MPEG slower on P4 ?

[KuroNeko]

I did a comparison check on the newer generation of mobo/CPU combo's. I used the DV file found at Teco's and the current CCE SP encoder demo.

The results I get are pretty weird - the 2.56GHz P4 system (Asus P4B533E with 1 GB DDR266 CL2) was thoroughly beaten by the Athlon 2100.

Especially notable was that the analysis run (3 pass encoding) of the 22 sec videofile takes about 23 seconds on all P4 systems (and Celeron 1,2GHz) I tried, with little gain by going higher clocked CPU's, whereas the Ahtlons nicely went down in time as CPU speed increased.

I tried Rambus, DDR and ordinary SDram setups, doesn't seem to make a difference, neither does the harddisk speed.

So now I am wondering if the Canopus DV codec used to decode the DV file is maybe causing the problem for the Intel systems, not being optimised for P4. The CCE SP encoder, according to CinemaCraft support, is supposed to perform optimally on P3 (dual) or P4 machines, but I see the Athlon being a lot faster.

Does anyone here use CCE SP and/or Canopus DV codec regularly and can comment on this?

Thanks

Neko

[keithm]

perhaps the athlon is just a better chip

[LvR]

You perhaps running a "B" step P4 there?.

Sort of a stink was mentioned (Inquirer or Register?) not too long ago about those - a problem existed that very few people would run into in real life - perhaps you are one of those?. Suggested we replace our B steps with Cs

EDIT: OK it was the Inquirer - lots of links from here.

[KuroNeko]

Keith, the Athlon may be a better chip, but all benchmarks show the Athlon to be appr. as fast as the P4 (Athlon 2000+ vs P4 2GHz), with some variations in particular areas, but none as large as I have noticed in my tests.

I do recall a discussion about difference in file size when rendering on P4 or Athlon, indicating different calculations used (slight variations). I hope the Athlon isn't taking shortcuts that lessen quality - I only speed tested, didn't do quality comparisons

As for the B-step, I tested various P4 setups (about 12 I believe), with some very recent 2.56GHz machines. I doubt these are still B-step versions, but didn't check every CPU. I'll have a look, but from the INQ article I don't think this is the problem. It's as if something in the setup (Canopus DV codec, CCE SP) is actually holding back the P4, perhaps by using an algorithm particularly unsuited to the P4's long pipeline... I am not knowledgeable on these issues, that's why I'm asking here.

Neko

[wkulecz]

Welcome to the real world where the only "benchmarks" that matter are the applications you actually need to run!

Try Doc's MSP vs. Premiere rendering and mpeg encoding test.

Here's a page with some results:

http://www.hyperactivemusic.com/MS%20Pro%206.0.htm

--wally.