Encoder speed in DVD-RB using single/multiple encoder processes

[UncasMS]

Having upgraded from single to dual core recently i was interested what difference it would possibly make to use DVD-RB Pro's option:
"Multiple Encoder Processes"

I used Harry Potter and the goblet of fire - 226.270 frames/150 min on an AMD Athlon 64 x2 4600 AM2, 2GB Ram with 2 out of 3 sata hdd for encoding and DVD Rebuilder Pro 1.22

Encoder       Single Encoder Process   Multiple Encoder Processes
CCE           38 min                   37 min   
HC 020        202 min                  115 min  
Procoder 2    125 min                  127 min

CCE was used in 2 pass mode
HC 020 with settings BEST
Procoder 2 was set to MASTERING Quality


CCE is known to use all cpu power it can get hold of without having to tinker with it and demonstrates that it's code is up 2 scratch.

Procoder2 doesnt seem to like interference as well and is best left alone running only once instance at a time.

Hank315's latest version of his great freeware encoder HC 0.20 clearly gains in terms of speed from the rebuilder option "multiple encoders" and drops down to 57% of the time using two encoders instead of one!

[blutach]

I read somewhere that CCE couldn't use multiple process so the result is no surprise. But my experience with Procoder is enhanced times - need to do a test to see but I am very surprised at the result.

Regards

[BR7]

That's why I thought ProCoder couldn't run multiple process because when I have it running normally it uses 100% of my CPU Power. Being PC2 used all that power I just figured it wouldn't benefit running multiple process

[techreactor]

I agree with BR7

CCE uses 100% CPU in single instance of encoder
Procoder uses 100% CPU in single instance of encoder
AutoQmatenc uses 100% CPU in single instance of encoder
HCEnc does not use 100% CPU in single instance of encoder

Therefore using multiple encoder processes with HCenc will benefit and reduce the total time taken, for others the multiple encoder processes in DVD-RB pro is virtually useless.

[UncasMS]

hmm, made a small mistake here, folks

i ran cce overnight and didnt look at the cpu load - i have startet cce agian this minute and i must correct myself:

cce does NOT use all the cpu power it could get - c.f. attached screenshots
(and it of course does not run two instances of its encoder)


could someone provide us with a screenshot showing cce using almost 100%???

[jdobbs]

I'm really surprised at the reported times with HC v0.20. On my system (AMD Opteron 165 [dual core] overclocked) -- the HC times when set to "Multiple Encoder Processes" are very close to CCE. (within about 15-20% or so). Hank315 did s series of tests also -- and his showed much faster speeds than the number reported here.

My own experience was that ProCoder was getting improvements as well -- although nowhere near what you'd get in HC. But I haven't tested that in a while. I'll try it again.

"Multiple Encoder Processes" will also improve QuEnc encoding times. By the way -- it has no effect on CCE -- so the one minute difference was just the resolution of the timer.

[UncasMS]

i'll take notes of whatever i'll be testing next and post it here

btw does cce produce roughly 100% cpu load on your machines, jerry?

[jdobbs]

No. It uses both processors -- but typically uses about 80% of each.

[UncasMS]

ok thx

that's identical to my screenshot

[hank315]

Same movie, (PAL) Harry Potter, Goblet of Fire, 226270 frames (movie only)
Intel E6600, 2 GB ram

 
[FONT=Courier New]Encoder                          Multiple Encoder Processes[/FONT]
[FONT=Courier New]CCE basic                        43 min   [/FONT]
[FONT=Courier New]HC 020 best                      57 min  [/FONT]
[FONT=Courier New]HC 020 normal                    38 min  [/FONT]
[FONT=Courier New]Procoder 2 mastering quality     63 min[/FONT]

CPU load for HC and Procoder 100%, CPU load for CCE approx. 70%.
CCE speed is pretty constant at 7.0 during the encode,

[UncasMS]

hi hank,

thx for your times, which of course make me wonder quite a bit

i'll have another run with your encoder, hp4 and movie only this time (though that isnt responsible for the severe difference) and report back here

[UncasMS]

i have restartet hp4 once more with hc/rb in multi processor mode and it took some 15 min for ~11% so i dont get anywhere near your encoding times, hank

dvd-rb's idct + matrices are set to encoder default

hc uses all the cpu power there is, no process blocking as background task (c.f. screenshots)

any idea why my times are so much different from yours?

[hank315]

Hi UncaMS,

My settings are the same as yours so I encoded it again and looked at the speed in the same segments as in your screenshots.
Your AMD does 35 fps (first pass) and 40 fps (second pass), the Intel does 65 fps (first pass) and 75 fps (second pass).
That's a big speed difference...

BTW, the Intel is approx. 30% overclocked.

[UncasMS]

fair enough

[rahzel]

CCE is faster on my AMD X2 machine as well. i can do 3-passes with CCE to equal HC and Procoder with multiple encoder processes running. with the film Gladiator R1 NTSC, all 3 encoders were around 100 minutes with the best possible settings for each encoder, and 3-passes with CCE SP.

specs:
AMD X2 3800+ @ 2.5Ghz
2x1GB OCZ Gold PC4000 3-4-3-7