Buying a new notebook as editing system(need hardware tipps)

[the Black Monarch]

*wanders back in, head no longer hung in shame*

Sorry about that. The actual reason why most quad-cores are a waste of money right now is that Virtualdub, which will be your primary consumer of CPU cycles, does not efficiently utilize multiple cores:

http://www.pcper.com/article.php?aid=63 ... pert&pid=8

Notice than in SSE4, the Q9650 outperforms the E8500 by only about 10%, despite costing 63% more on Newegg. The i7 chips fare much better, but motherboards for them are much more expensive than their LGA775 counterparts.

[Kariudo]

It has less to do with vdub and more to do with the codec itself.
I'll do some test encodes tomorrow (q8200 @2.33. I'll do runs with 3 cores disabled, 2 cores disabled and no cores disabled)

Q9400 is about $25 more than the E8500, with intel's VT
Q8200 is about $30 cheaper, but no VT

both surpass the e8500 in rendering out of vegas link

It's surprisingly hard to find comparisons on video encoding

[the Black Monarch]

Q9400 is about $25 more than the E8500, with intel's VT
Q8200 is about $30 cheaper, but no VT

And both will be slower than the E8500.

I'll do some test encodes tomorrow (q8200 @2.33. I'll do runs with 3 cores disabled, 2 cores disabled and no cores disabled)

Oh, that should be quite unnecessary. Let's not drag this thread out any more than we need to.

(for the record, I just did a mini-test on my dual-core system, and CPU load stayed consistently between 55% and 65%, with neither core ever experiencing full load. Latest versions of both Vdub and XviD were used.)

[Kariudo]

Q9400 is about $25 more than the E8500, with intel's VT
Q8200 is about $30 cheaper, but no VT

And both will be slower than the E8500.

boy, you just like to pick and choose your information don't ya?

I'll do some test encodes tomorrow (q8200 @2.33. I'll do runs with 3 cores disabled, 2 cores disabled and no cores disabled)

Oh, that should be quite unnecessary. Let's not drag this thread out any more than we need to.

(for the record, I just did a mini-test on my dual-core system, and CPU load stayed consistently between 55% and 65%, with neither core ever experiencing full load. Latest versions of both Vdub and XviD were used.)

Not at all, it is in the name of denouncing the perpetuation of computing myths after all!
Onwards I say.

all encodes were the same 2000 frames from the first episode of Gungrave (960x480)
The most recent versions of each codec (and virtualdub) were used
Defaults for all codecs were used, with the exception of options that allowed multithreading.

Vdub doesn't hand me the CLI parameters it passes to xvid, so the meGUI xvid test was done with parameters as close to Vdub's as I could get. The CLI parameters from meGUI were fed into xvid_encraw CLI

divx didn't like 960 x 480 on the home theater (ie, default) preset, so encoding was done at 720x480 for divx

Divx 6.8

Code: Select all

quad core: 50 seconds
dual core: 62 seconds
single core: 82 seconds

%improved 
64% (single to quad)
32.26% (single to dual) 
24% (dual to quad)

Huffyuv 2.1.1

Code: Select all

quad core: 21 seconds
dual core: 47 seconds
single core: 64 seconds

%improved
204.76% (single to quad)
72.97% (single to dual)
76.19% (dual to quad)

Lagarith 1.2.1

Code: Select all

quad core: 19 seconds
dual core: 48 seconds
single core: 84 seconds

%improved
342.11% (single to quad)
75% (single to dual)
152.63% (dual to quad)

Xvid 1.2.2

Code: Select all

quad core: 41 seconds
dual core: 69 seconds
single core: 122 seconds

%improved
197.56% (single to quad)
76.81% (single to dual)
68.29% (dual to quad)

Xvid 1.2.2 (using meGUI)

Code: Select all

quad core: 43 seconds
dual core: 75 seconds
single core: 102 seconds

%improved
137.21% (single to quad)
36% (single to dual)
74.42% (dual to quad)

Xvid 1.2.2 (using xvid_encraw CLI)

Code: Select all

quad core: 45 seconds
dual core: 69 seconds
single core: 99 seconds

%improved
120% (single to quad)
43.48% (single to dual)
53.33% (dual to quad)

Divx's enhanced multithreading shaved about 4 seconds off the quad core encode and 10 seconds off the dual core encode.

Vdub always utilized more of every core available than the CLI...save for some of the single core tests where even the cli completely loaded the core.

...I'd say vdub handles multithreading just fine, and that a quad core is definitely worth it for encoding

[NeoQuixotic]

Virtualdub also has an option for multithreading, at least in the experimental builds. It is off by default, but I really only use Vdub for processing Lagarith files and Lagarith already is multithreaded itself. And then I use MeGUI for encoding H.264 MP4s and I've been able to max out a dual quad Mac Pro @ 3.0Ghz running Windows XP Pro.

[Kariudo]

the threading option is in the latest release of vdub
setting vdub's threading to 1 gave some improvement in my tests, usually a few seconds. I only tested threading = 0 with xvid though

obviously, threading was = 0 for all the single core tests

Code: Select all

xvid (threading = 0)
46 (quad)
68 (dual)

xvid (threading = 1)
41 (quad)
69 (dual)

[the Black Monarch]

I started suspecting shenanigans as soon as you mentioned "CLI" and "meGUI". Just what the hell were you doing?

Then I saw this:

152.63% (dual to quad)

At first I was like

But then I was like


Leave the benchmarking to the professionals.

[Kariudo]

I started suspecting shenanigans as soon as you mentioned "CLI" and "meGUI". Just what the hell were you doing?

Benchmarking of course. You did mention something about vdub not being able to handle multithreading efficiently if I recall, so I went and compared vdub to a CLI (which would show how efficient the codec itself is at multithreading) and another widely-used GUI (for comparing the efficiency between the two GUIs).
It's really pretty standard stuff ya know...having controls in your experiments that is.

Then I saw this:

152.63% (dual to quad)

At first I was like

But then I was like

Yeah, it was hard for me to believe that the Lagarith encode time jumped from 19 seconds to 48 seconds when going from quad -> dual, so I repeated that encode a few times to make sure it wasn't a fluke (and it wasn't).

You're starting to sound pretty desperate here ya know, completely disregarding all of the other results I posted. The trend showed that going from dual -> quad for encoding gives similar gains as going from single -> dual (which was essentially what we were arguing about in the first place).

...Now that I take another look at it, my percentage equation is off.
It's the speedup that I was calculating there, but thinking it was percentage and not speedup, when the numbers looked too high I changed the equation from (Dual/Quad)*100 to ((Dual/Quad)-1)*100.
Speedup is (Dual/Quad), so the lagartih speedups are:
2.52X (Dual to Quad)
1.75X (Single to Dual)
4.4211X (Single to Quad)

and so on and so forth for the other results

Leave the benchmarking to the professionals.

And perhaps you should leave this forum to people who actually know what they're talking about

[milkmandan]

You will want:

Windows XP Professional (Home can't handle multi-core and Vista just plain sucks)
The fastest dual-core CPU available (XP Pro can only use 2 cores)
A high-resolution screen
DDR2-1066 or faster memory

Not much else matters. You probably won't be using more than 2GB of RAM, nor will you notice any improvements in speed past DDR2-1066. The video card is wholly irrelevant, as it's not where the rendering and encoding are done. You're using an external hard drive, so HD size is irrelevant. Just get that 3.33GHz Wolfdale and a big screen, and you're gold.

stop digging up old threads and posting misleading information, this is like trolling+necroposting.
XP Pro supports quad cores.
your suggestions are even detailed. "high resolution screen"? wut?

[the Black Monarch]

Well, I WAS going to continue to trust the results of impartial professionals rather than those of an amateur who had a debate to win and whose results included some extremely anomalous-bordering-on-ridiculous data, but your airtight ad hominem attack has convinced me otherwise!

Anyway, the matter about XP supporting more than 2 cores was taken to NeuralClone over at Ex Isle. This is a guy who really, really knows his sh!t, and even he wasn't sure what the core limit was, but confirmed that XP Home was limited to 1 physical CPU, and XP Pro limited to 2.

your suggestions are even detailed. "high resolution screen"? wut?

Yes. As in higher = better, but there is no "magic number".