• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Q6600 video encoding performance

synaptical

Limp Gawd
Joined
Aug 29, 2004
Messages
194
Just got my Q6600 set up this weekend, and I am not sure about the performance I'm getting.

First, there's a problem with the linux kernel in frequency scaling on these processors, where you can only run it at two speeds, 1600MHz and scaled to 2400MHz. With an overclock, it still will only scale to 2.4GHz unless you disable the cpufreq modules. Once you remove the modules though, it shoots right up to whatever you have it overclocked to.

For the "benchmark," I encoded a half hour of mpeg video from mythtv to mpeg-4 with Xvid4, using 2-pass encoding at average bitrate 1000 kb/s, and mpeg quantization. For audio, I just kept the original, which I think is a medium bitrate mp3, like 192 I believe.

I performed the first test with the frequency scaling active. All cores were reporting 1.6, and the encodings took about 30 minutes each pass, which to me seems a little slow for only half an hour of video. My old single core A64 at 2.2GHz could encode it about the same rate. The reported CPU speed rarely if ever went above 1600MHz. I think one time I saw the third core shoot to 2.4 for a second. I really wasn't watching it the entire time, though.

For test 2, I unloaded the scaling modules, and the CPUs were all reporting 2.4GHz. Encoding the video took about the same amount of time - about 45-50 minutes for both passes. Again, that seems really slow to me. I will test again and be more precise, but for the first run I just wanted to get a general idea.

Compared to my 3200+ I was expecting more of a performance increase, but I'm not an expert in video encoding and so I'm not 100% sure what to expect. But I thought even at a lower frequency, the 4 cores would divide the processing and probably give a 20-25% performance increase at least. I definitely notice a more robust desktop and improved multitasking. Are there any more precise benchmarks I can do in linux? I would really like to find out the full potential of this chip.
 
Well, I only encode using x264 but I know that x264 is multithreaded. Encoding a 2hr 1280x720 video takes about 1 hour and 15 min per pass. It sounds like your encoder isn't multithreaded..even at 2.4ghz, with a multithreaded encoder 720x480 videos should encode at 40-60+ fps. The few dvds I've converted chugged away at 120fps (on an oc'd Q6700). I'd check on the encoder first.
 
Definitely check to see if the encoder is multithreaded and how multithreaded it is. There are still some that do not use more than 1 or 2 cores even when you have more cores than that available. Normally that is going to be where your problem is.

Also make sure you are using the exact same quality settings that you used with the other processor. If you're using higher quality settings with the new setup, the results won't be directly comparable.

 
I can typically encode an entire movie in about 20-30 minutes on my Q6600. I'd venture to say the program you are using is not multithreaded.
 
I think it is to do with the codec. I'm using avidemux, and I don't seem to have a lot of encoder options. So far, these are the results I have:

MPEG-4 ASP (lavc) - single pass, constant quality - 120 fps

MPEG-4 AVC (x264) - single pass, constant quality - 64 fps

MPEG-4 ASP (Xvid4) - single pass, mpeg quantization - 33 fps

MPEG-4 ASP (Xvid4) - two pass, avg. bitrate 1000 kbps - 38 fps

So it looks like "lavc" is the only multithreaded codec so far. I would just use that, but it will only encode single pass. Any attempt to do 2-pass crashes the whole program with some error about using "too high or too low a target for 2 pass." I guess I need to learn more about encoders and probably mencoder or some other program. thanks all.
 
Back
Top