Log in

View Full Version : Anyone using 3.05ghz P4 for CCE Encoding? New P4 or Dual Xeon?


Roveer
26th November 2002, 13:56
I'm at a crossroads as to whether to build a CCE encoder using a single P4 3.06ghz CPU or to build a Dual 2.4ghz Xeon. I going to use RDRAM (RAMBUS) or Dual Channel DDR, on either system. Memory is important, so only the fastest will do.

I do all my encoding with CCE 2.5 with AVISynth and really dont want to switch (except to other CCE versions).

I've read reviews, and CCE wins just about every one, plus I'm very happy with the output. I've decided that I'm going to stick with 3 pass, which today takes 6-8 hours on a 2 hour movie on my 1.6ghz P4. I'd like to improve on that number.

If anyone is using either a fast dual, or a new 3.06 ghz P4, please share your experience. Your comments will help me and others who find themselves in the same situation.

Thanks,

Roveer

colebert
26th November 2002, 16:39
If you use AVISynth, duals won't matter cos its not coded for SMP.


-Cole

LB
27th November 2002, 22:34
It's common sense to encode CCE using VBR intead of CBR. And it's also intelligent - yet not everyone does it this way - to use a huffy avi file as your source for CCE. ESP. if you have dual cpus. Quick reasons why:

1. Huffy avi is essentially a lossless compression codec. A two hour movie will give you a 30gb huffy avi file. You create this in VDUB. Benefit? You will now be feeding an uncompressed avi into cce instead of a filtered up, ivtced, bicubic resized avs file so the rendering time can in most cases be 500% FASTER!

2. Avisynth + VDUB = SMP AWARE. What does this mean? It means that while Avisynth is not SMP compatible, and neither is Vdub, they both when used together are AWARE of dual cpus and thus will each utilize their own processor. In english? Avisynth will render on one cpu while vdub renders on the other. This will drastically speed up the conversion to a huffy avi.

So basically, even if you don't have dual cpus, you still should use a huffy avi. You will see ~ 300% increase in speed. Couple that with SMP + HUFFY w/AVS and you have a huge improvement in speed.

Finally, 3ghz xeons probably will not be released until FEB at the earliest. There are no current boards out for them yet. ;( On another s*tty note, I just checked pricewatch and the 3ghz p4s are 799$ ;p Sucks eh? I'm debating on whether or not just to L5 mod some 2.8 AMD's and run them. Oh well.

LB

Roveer
27th November 2002, 23:18
Are there any guides as to how to create a huffy avi? I've been reading about this, and think I want to give it a try. Your post describes this at a somewhat high level, but I'd need more information. If there are no guides, can you step me through the process. I'd like to see just how fast this is.

Roveer

colebert
27th November 2002, 23:44
I have 2x9G 10K RPM drives and 1x36G 10K RPM drive. Huffy might not work too well for those who go for speed and low latency in their drives (i.e. U120/U160 SCSI.)

Sounds intriguing though.

-Cole

TyMac
28th November 2002, 11:46
Go after a dual xeon rig! Get a encoder that supports multiprocessing an run with it. Xeon's have HT so you could have an app think it's using 4 processors... TMPG supports dual processors and rempeg can run parallel threads... maybe more than 2... I'm not sure.

TCCK
29th November 2002, 19:59
P$- 3.06 with hyperthreading is the way to go as it gives you the cheap "daul CPU" without the cost of the expensive MoBo or second pricy CPU.

I have a 3.06 with an ASUS P4G8X motherboard and all my high end customers I am encouraging to go this way as it has the room to expand with the future technology.

Sure go Athlons if you are on a budget but you WILL NOT get the performance as compared to the Intel P4 -3.06.

I test these daily, I KNOW!

There are a lot of AMD fans that will now flame me for my statemnets here but I do not care as I have done the tests and the proof has laid down in favour of the Intel P4 - 3.06.

Why do you think that NO MAJOR server manufacturer is using AMD processors for their servers when they are supposedly better!? Thier not and that is why no servere manufacturer uses AMD!!!

The dual 2.8 XEON would beat the P4-3.06 but at a massive amount more $. (I also have a dual 2.4 XEON system and I prefer my new P4- 3.06 for most stuff but I still encode on my dual 2.4 as it is not being used for anything but that!)

Roveer
29th November 2002, 20:25
Right now, I'm deciding on a P4 3.06 with the Asus Mobo, or a Dual 2.6ghz Xeon (533fsb), with a iwill dp533 mobo. Still can't decide, but time is not an issue. I'm not an AMD guy, nothing against them, but I'm going to stick with Intel for now.

Roveer

TCCK
29th November 2002, 21:58
Really that one comes down to your budget. The dual system is going to cost more and reality get maybe just a bit more performance.

It all depends on how much you want to spend and do you want to be able to upgrade it.

The ASUS board with the 3.06 will upgrade for quite a while, compared to the dual CPU board will need a new board sooner (support for next gen. HT XEON's) and when you upgrade it is not just 1 cpu but 2 you have to buy and the XEON are a fair amount more $.

orionware
30th November 2002, 16:30
The intel does have the FPU advantage, however in real life examples I haven't seen much of a difference. I would say if you are going to spend every day doing fpu intensive processing, it might be worth the extra cost of the Intel. However, in most every day real life applications, the AMD wins hands down.

Why do you think that NO MAJOR server manufacturer is using AMD processors for their servers when they are supposedly better!? Thier not and that is why no servere manufacturer uses AMD!!!


Because Intel is a marketing machine who already has a well secured foothold in that market. IBM, Sun and Dell have all dabbled with AMD chips but have been lured to halt development or changed plans because of Intel responding with massive price cuts often resulting in selling chips at a loss to maintain the position. Intel is the m$ of the hardware world when it comes to bullying manufacturers and VARs. We currently have three dozen xp/mp systems running everything from DNS, Virtual Host Web Servers, Mail servers and Oracle/MS SQL boxes. Haven't had one CPU flake or fail.


Now... I have an XP2700 with the new Dual DDR Asus Mobo (A7NX8) system and a 3.06 that I built for a friend. I played with it for 5 days before he picked it up and I found the AMD system is about 5-7% faster when using CCE. I would expect that the Dual DDR has something to do with it as I picked up performance just going to this mobo. I think AMD and Asus make a great match. Solid and priced right. And from what I have read on the new Hammer (Opteron) chips you might want to wait. A 64 Bit chip that will be backwards compatable with the 32 bit stuff unlike the Itanium.


The difference between the 3.06 Intel and the 2700XP was about 325 bucks when they were purchased.

auenf
30th November 2002, 16:35
Originally posted by orionware
The intel does have the FPU advantage, however in real life examples I haven't seen much of a difference. I would say if you are going to spend every day doing fpu intensive processing, it might be worth the extra cost of the Intel. However, in most every day real life applications, the AMD wins hands down.

actually the P4 FPU is weak, and intel has been trying to push SSE2 as a replacement.

Enf...

orionware
30th November 2002, 16:51
Yea. Unless something is SSE2 optimized you do see a hugh dropoff. SiSoft's benchmarks have shown the Intel SSE2 FPU faster but overall processing power weaker than AMD.

http://zippy.sgsnet.com/~mchytrac/ebay/2700xp_op.jpg

The reference cpu is an XP2700 / 1 Gig Dual DDR / Asus A7NX8
(Drystones 9189 Mips / Whetstone FPU 3723 MFLOPS)

The CPU's being compared are

Intel P4 2.8Ghz 512L2 (8196 MIPS FPU/2408 SSE2/5250 MFLOPS)
Intel P4 2.66Ghz 512Ls (7113 MIPS FPU/1496 SSE2/3492 MFLOPS)
Athlon XP 2200+ 256L2 (6775 MIPS FPU/2752 MFLOPS)
Intel P4 2.4Ghz 512L2 (6403 MIPS FPU/1346 SSE2/3144 MFLOPS)

LB
1st December 2002, 03:44
I beg to differ bro. Comparing a single cpu intel to dual AMDs is like comparing apples and oranges.

Now, there have been reviews of the hyperthreaded 3.0s up against single AMDs at toms hardware.

http://www17.tomshardware.com/cpu/02q4/021114/p4_306ht-15.html

As you can see, a hyperthreading enabled 3ghz intel is only able to achieve 1FPS, yes, go read it for yourself, 1FPS more (264FPS v. 265FPS) encoding with Divx5.02.

Now, there are many people that dislike divx 5.02, as do I. But, with limited benchmarks available, we should look to an application that will give us an idea of what to expect for our purposes. So, does that test utilize hyperthreading? Of course. Does hyperthreading make a difference there? No. (Now, none of us know if the cpu was already being maxed out at 100%. If it was, then there is obviously no room for a benefit with hyperthreading, but if it wasnt....) And, compare that to the AMD which is only available at 2.8-SR and you see that the intel is only chuggin along 5% faster. Now we also have to consider the fact that dual xeon 3ghz are not even avail. yet. (This isn't a xeon being tested). So tack on another 5 month wait to that + a high price tag. Aight, let's look at some mpeg2 encoding. Scroll down in that link I gave you to the second test.

TEST2. Amazing eh? We see that, although its marginal, the 2.8-SR AMD *BEATS* the 3GHZ INTEL at mpeg2 encoding in that particular application. Woah. Let's all take a few steps back to think about that for a second.

Conclusions. So what can we draw from this? Well, it's obvious that for encoding purposes, it doesn't look like Hyperthreading is worth a hill of beans. This is just from these tests, but that's the impression i've recieved. Also, the entire conversation is about dual processors. No xeon's with HT are out at the moment but we have seen there is no marginal difference with it. So, let's compare dual processors. The higest speed dual intel out at the moment is 2.8s and the fastest AMD for duals is 2.6-SR (w/l-5 mod). Also, while it's not a SMP test, we can see how a 2.6-SR amd beats down the 2.8 intel at single cpu mpeg2 encoding, so it's fair to assume that roughly the same will occur in a dual situation for each chip. So, at the moment, AMD has the clear advantage in the encoding realm. Weighing in the cost factor is just another *huge* weight on the end of the scale for AMD. So, for the SAME price of a SINGLE 3ghz Intel chip (699$) you could get DUAL AMD 2.6's + a Motherboard and still have money left over to take your G/F to the movies! I'll say that again...

FOR THE SAME PRICE as a SINGLE 3GHZ Hyperthreaded Chip and NO motherboard, you could get DUAL 2.6-SR AMD's + SMP Motherboard and still have money to spare.

SO, long story short. If you are looking to encode, before you even consider an intel, look at dual AMDs since there is NO comparison.

Cost and performance weigh in favor of AMD.

Before I go though, I would like to make one thing clear. You are likely to see a 90-100% increase in performance for a SMP compatible encoding application by going to dual chips of whatever single chip you are looking at for those of you that are considering a single cpu setup versus duals. Although, a lot of dually boards can be troublesome so it would be wise if you do decide to take that path to check out 2cpu.com and their forums since you can get a good indication of which is the best dually board for you. My a7m266 was a true pain in the arse to get stable. It has real irq issues. But it's stable now and that's all that matters. So, make sure you know all about your board before ya pick one.

LB

orionware
3rd December 2002, 20:31
http://www.2cpu.com/Hardware/GA-7DPXDW+/

Interesting benchmarks concerning this subject..

I'd go with the Dual AMD system over the Xeons

bumblebe
7th December 2002, 08:21
Well I just upgraded my trusty Dual Celeron with an el-cheapo P 4.24b ($200) that runs cool at 2.8ghz+

Rempeg:
P4/2.8 - 10fps
C550 - 3fps
2x C550 - 3.5fps

Hehe...

TRILIGHT
8th December 2002, 06:45
Originally posted by TCCK
I have done the tests and the proof has laid down in favour of the Intel P4 - 3.06.

Can you please share the encoding speed you get in CCE? I would like to know. I plan on upgrading and was looking at this chip but I would like to know what tests you've done (specifically CCE) and give me a number. Thanks!

Oh, and as for people being hooked on AMD, yes they are fast and are cheap. A lot of times they look good on paper. However, in my experience, you will get a LOT more stability and compatibility out of an Intel/Asus processor/mainboard combination than you will out of an AMD setup. Do I want a fast processor? Sure! Do I want it at a good price? You bet! Do I want to have bullshit little issues with certain hardware combinations or unknown software issues? HELL NO!!!!

auenf
9th December 2002, 11:28
Originally posted by TRILIGHT
Oh, and as for people being hooked on AMD, yes they are fast and are cheap. A lot of times they look good on paper. However, in my experience, you will get a LOT more stability and compatibility out of an Intel/Asus processor/mainboard combination than you will out of an AMD setup. Do I want a fast processor? Sure! Do I want it at a good price? You bet! Do I want to have bullshit little issues with certain hardware combinations or unknown software issues? HELL NO!!!!

does TRILIGHT want a challenge? NO!

Enf...

bumblebe
9th December 2002, 19:30
Originally posted by TRILIGHT
Can you please share the encoding speed you get in CCE? I would like to know.

On my P4 2.4b overclokced to 2.9ghz, I get x1.9 using VFAPI. In Tmpeg I get about x0.7 with a DVD2AVI source

TRILIGHT
9th December 2002, 20:14
Thanks for the info, bumblebe!

bumblebe
9th December 2002, 20:21
sure, how does it compare to your present system.

I forgot to mention that I get about x0.4 in Rempeg at 720x480 but thought you wouldn't necessarily be interested in that info ;-) I must say that since Rempeg let me down with Titanic, I'm taking a dimmer view towards it!

TRILIGHT
9th December 2002, 20:41
I won't get into my thoughts on ReMpeg. Also, is the speed you are reporting in FILM format and going straight? Are you using any filters? Performing any IVTC during the encode?

Right now, I have a P4 1.4Ghz clocked to 1.7Ghz and I usually get 1.1x on average in CCE without running though any filters or processing.

bumblebe
9th December 2002, 20:45
hehe...I was willing to be lenient with Rempeg because it was convenient but now that it fails on some moveis, it's got little going for it.

In CCE I was feeding a FF dvd2avi project via VFAPI (sp?) and nothing else. In Tmpeg, straight dvd2avi but at 29.97 this time.

dannyv
12th December 2002, 20:47
Originally posted by TRILIGHT
Can you please share the encoding speed you get in CCE?

With cce I encode a 2:16 movie (startrek the motion picture) with 1 pass in 1:07. Did it just last night. My specs are below.