CCE Speed Issues

[TeeeRex]

Hi,

My first question to the forum actually after lots of replies so here goes ..

I'm using CCE 2.50 and AVISynth (via DVD2SVCD) on two PC's - one is a 1.6GHz P4 and the other is a 1GHz P4. Both systems are running XP Pro with no other services other than standard.

If I try a 4 pass SVCD VBR encode from a DVD RIP - with the 1.6GHz I get a max encode rate of 1.07 and with the 1Ghz I get 0.8 for each pass.

My question is this - if the encode is proportional to the processor speed - then why is the 1.6GHz not encoding faster ? Is there a bottleneck with AVISynth or is the bottleneck elsewhere in the process ? Can CCE be 'tuned' ?

Thanks,

TeeeRex

[snowmoon]

Your bottleneck is probably in you DVD decoding. Speed is also related to cache size and other nuanced of the processor. So apples to apples the speed difference will be perportional to the time your DVD2SVCD spends encoding vs doing io.

So I doubth their is a way to speed things up without compromising something.

[Nelson37]

Disk IO is still relatively the same in both machines, this is probably the main speed limiter in comparing the two.

Double check bus and memory speeds in the 1.6, these are often defaulted to a slower speed, and/or make sure that the 1.6 has the correct hi-speed mem chips.

[TeeeRex]

Nelson37 & snowmoon,

Thanks for the replies - I'll check the hardware settings tonight to see if I can increase them. I know it's not IO as I can RIP at 6x with no issues and I'm running RAID0 on 7500rpm drives!.

Do you know if there are any 'benchmarks' as such with typical hardware and a set software configuration - I've read in the forum some people are saying that can encode at 2.5x using CCE ? It would be nice to see average results so people know what to expect ?

Cheers..

[DJRumpy]

TeeRex, remeber your not going to see in increase comparible to MHz on your processors either. Just because your processor MHz has almost doubled, it doesn't mean your encode speed will double. Any benchmarks from processor to processor usually show small to marginal increases between processor speeds (i.e. 1.0 to 1.5, etc). Say 30 to 40 percent for a jump from 1.0 to 1.7 (this a complete guess, no flames please...). Your not going to get the full 70% increase in speed.

Disk IO is pretty much irrelevant when it comes to encoding MPEG-2. The bottleneck is almost all CPU/Memory/Bus Speed.

[dc91gt]

P4 has no barrel shifters - intels biggest mistake in years.

get an athlon.

-d

[vitualis]

That comment makes absolutely no sense at all.

It also doesn't help that the top P4s are somewhat faster than the top Athlons.

In any case, both CPUs are excellent and you can read up the differences between the platforms on many hardware sites on the net.

Let us not start another pointless CPU flame war...

Regards.

[Executioner]

Just for conversation sakes, I have a dual PIII-550MHz rig running Win2k pro with 328 megs of pc100 ram. When using CCE 2.50 with 4 pass, I get 0.6. A typical 2 hour movie takes about 16 hours from start to fininsh using DVD2SVCD. Most of this time is the encoding My best guess is the eicoding takes ~14+hours my my rig.

[LeeBear]

TeeeReex: Are you sure both system's are P4's? I don't believe Intel ever realeased a P4 CPU slower then 1.3 GHz, so unless you underclocked a P4 I doubt the 1 GHz machine is a P4, it's most likely a P3. If that's the case that's why you don't see a linear encoding time increase with the faster CPU. The P4 is MHz for MHz slower then a P3 chip (on most stuff anyways) ie. a P4 1 GHz is slower then a P3 1 GHz. That's probably why you're see the non linear speed increase.

-LeeBear

[TeeeRex]

Leebear - Yes you are correct the 1GHz is indeed a P3 (my company laptop) so my mistake...

I thought CCE & TMPGenc were optimised for a P4? (SSE2) - in which case wouldn't that reverse your logic ?

To be honest with you I'm not particulaly fussed at the difference in speed - I just wanted to know why there was only a marginal difference in encode times when one system is clearly a higher spec than the other by some 600Mhz.

Weird...

PS - I am right in thinking the that CCE 'speed' setting is RT - ie 1.0 is RT with 2.0 being double RT ?

[fantomlord]

PS - I am right in thinking the that CCE 'speed' setting is RT - ie 1.0 is RT with 2.0 being double RT ?

that is correct.

[offline]

That comment makes absolutely no sense at all.
It also doesn't help that the top P4s are somewhat faster than the top Athlons.

barrel shifters simply enable more data to be pushed through the
cpu in a given number of cycles. Intel developed them in the 386
but now have dumped them for the p4, seriously lowering performance.
Why? marketing. Without them, Intel can boost core speeds and claim the crown of the top cpu in terms of Mhz. Intel know that the Mhz rating
is still the way most people (incorrectly) assess a processors speed and
are happy to reduce performance to increase Mhz.

[snowmoon]

Intel is pushing SIMD instrucions at the detriment to some other instructions. SIMD is the way of the future and with proper compilers you CAN take full advantage of the benifity of the P4.

Can you blame them for trying to shift away from a programming methodology that will only hold them back in the long run. Look at how AMD is really starting to stuggle to keep the MHZ comming, intel is slowley but effectlvly turning up the MHZ every chance they get. P4's run cooler and more efficiently than AMD's too.

So what you call marketing many would call strategic planning for the future.

Dom't get me wrong I think that both chips have advantages. I personally would like to go back to AMD's but the VIA mb's are not worth the agrivation.

[offline]

AMD is really starting to stuggle to keep the MHZ comming

True... but a new patient has been registered by 3rd party researchers, overcoming the Ghz limitation.

So what you call marketing many would call strategic planning for the future.

I always approached strategic planning and marketing as two sides
of the same coin. Neither is focused on the best interests of the consumer.

Intel cut out a full speed full size predecode cache and put in a 16kbyte post decode cache for the P4. Is this way of the future or just a cynical
attempt to drop core size at the cost of performance for the quick buck?

SIMD is the way of the future and with proper compilers you CAN take full advantage of the benifity of the P4.

If you are using code that is optimized just for SIMD it is faster than generic code, but the same goes for athlon, and generic code runs faster on athlon than p4.

btw I use a p4 at work as the chipset is more stable. I don't use
athlon at all at the moment. Like you I think both cpu's have their advantages and disadvantages. I just find Intel to be less consumer
friendly than AMD.

[LeeBear]

Whether you like Intel or not you have to admit they did the right thing with the P4. While one can argue it's not the most effecient CPU desidn it's long pipeline allows it to scale very well. The P3 architect would never be able to reach the GHz speeds that the P4 is capable of, even AMD's Athlon processors are nearing the end of what the Athlon architecture is capable of. While AMD has resorted to releasing paper launches of it's CPU (cough, cough, Athlon XP 2800+) and increasing the FSB to boost performance, Intel will have no problems putting out 3 GHz chips, and the P4 architecture should easily scale to 4-5 GHz. As of now Intel's fastest P4 is faster then AMD's fastest Athlon, so they own the performance crown. Intel's CPU's also have the highest MHz rating so they own the marketing crown as well. It's a win win situation for Intel right now, and if AMD doesn't get there Hammer CPU's out soon they will be in trouble.

-LeeBear

[offline]

did the right thing with the P4.

for themselves.. sure. Early purchasers were completely screwed via cpu pin changes, lower speeds for their dollar, and memory support - not to mention intel's unnecessary changes to the ATX form factor.

P3 architect would never be able to reach the GHz speeds that the P4 is capable of

hmm.. how about a celeron (with p3 core) running at 3GHz?

and if AMD doesn't get there Hammer CPU's out soon they will be in trouble

true. sadly true.

[Nelson37]

Intel and AMD will continue to play leapfrog and hype their products, (A competition from which we all benefit), debating architectures is pointless, REAL-WORLD performance and Price are only valid comparison. In 6 months it will all be different.

Disk IO is a factor in encoding, and if identical in both systems, would tend to diminish processor speed gains. There may also be variants in the MOBO's and cache config, several possible factors.

[snowmoon]

Chipset's are a huge thing for me. Dispite how much I like AMD chips themselvs in terms of dollar/power I refuse to fight with my MB to get it stable. I know things have gotten better over the years, but with VIA it seems like there is a new incompatibility every week. Add to that the heavly slanted market twords intel ( optimizations are done for the intel chips long before amd chips ).

I love the chips, I hate the chipset.

[mrbass]

http://www.mrbass.org/dvd2svcd/ccespeeds/
those are my test results....remember in the newbie reference near the end it'll show you which various variables affect encoding speed. All ballpark estimates mind you.

[Hex]

Intel can boost core speeds and claim the crown of the top cpu in terms of Mhz. Intel know that the Mhz rating
is still the way most people (incorrectly) assess a processors speed and
are happy to reduce performance to increase Mhz.

That is a poor argument. Intel uses a longer pipeline, therefore each stage is shorter and fewer transistors are needed. Higher clock speeds are also obtained because of this. The whole AMD/Intel argument is like complaining about a Long Jumper who clearly jumps further, but took more steps. Who cares! The end result is a further jump/higher performance. High end Intel CPU's clearly outperform anything AMD can deliver. When hyperthreading is used on the new 3.0Ghz, look out!

[offline]

That is a poor argument.

Whatever the bias, at least I addressed the issue whereas you
fluff on about the pipeline and end up saying "who cares?"
That's a non-argument in my book.

Intel uses a longer pipeline, therefore each stage is shorter and fewer transistors are needed. Higher clock speeds are also obtained because of this.

?? of course.. when someone says "drop core size" they are
saying exactly that but in a simple, less convoluted fashion. Read
what I said carefully.

High end Intel CPU's clearly outperform anything AMD can deliver.

point out exactly where I said that p4 was inferior to AMD please?
This is not an intel vs amd thread.

[Kevin Henry]

I made the mistake of buying an AMD processor a year or so ago. It was reasonably fast (was the fatest chip you could buy at the time) but my GOD did I have problems with it. Eventually, it fried itself.

I bought a P4 2.26Ghz 533MHz FSB and SiS645DX mobo recently - a vast improvement in both stability & speed. Also, in most things it benchmarks faster than the Athlons that were avaliable at the time. And it was the same price.

Intel is still the best choice for CPU. Intel & MS got thier noses up each others asses so that makes sense for a start. The 'Lord of the Rings' was made on Dual Xeon CPUs (the guy who built the machines is my best friend). HP uses Xeons for it's servers. The list goes on.

It's only home users who really go for AMD cpus, and thats cos they were cheap. Gaming is a little faster on AMDs SOMETIMES but for those of us who do other stuff - like encode SVCDs - P4 is still a better chip.

Tempratures the chips run at, compatibility with memory types and the STONKIN' FSB that the 533MHz P4s have make alot of difference - my 2.26GHz 533MHz FSB out-performs a 2.4GHz 400MHz FSB. Also, consider the future. My CPU will work with 400MHz DDR, and will probably be competitive for longer than the AMDs will be, by upgrading mobo & memory.

Now, with the new P4's forcing down prices, why waste your money on Another Malfunctioning Device??? Many years ago, I bought an AMD K6-2 350 and took it back to my supplier the next day. I wish I could have done the same with the Asslon. And DONT EVEN mention that Duron P.O.S that I had for a month while waiting for the P4 prices to drop ...

There's my two cents worth. Experience is the teacher for those of us who dont believe the liars on the overclocking sites or Tom's benchmarks.