Duel P3 Vs. AMD @ 1.33

[Guest]

I am in the mood (and my wife finally said ok) to buy a new Mobo and CPU. I really want a powerhouse to crunch video and other very CPU intensive tasks. The applications I use will take advantage of multiple CPUs.

I guess my question is to anybody who has experience is,

1) Are AMD's REALLY just as (par for par) stable as Intel.

2) Is the price difference (~200) worth the duel P3-1ghz over the AMD?

2a) How much faster would the P3 be?

Thanks for any insight you have!

 

[Guest]

<<<1) Are AMD's REALLY just as (par for par) stable as Intel.>>>
in my opinion no but i'm sure you hear otherwise

<<<2) Is the price difference (~200) worth the duel P3-1ghz over the AMD?>>>
maybe, it depends on if your appz will benefit greatly from 2 cpus and if 100% stability/compatability is really important.

<<<2a) How much faster would the P3 be?>>>
faster sometimes slower sometimes, again it depends on the app.

if i were you i would wait until northwood comes out and either get 1 or dual p4's. the p3 system you would buy wouldn't have a upgrade path.

 

[Guest]

>1) Are AMD's REALLY just as (par for par) stable as Intel.

I've run mine for days at a time with number-crunching apps maxing out the CPU with no problems. I haven't had it long enough (about 2 months) and I've been mucking around to much with it to make any real stability comparison though. If your uptime isn't measured in months, it's unstable

>2) Is the price difference (~200) worth the duel P3-1ghz
>over the AMD?
>
>2a) How much faster would the P3 be?

I can only suggest that you try to find benchmarks for each processor. Also understand that, even for apps that can use SMP, you are not going to see 100% scaling.

That said, here is a <i>very</i> crude comparison...

Here is a <A HREF="http://www.linuxhardware.org/features/01/05/03/167228.shtml" target="_new">link</A> to a comparison of a 1GHz PIII and a 1GHz TBird on <A HREF="http://www.povray.org/" target="_new">POV-Ray</A>.

The scores are:
PIII: 21
TBird: 28

Now come some HUGE assumptions/simplifications:

Assume the PIII scales perfectly:
Dual PIII: 42

Assume the TBird scales linearly with clockspeed:
1.33 GHz TBird: 37.2

This shows about an 11% advantage to the dual PIII.
But as I said, there is some huge speculation involved in both numbers.

My guess is that the perfect SMP scaling assumption on the dual PIII is a worse assumption then the perfect clockspeed scaling assumption on the TBird. Would it amount to the 11% difference? I have no idea.

Some other notes:
The benchmark was performed with CAS 2 SDRAM. Getting a DDR based TBird board might help a little.

If an upgrade path is important to you, the PIII has none. The AMD solution does.


In theory, there is no difference between theory and practice.
In practice, there is.

 

[Guest]

"1) Are AMD's REALLY just as (par for par) stable as Intel."

The cpu's: yes. no doubt whatsoever. The system ? Yes, but only if you pay attention to proper cooling,and a decent PSU. AMD systems tend to be more pickey about these than intel systems. But, to give you an idea.. Anandtech runs its webservers on AMD systems.. that should tell you something about their stability (and performance).

2) Is the price difference (~200) worth the duel P3-1ghz over the AMD?

Hard to tell.. for properly multithreaded apps, the dual P3 is likely faster than the amd setup.. by how much ? depends on the apps. it might be helpfull if you named the applications. And even then it depends how much $200 is worth to you.

Personally, I wouldnt go for a dual P3 setup.. its a dead end street. If you can wait a bit longer, a dual Palomino would be *exactly* what you are looking for.. If you cant wait and have money to burn, a P4 1.5 or 1.7 is probably also a better choice than a dual P3.. Especially for video processing where SSE2 might make a real difference.

Dual P3 dont really scale well.. Their architecture is far from ideal for multiprocessing. Most apps only gain about 30%-50% from the second cpu. The AMD 770(i think), anyway the SMP solution, looks very promising. Some beta tests I've seen report gains of close to, and even *beyond* 100%. (Dont ask me how a dual system could be more than twice as fast as a single cpu with the same clock.. I just read it somewhere. maybe the chipset is faster by itself, even in single cpu mode).. but thats all speculation. We'll have to wait and see).


<P ID="edit"><FONT SIZE=-1><EM>Edited by bbaeyens on 05/08/01 04:41 PM.</EM></FONT></P>

 

[Guest]

Thanks for the info, I was pretty excited about the DUEL AMD solution until I found out that the AMD solution was going to be server level only (found at 2cpu.com) and would cost upwards of $900 just for the mobo. Oh well... I really like the IDEA of 2 cpu's but it may not be practicle anymore (Duel p4's will only be workable for XEON procs).

Thanks again, I'm leaning towards the single AMD.

Do you have a timeframe on when the next

 

[Guest]

Firstly Gkfisher be carefull who you listen to in here. If someone here recommends something make sure they actually use it. You wouldn't believe the crap I've heard in here that misleads alot of people. Most people simply hide behind a bias...be it Intel or Amd and attempt to mislead by posting irrelevant or more biased links. Basicaly the Dual Intel configuration is the only way to go for video. Especially if you are using Premiere or After Affects...(notice how nobody talks about software)? Both these applications are multi-threaded in i-code. That is the operating system can analyse 1 thread without having to wait for an answer from its sequencial thread.

1) Are AMD's REALLY just as (par for par) stable as Intel.

You can answer this question for your self if you take some time to look at the posts. The Amd problems out number the Intel issues by about 20 to 1. Regardless of wether its the cpu or chipset doesn't matter to the consumer...they just want a reliable machine. Also company like Pinnacle, Avid and Media 100 write alot of their code for the Intel instruction sets (primarily) so check the details of your capture card to determine any possible compatability issues or performance issues.

The cost difference is entirely up to you to consider...but next time you render a movie or if you have a real time affects card....boot up your cpu graphs (task manager in Windows) and watch them flutter between 95-100% while rendering (k...you got to test this on somebody else's dually). Duals kick but. I have two duallies at home (2xp2 450mhz) and (2xp3 733 mhz). I use premiere mainly for basic cutting and transitions and after affects for titling and the available plug-ins. I am a mechanical engineer but I do all our product demo's and some home video's too.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Guest]

Most of these people have never edited video. So your best bet is to post in the video editing forum where you'll get replies from power video users. If you seriously consider someone's post make sure you ask them some intense editing questions, otherwise ignore them. Or better still ring up some cable or free to air networks..."get some real answers". Good luck.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Guest]

I am in the same boat, and I am waiting untill the dual Athalons are available. The dual P3 is no slouch but they are only capable of using 160% of the 200% of the processors availability. From what I understand it is a limitation of the buss. From what has been reported, yet to be seen of course, the AMD buss will utilize all 200%. I do alot 3D rendering, and the 1333 was just posted on the Highend testcenter as the same as 2 P3 800's and only 2 seconds slower to render the same scene as dual p3 933's. If the Dual AMD does scale to 200% there will be a stampeed from 3D and video pros to the dual Athalon systems.
It seems that most of the incompatability that has been reported with the AMD processors is due to the fact that larege numbers of AMD processors are in self made machines. For cost cutting overclocking or whatever reason. Most manufacturers have ironed out the incompatabilities between components, drivers and whatnot.

Anim88tor

 

[Guest]

Yep I can see it now. You'll be working on the final render and your hyped for it then the blue screen comes oh no it cant multithread. Then 7 years of work gone and you kill your self. Tough luck.

SPUDMUFFIN

<font color=red>Being Evil Is Good. Cause I Can Be A Prick And Get Away With It.</font color=red> :lol:

 

[Guest]

Thanks for all the advice... I am leaning towards going with an AMD & IWill KK266-R. I would love a duel AMD setup, but don't want to pay out the $800+ I'll need just for the dual AMD board.

If anybody else has any thoughts, please post. Thanks!

 

[Guest]

>I am leaning towards going with an AMD & IWill KK266-R.

I have a KK266
Make sure you flash the bios up to the latest revision.
I had some PCI bus throughput problems until I did that.

In theory, there is no difference between theory and practice.
In practice, there is.

 

[Guest]

"The dual P3 is no slouch but they are only capable of using 160% of the 200% of the processors availability"

Well you can't complain about 200% and 160% isn't bad either. I don't know how your machine is configured but when I render I never get below 95% utilisation on both cpu's. No real animator sits in front of their machine while it renders, most real profesionals want viewport speed. If they continually render stills to sort out materials they turn off all their other objects. Therefore once again analysing geometry and motion is the name of the game, another thing duallies excel at, especially when the polygon count exceeds 6 figures.

If Amd does get there SMP situation sorted (when you think about spreading threads and sharing violations that may occur among common resources, smp is a difficult think to make work) it will still have to prove itself reliable, and as most people know the industry usually waits 6 months or a year before it upgrades to anything new. Therefore there is much to be proven before you can consider the AMD platform a real solution for CAD, 3D, video or server.

"It seems that most of the incompatability that has been reported with the AMD processors is due to the fact that larege numbers of AMD processors are in self made machines. For cost cutting overclocking or whatever reason"

Alot of people who build Amd's built plenty of Intels before without anywhere near the amount of problems they have now with Amd's. Yes of course you can build a working Amd setup, but Jesus couldn't they make it smoother sailing?

Anyhow gfisher, your best bet is to still ring up the cable and free to air networks and talk to some of their hardware engineers. Alot of people make posts to nurture there own false sense of self worth, so before you waste your money talk to real people (those actually employed to edit video) and get real answers (something that works fast and reliably, not riddled with problems) and make your own judgement. Good Luck.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Guest]

". Yes of course you can build a working Amd setup, but Jesus couldn't they make it smoother sailing? "

I wouldnt know how, really..I have setup about 4 AMD machines recently (all with KT133A boards of MSI), and I have not come across any issue to speak off.. okay,.. maybe one, with an old turtle beach montego audiocard under win2k, but that was easy to solve. Have you ever built a duron/atlhon machine ? IF not, how would you know ?


"Therefore there is much to be proven before you can consider the AMD platform a real solution for CAD, 3D, video or server."

True enough. although one has to admit that AMD has yet to produce a troublesome chipset. I'd be very warry to buy a VIA dual machine, wether it be with intel or amd cpu's however.

 

[Guest]

>Well you can't complain about 200% and 160% isn't bad either.

Huh?
He's saying that you <i>won't</i> get 200%

And 160% is pretty bad. Take a look at the SPEC Rate benchmarks to see how other SMP architectures scale.

The only dual PIII is a <A HREF="http://www.spec.org/osg/cpu2000/results/res2000q4/cpu2000-20001204-00423.html" target="_new">Dell 420 Precision Workstation</A> with RDRAM. There is an <A HREF="http://www.spec.org/osg/cpu2000/results/res2000q4/cpu2000-20001121-00355.html" target="_new">identical machine</A> (with only 1 cpu of course) in the single CPU results.

For comparison, here are two Alphas:
<A HREF="http://www.spec.org/osg/cpu2000/results/res2000q4/cpu2000-20001023-00275.html" target="_new">AlphaServer ES40 Model 6/833, Single CPU</A>
<A HREF="http://www.spec.org/osg/cpu2000/results/res2001q1/cpu2000-20010309-00516.html" target="_new">AlphaServer ES40 Model 6/833, Dual CPU</A>

These results include run times for each of the individual benchmarks that make up <A HREF="http://www.spec.org/osg/cpu2000/results/" target="_new">CFP2000</A>. On the dual CPU results, two sets of runs are started simultaneously. If a architecture scales perfectly, those runtimes will be identical to the single CPU scores where only a single run is started.

Here are results showing how much longer the two runs on a dual box take wrt a single run on a single box:

Benchmark PIII Alpha
168.wupwise 12.47% 0.38%
171.swim 89.51% 48.13%
172.mgrid 32.47% 19.00%
173.applu 36.98% 20.46%
177.mesa 4.60% -2.02%
178.galgel 45.42% -25.53%
179.art 78.38% 1.50%
183.equake 28.99% 0.71%
187.facerec 16.01% 8.25%
188.ammp 13.99% -1.55%
189.lucas 31.20% 17.52%
191.fma3d 16.86% 11.59%
200.sixtrack 0.48% -4.32%
301.apsi 16.87% 2.89%

Still think the PIII SMP scales well?
I am a little curiuos how the dual alpha actually was faster then the single on some marks. Any ideas?

This also shows how much SMP scaling depends on the app.
If you app has very low bandwidth utilization and largely (or completely) independent threads, you might get close to that 200%, similar to the sixtrack benchmark above.

>I don't know how your machine is configured but when I
>render I never get below 95% utilisation on both cpu's.

That doesn't neccesarily mean that you're getting good scaling. Time one of your "real animator" runs, then disable one of the CPU's. You can do that in Linux just by booting to a non-SMP kernel, not sure what you can do under ms os. Now time the same run again with just the single CPU and do the math.


In theory, there is no difference between theory and practice.
In practice, there is.

 

[ajmcgarry]

>> 1) Are AMD's REALLY just as (par for par) stable as Intel.

I don't really think there are any stability problems with the Athlon or T-Bird. The K6 though I would say has problems but your not going to go there anyways. The source of any problems with Athlons lie in the chipsets from Via. They tend to sacrifice extra stability for speed.

>>> 2) Is the price difference (~200) worth the duel P3-1ghz over the AMD?

Intel's greatest downfall. They only ever reduce prices when they are stopping support and something new is one the way.


2a) How much faster would the P3 be?

I think the question should probably be: How much faster would the Athlon be. But then again get yourself a P3 that has a hefty cache and you'll notice the difference. In my eyes the P3 Xeon 1MB cache is the best value for money.
There are reports of Athlons with 512KB of cache. I've seen adds on the Cnet auctions site, but I wonder how hard these are to come buy, as u don't here much about them in this forum.



I think it's down to a throw-up between two systems if you are really serious.
A dual P3 Xeon 1MB cache (if you want one right now), or
A dual Atlon ~1GHz with the AMD 76xMP series chipset, when it becomes available. (Which should be soon given Tom's report today)

I would also consider an ATI graphics card as the way to go for Amateur Video editing. You get some nice tools with the cards and they have the best video quality of the mainstream cards. Look at the Radeon All-In-Wonder 64MB. Unless of course u are going for something a bit more professional.

<font color=red>Why don't you ever see the headline "Psychic Wins Lottery"?</font color=red>

 

[Kelledin]

I don't know how your machine is configured but when I render I never get below 95% utilisation on both cpu's.

The other 5% of the time only accounts for the time when a CPU is waiting only for a device to raise an IRQ. If a CPU is waiting on the memory bus (or something else that doesn't raise an IRQ), the CPU still gets marked as busy, even though it's not getting anything done.

Work-done-per-unit-time (a.k.a. Benchmarks) is the only real measure of how well your CPUs are being utilized.

Kelledin

bash-2.04$ kill -9 1
init: Just what do you think you're doing, Dave?

 

[Guest]

hehehehe.....aaah you got to laugh, some poor guy wants to come here and ask about which computer he should buy. He is interested in video and tells us the cost difference to him is about $200. And look at the responses....we've got one slut running around the net gathering benchmarks and another who wants to blow money on a couple of xeons and then go buy an Ati card. Bwahahahahahahahaha........hehehehehehehehehehehe......ahahhahaha.......non of these bitches have ever cut video, wouldn't now the difference between cinepak or sorenson, a cross disolve or a page peel, ntsc or pal. Ahahahahahaha.....hehehehehehehe.....you can always spot the people who never use the software. Why does everyone think its all about render times? Do you really believe that a quicker render is more important to this guy than the ability to see his affects and transitions in real time? If he made a film that that lasted an hour (about 108000 frames) it would probably take him 20 times as long to decide on how he wants to cut it than the time it takes to render it. The only time you will ever sit motionless and wait for anything to render is if you can afford a render box. And if he has $20-$40k to spend he wouldn't be asking anyone questions in here. Dual xeons mated to an Ati card? God freakin help us all.
Let me tell you mules something. This guys hardrive and raid setup and his video card are alot more important to video than than his cpu dilema.
What I'd like to know is why companies keep producing multi-threaded software when they know that most of their users will be single Amd/Pentium or dual Pentium users? If the pentiums don't scale up so well why would they bother? Maybe you should mail your buullshiit to microsoft,kinetix, alias wavefront or Adobe? Show them your post and tell them your alot smarter than all their engineers and all their software developers, then try and convince the end users their stopwatches are all wrong and their cpu graphs are inverted. When they pick themselves up off the floor and give their faces a chance to stop acheing after laughing so hard they might scan in their butts and email them to you with a big "kiss this" sign on it as well.
Have you ever edited video ERGEORGE you silly bitch? Then fuuck off before the guy wastes his money just to find out he still can't get the job done. The funniest thing is I don't think gfisher even gives a crap, he hasn't bothered to reply to anything.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Ncogneto]

I've been following this post and you seem to make sense. I will not try to pretend I know anything about this at all for I have never used this type of software before. My interest lies in the fact my younger brother is currently taking classes at college to do this exact type of thing. So I am trying to gather as much information as possible to build him the best machine possible for this type of task for the least amount of money. What you say about raid setups and video cards makes a great deal of sense, as that is what I have already gathered up to this point. It appears that while Engeorge is comparing Rendering times you are saying that the viewport window is the app that one should pay more attention to. This makes sense as well, but what if both are of importance? Please forgive my ignorance in this matter, but one has to start somewhere. What would you suggest to a college student on a budget as the best cost effective solution for doing this type of work?

Finally, and at the risk of getting laughed off this thread, you mention render farms. My question is there any software out there that would allow you to render using distrubitive computing? Is this not feasable? for instance, say I have 6 computers all greater than 1 gig ( all single cpu macnines ) all tied together on a 10/100 lan. would there be anyway to utilize them all to do a rendering project? Or is this just not possible with the software that is available?

A little bit of knowledge is a dangerous thing!

 

[Guest]

Well hello Ncogneto. Thank god someone is actually interested in the real issue!! When gfisher talks about video I am assuming he means video shot with a camera, not animation. Its important to know what your brothers specialty is. If its animation then firewire in/out might suffice on a budget, just for overlays or blue screen for his animations. If he is cutting alot of footage and applying special affects he will want a capture card that will give him real time visuals. As most of the footage is already rendered by the camera (be it mini Dv, Hi-8 etc) you don't need to worry about render boxes or render farms (particularly if he's still a student). There aren't even too many professionals out there who would cut a feature length film everyday, most have the script writers or director in their ear so sometimes it can take days to cut it and less than 2 hours to render it. However animation is VERY different. The greatest thing about transitions and affects in real time is the fact that you dont have to render the movie to see the affects, thats why these cards sell and its a very very important feature if you want to create gripping video. If its animation then Cpu, graphics card and ram will be the major issues. If its footage then capture card and drive space/speed will be your biggest issues. We use to create 30-40 minute training video's for work on a Pentium 2 500Mhz. Our biggest concern was having analogue and digital inputs and outputs....hehehe. In Adobe Premier you have a slider in the timeline that will allow you to render as much or as little footage as you like even if you don't have a real time card. Its a bit of a pain with affects and transitions but it still served our purposes. Let me know more about your brothers situation and we'll come up with something. If its footage he's rendering do some research into the capture cards as the rest of your budget might revolve around this. Good luck.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Guest]

He asked whether he should go with a 1.33 GHz AMD or a dual PIII. My posts address that. And no, I've never "cut video". Look at my first post. I told him to find benchmarks for the software he wants to run. Anything after that is speculation & discussion.

And why do they still produce multithreaded software? Because, until recently, a dual PIII was definitly, no questions asked, the way to go for this stuff. It may not be now. And almost certainly won't be with the next clock step on the AMD side. Or the P4 even. And when dual Athlons & P4's come out multithreading will be important again.

And there is more to SMP then 2 CPUs. If you're really serious, then you go out and get a 4-way or better box. Even with the PIII's crappy scaling, a 4-way box will do very well, although you'd have to question the utility compared to the cost of a small render farm.

So, to question why software houses produce multithreaded apps is pretty ignorant in my view.

And if you think this guy should be more concerned with his harddrive & videocard, fine. I addressed the question asked. As you say, I've never "cut video", so I can't really speak to that.

And finally, can't you say anything without profanity and name calling. Maybe you should go back to playing with you imaginary friends.


In theory, there is no difference between theory and practice.
In practice, there is.

 

[Guest]

Yea, I guess I am a silly slut for gathering & posting benchmarks. After alll, nobody here is interested in objective facts. We should all just listen to tonestar. He is the man.

Oh wait, that's right. tonestar is a demonstrated liar & fraud. Maybe I'll be looking at those benchmarks anyway...


In theory, there is no difference between theory and practice.
In practice, there is.

 

[Guest]

Hello ergeorge!! Wakey wakey, hand off snakey!!
On page 1 I said:
"Basicaly the Dual Intel configuration is the only way to go for video. Especially if you are using Premiere or After Affects.."

On page 2 you said:
"And why do they still produce multithreaded software? Because, until recently, a dual PIII was definitly, no questions asked, the way to go for this stuff. It may not be now. And almost certainly won't be with the next clock step on the AMD side. Or the P4 even. And when dual Athlons & P4's come out multithreading will be important again."

Gkfisher price range extends to a single Amd or dual P3 setup. So I guess you agree with me then. So what on earth are you argueing about now? Nice to see you admit you don't know anything about video. You have grown as person today, even if it is only a little bit.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!

 

[Guest]

You can't read your own post, can you?

I repeat, <i>until recently</i> there was no question that the dual PIII was the way to go. And again I quote <i>"It may not be now."</i>

Now, I see it as pretty much neck & neck. It comes down to specific applications & benchmarks. But the PIII is at the end of its line, He'll have a choice to upgrade to an even faster AMD chip in the future.

You need to work on your reading comprehension. Or are you deliberately mis-interpreting my post?

In theory, there is no difference between theory and practice.
In practice, there is.

 

[Guest]

ergeorge, thanks for saving me some typing.

On a side note.
It seems that there must be a problem with the theroy of natural selection. Shouldn't tonestar and spunk monger have stepped in front of busses by now?
Maybe they are in "assisted living" situations, you know where there mittens are pinned to there coats.

Anim88tor

 

[Ncogneto]

No he is not into shooting movies with a vcr he is learning 3d studio max and a few others ( they escape know but he will be over sometime today). So far he has a Gforce2 pro ( he does gaming as well) and two maxtor 7200's ( 2 meg cache) in a raid O array. Firewire may still be a bit faster but this is not to shabby. The raid array is using a modded promise 66 controller ( add one 100 ohm resitor and whalla) which performs really well. Doing a little research into doing somethig simular with his video card but still not sure on that. He currently has a one cpu setup ( AMD 1.2 gig) but duals maybe in his future. Dual p4's will certainly be out of his price range any time in the near future ( even with me doing it at cost) Dual p3's is a possibility but stuck at 1 gig for now. Dual athlons maybe a possibility also but from what I here at first the motherboards will be very high and geared more towards the server market. But, others are supposedly on the way. A dual athlon my give him a little more room to grow, not getting stuck at 1 gig, but agian he will be navigating uncharted waters here. So i take it there is no such program that would allow rendering to be distrubited across a network? this seems, in theory, to be something that could be done ( I think the whloe idea of distrubited computeing should be exploited much more) but does it exist?

A little bit of knowledge is a dangerous thing!