Gigabyte GA-P35-DS3R Gigabyte GA-890PA-UD3H
CPU Intel Core 2 Quad 6600 AMD Phenom II X6 1055T
CPU cores 4 6
CPU clock 2.4 GHz 2.8 GHz
RAM 8 GB 16 GB
RAM type DDR2 DDR3
RAM clock 800 MHz 1333 MHz
Video card XFX 9800 GT XFX 9800 GT
Video card RAM 512 MB 512 MB
Video bus PCI-E x16 PCI-E 2.0 x16
Disk controller Intel ICH9R SATA AMD SB850 SATA
Hard disks 2 x 1.5 TB Seagate Raid 0 1 x 2 TB Seagate

Software-wise, both systems are running the same Windows 7 Ultimate x64 with all the latest patch levels as of this week-end, same 280.26 of the nVidia video drivers, same revision 2930 of PowerDirector 9 Ultra6.4 Of course other drivers such as chipset and storage vary.

I trusted that the new machine would be be significantly faster at video production, given the 6-core vs 4-core CPU, and the higher clock speed. The older Intel Core 2 Q6600 CPU also lacks some instructions that are in the Phenom II X6, such as SSE4, which PowerDirector 9 Ultra64 can take advantage of.

However, I like to trust but verify, so I spent a few hours on some performance tests, which were quite surprising. My test consisted of the latest video clip I produced . The source file was 589 MB in size, in AVCHD format, from a Canon HG21 1080p HD camcorder, and the clip was 3 minutes and 22 seconds long.

I proceeded to do some renderings. The new machine didn't feel like it was significantly outperforming the old one, if at all, which is the reason I set the performance tests in the place. I figured this might be due due to the faster RAID 0 storage in the Intel-based machine. So, I tried to eliminate this variable. I downloaded and installed the DataRAM RAM disk, which is free for RAM disks up to 4 GB. I created a 4 GB RAM disk.

I copied the source file to the RAM disk on both machines. I set PowerDirector to also output the renderings to the RAM disk. Monitor preview was turned off. Nothing else was running during the tests on either system, except OpenOffice on the Intel box to write down the rendering times. Here are the results I collected. All values are rendering times in seconds, so the shorter the better.

AVI MPEG-2 BD 1920x1080 H264 1080p H264 1080p WMV/LAN Best MP4 Best MP4 Quicktime Full HD
Fast rendering

Yes No
Yes No
H/W encoder

Yes No
Yes No

AMD – PD8 3022 169 179 177 413 167 N/A N/A N/A
AMD – PD9 U64 2930 375 376 376 376 376 376 376 630

Intel – PD8 3022 204 210 252 599 203 N/A N/A N/A
Intel – PD9 U64 2930 192 193 195 451 193 307 654 829

As you can see, the new 6-core AMD box is significantly slower than the old 4-core Intel box in most tests. The AMD is almost half the speed, except for Quicktime and H264/MP4 with hardware acceleration disabled. There seems to be some bottleneck on the AMD box as most of the rendering times are the same, except for the Quicktime case. I have yet to figure out what it is.

I peaked with the task manager for a few seconds on both machines systems during some of the tests to check CPU utilization. The CPU utilization was usually much higher on the Intel than the AMD system. Sometimes it was only in the 5-15% range on the AMD system. Stil, even on the Intel box, I never saw the CPU utilization go much higher than 70%. Both systems have their power profiles set to "high performance". Both are running without a page file. Neither is overclocked.

Perhaps PowerDirector 9 Ultra64 is optimized more towards the Intel CPUs than AMD, but this still seems like something else is wrong. I haven't run other benchmarks other than Windows experience tests, which show 6.8 for Intel and 7.2 for the AMD machine in the CPU areas. I will try to find some and run them. Perhaps there are some BIOS power management options that affect the AMD performance as well. I left most of the advanced options as defaults.

Update: I ran Sisoft Sandra benchmark software. The performance is as expected for both the Intel and AMD box. The AMD box is about 50-70% faster in most tests, as one would expect. So this issue would appear to be specific to PowerDirector.

Update 2 : I downgraded from PowerDirector 9 Ultra64 build 2930 to PowerDirector 8 (32-bit version) build 3022 . I reran my tests and added the results to the table above.

In summary :
On the Intel Q6600 box, PD9 Ultra64 takes 5 to 25% less time to render than PD8 .
However, on the AMD 1055T box, PD9 Ultra64 takes 110 to 120% longer than PD8, except for the non hardware-accelerated H264 case. Something is definitely wrong with PD9 Ultra64 on AMD.

No comments:

Post a Comment