Any major differences between Pentium/AMD nowadays?

[Durk]

OTOH, intels dothan/centrino is really an amazing CPU. low power consumption, low heat, long battery life, and performance that is up there with the +3 GHZ/performance rating desktop CPUs. incredible. in addition to making it the obvious recommendation for laptops, i really hope intel makes the switch and gives up on the P4 and starts to put
dothans into desktops, cause i hear theres still a lot of, erm, "headroom" left.

Don't get too excited yet. A couple of days ago I read a clever comment from a Dutch guy. He has some good points, I think. So I can't resist sharing this with you all. It's a comparison between the Pentium M (Dothan) and Athlon 64. I will translate the interesting pieces:

Scalability.
--------------------------------------------

The Pentium M is produced with a 0.09 micron process while the Athlon 64 is still on 0.13 micron. Having the Athlon 3000+ (which is castrated and a low-end model) perform slightly better overall than a Pentium M 2 GHz (the high-end model which is almost at its top; not a lot headroom left for a speed upgrade with the same process), this tells us something about the future of the Pentium M in desktops. Intel will have a hard time to produce a Pentium M which will perform on par with the high-end Athlon 64 (FX) models (which are, by that time, also on 0.09 micron and will allow much higher clock speeds).

Power consumption.
--------------------------------------------

Everybody's raving about the low power consumption of the Pentium M. Well it's not consuming a lot, but comparing a Pentium M 1.8 (Dothan) with a 1.8 GHz AMD K8-based processor, doesn't make the K8-core put in shame:

Specs tell us that the Pentium M consumes 21W TDPtyp and the K8 35W TDPmax. Intel never spreads around TDPmax values, but luckily this site has done some measuring. It seems that the TDPmax for a 1.8 Dothan-based processor is 26.817W (in full speed mode). Now it's easier to compare.. as you see, there's a difference of 8.2W.

The actual difference of the systems carrying these processors is even smaller, because the on-die memory controller of the K8-based processors make K8-chipsets consume less power. Intel needs an off-die memory controller in the chipset for their Dothan-based processors. Too bad, because that's the most consuming part of a chipset. So the 8.2W difference is in real life even smaller.

Well and there's the 0.09 micron version of the K8 coming, which will consume less power and perform better than their predecessors.

64-bit support.
--------------------------------------------

While the lastest Xeon (Nocona) and Pentium 4 (Prescott) processors carry 64-bit support (Intel calls this EM64T), the Dothan does not. It will require a complete redesign to implement this in the Pentium M. This takes a LOT of time.

Concluding.
--------------------------------------------

The Pentium M is a very nice processor for sure, but it won't be too easy to bring this technology to the desktop.

and not to forget, the dual G5s are actually the first apple machines in

years that have comparable performance to the fastest pcs, sometimes even better (especially

when it comes to logic audio, which seems to be highly optimized here). pricey, yes, but

other than with the late G4s, now you actually are rewarded with great performance for that

price again.

Too bad I never saw good reviews comparing the G4/G5 with Intel and AMD processors. So I can't confirm your point. Do you've got some usefull links?

either way, with a few exceptions, its rather hard to find a truly bad CPU on

the market these days. its not like a fast P4 is a bad choice either. good times to buy a

daw computer

Some people still have problems deciding which kinda processor they will get...

[bebop]

egbert wrote :

your audio card (normally) has no role in mixing multiple tracks together - the audio engine of your sequencing package does that for you if you are mixing to stereo outs. If the audio quality from the card is not so great it is just as bad with one track as with 100.

Still depends if the sequencer has got an offline rendering engine...

But anyway offline or online rendering doesn't matter that much when you consider both are about summing the tracks : past a certain amount of tracks playing at the same time ANYWAY it'll sound muddy because you'll just have added too much sound in your project.

Remember long before being an art of sound, music is an art of time and thus an art of silence.

There was a topic recently asking when a song was "going south", and that's it precisely, too much is not equal to better at all. This is why I say in most cases you don't need 32 tracks to express a musical idea, and even 16 plain and all different tracks is yet probably too much too. Also when you play say a piano in a track, you hardly use more than 64 voices of polyphony, so for synths you don't need to be able to play 256 voices x 32 synth tracks, because musically appart from a great cacophony it won't lead you anywhere, etc. Of course for the mixing you can end up with 32+ tracks easily (f.e. 10 drums, 16 vocals, each instruments with it(s) send(s) separately etc), but you don't want to do this with vstis, with audio instead, and you don't add 32 different reverbs neither anyway, because with say 8 types of reverbs you can do it all, etc.

I think you get the argument, it doesn't matter really what processor you use, for in any case (say almost any case), above 2ghz any cpu is able to handle 12-16 vstis tracks with a low latency, faaaar enough to express your musical ideas. Still not convinced? Take your favorite LPs and listen to them carefully, make the count of the instruments and tracks, you'll see, 16 is quite a lot already (even if as mentioned at the studio they needed 48 tracks just for the vocals indeed etc, but 48 audio tracks has only little to do with the cpu...)

Now it doesn't help if it comes to decide if AMD's are better than Intel's, but as they're usually cheaper, as people say here and there AMD are better for audio applications, there's no reason to doubt, AMD is probably the best way to go yes.

But I'm happy I don't have to make a decision between AMD64 and Pentium 4 today (or this year at least), as the question of 64bits support isn't cleared yet, specially when Intel announces 64bits p4 to come...

[DevonB]

But anyway offline or online rendering doesn't matter that much when you consider both are about summing the tracks : past a certain amount of tracks playing at the same time ANYWAY it'll sound muddy because you'll just have added too much sound in your project.

Remember long before being an art of sound, music is an art of time and thus an art of silence.

There was a topic recently asking when a song was "going south", and that's it precisely, too much is not equal to better at all. This is why I say in most cases you don't need 32 tracks to express a musical idea, and even 16 plain and all different tracks is yet probably too much too. Also when you play say a piano in a track, you hardly use more than 64 voices of polyphony, so for synths you don't need to be able to play 256 voices x 32 synth tracks, because musically appart from a great cacophony it won't lead you anywhere, etc. Of course for the mixing you can end up with 32+ tracks easily (f.e. 10 drums, 16 vocals, each instruments with it(s) send(s) separately etc), but you don't want to do this with vstis, with audio instead, and you don't add 32 different reverbs neither anyway, because with say 8 types of reverbs you can do it all, etc.

Orchestral tracks are very regularily over 32 tracks of audio. The bad mix comes down to bad effects and bad technique, not track count.

Devon

[derek]

Don't get too excited yet. A couple of days ago I read a clever comment from
a Dutch guy. He has some good points, I think. So I can't resist sharing
this with you all. It's a comparison between the Pentium M (Dothan) and
Athlon 64. I will translate the interesting pieces:

sorry man but even though im a big believer in AMD and always wish them the
best for everything they do, that rant sounds so goddamn biased and
looking as hard as it can for weaknesses, that i would take it with a grain
of salt. or two. i mean, huh? the dothan is "almost on its top"? that goes
against more or less *everything* i read about it in various magazines, what with
the dothan just being out, being easily overclocked to 2,5 GHZ and having
more headroom than most CPUs have had in a while.

or about the power consuption...well, i will give that some thought when
i see a laptop that runs with an AMD 64 and lasts 5-6 hours with a regular
battery. until then, its theorizing far away from RL, and its wishful thinking.

its wishful thinking i would completely share, because the more AMD gets
from intels cake, the better for us customers. and i just like AMD, ive had
many desktops with AMDs in it and my next one will be AMD too. but its still just that, wishful thinking.

and that has nothing to do with it having 64bit support. as was mentioned in
this thread, for now (and some time being) thats more or less irrelevant for music. its about speed and how it performs in real life, today, not foggy speculations about what will and what might be some day.

allthough i have to say, in a way, the whole 64bit thing was a good return from AMD to intel. intel sorta "caught" AMD on the wrong foot when they managed to make the GHZ spec appear to be the only relevant number to the public.

now, AMD does the same with the 64bit hype, and intel, stuck with a P4 and dothan that dont do 64bit at this time, is caught on the wrong foot.

i like that :-)

Too bad I never saw good reviews comparing the G4/G5 with Intel and AMD processors. So I can't confirm your point. Do you've got some usefull links?

links, no. in german keyboards magazine there are very detailed and competent benchmarks on a regular basis. those benchmarks deal with the only thing thats important to us, and that is, how many tracks, how many plugins, what latencies with various soundcards can be run without crackles etc.

they measure this with a whole armada of standard tests they have come up with, plugins that depend on fast memory transfers, plugins that depend on raw CPU speed only, plugins known to be optimized for intel, plugins that are heavy on FPU, plus all tests are done with various latencies, and if that werent enough, under different hosts, so that you get a *real* picture of what to expect from a computer.

these are the best, most detailed and thoughful audio benchmarks i know.
they do this regularly with laptops of various kinds, custom audio pcs, standard pcs (for comparison), and of course macintoshs or all sorts too.

the dual 2 GHZ G5 was tested a while ago, and it was faster in many aspects than +3 GHZ pcs.
and i wouldnt suspect these people of being biased, because those were the same people that got themselves into the mother of all flamewars when they told the mac community that the G4 had lost the race already years ago.

:-)

im not much into macs anymore (my last mac was running macOS seven point something, that should give you an idea), but the numbers the G5s scored in these benchmarks were pretty impressive. especially under logic. note that they also tested some of the (now discontinued) single CPU G5s, which didnt have the performance to justify their price. it was impressing to see how good the dual models scaled, all the while not being that much more expensive. (relatively speaking of course)

[bebop]

DevonB... are you saying you compose orchestrations with 32 tracks + in your kitchen on a regular basis?

I mean that's really cool, and nowadays with modern samplers and soundbanks that's very well possible indeed, all you need is a small cluster of fast computers with fast hard drives, a few dozen of gb of fast ram, rather a large kitchen than a small, a solid musical formation, a desperating lack of social ability (that would lead you to work with a real orchestra) and a non less desperating joy in loneliness (same reason)...

But ok that's indeniably true and very relevant : track count has no importance at all, and in some cases what makes the difference between a kvrian track and a world wide hit is the 57th reverb they added here and not there...

ok ok...

Do you know people if there is a "world record in audio sequencer tracking" in the Guiness Book and what is that record currently?
Also what about a "fastest computer record"? I tell you I'm in the mood today I wanna beat this record now I wanna feel like the one guy with the 2 feet long...

[DevonB]

DevonB... are you saying you compose orchestrations with 32 tracks + in your kitchen on a regular basis?

What's to answer with your post? Obviously you already know it all.

BTW- Giga160 does 64 channels, and has for years. Doing orchestral work eats up slots very quickly. But I'm sure you already knew that though, right?

Devon

[bebop]

lol

And what about that one? http://www.amdboard.com/tyan_s4880_opteron_board.html
For the price of a comfortable car you'd get a nice machine to play with during long evenings in the winter...

When I think Bach composed some of the finest pieces of all times for a single violin with its 4 poor strings (barely 4 voices of polyphony!!!), and on the other side Mariah Carey some of the most boring songs of all times (the comptetition for the "most boring song's award" is still opened actually) with a glowing 128 tracks PT HD set up (I guess),
yes I keep saying the number of tracks has no importance at all, or only maybe to impress your mates, and even, ot sure it'd work.

I may also be completely wrong, after all, wasn't that Bach a real retard? I mean, eeek! Music with a solo violin, not very groovy isn't?

[DevonB]

When you quit acting like you're 12 years old Bebop, we'll continue. I have a feeling that won't come to fruition though.

Devon

[ObsoleteAcc99]

[DELETED]

[bebop]

I read reviews and magasines, like you do too I suppose.

And indeed that Tyan mainboard looks awesome, indeed some large sample sets like the Vienna Symphonic Orchestra are reviewed in magasines and they indeed test those on small domestic clusters, indeed I really mean what I say about the number of tracks, indeed even I think at some point technology is the enemy of arts, in the way it makes you think less [musical] techniques are required as long as you've got more technology. But that's suddenly completely off topic too, indeed.

Now about Mariah Carey I'm sorry, I shouldn't have done that, sorry sorry, I didn't realise how touchy a point she was for you

[Durk]

Wow, that message must have taken time!

sorry man but even though im a big believer in AMD and always wish them the
best for everything they do, that rant sounds so goddamn biased and
looking as hard as it can for weaknesses, that i would take it with a grain
of salt. or two. i mean, huh? the dothan is "almost on its top"? that goes
against more or less *everything* i read about it in various magazines, what with
the dothan just being out, being easily overclocked to 2,5 GHZ and having
more headroom than most CPUs have had in a while.

As you know, Pentium 4 Northwoods were easily overclocked as well. But those speeds never made it to the shop. 3.4 GHz was the max, but overclockers went a lot higher. Same goes for AMD Athlon XPs.

I don't expect the Dothan running higher than 2.2 GHz.

or about the power consuption...well, i will give that some thought when
i see a laptop that runs with an AMD 64 and lasts 5-6 hours with a regular
battery. until then, its theorizing far away from RL, and its wishful thinking.

The problem with testing laptops is all the different components. AMD-based laptops are more often sold as budget or mid-end solution than Intel-based ones. So, which one will have better batteries? Sure I can't prove it nor test is. But it's just an example how hard it is to compare laptops, because different screens matters, different graphics matters, etc.

But as you're saying, it would be nice to have some real life testresults.

Just wondering.. Why do I look biased when I'm saying something about numbers in specifications? It's at least more objective than saying: "My new Athlon 64 is running a lot longer than my Pentium M laptop".

and that has nothing to do with it having 64bit support. as was mentioned in
this thread, for now (and some time being) thats more or less irrelevant for music. its about speed and how it performs in real life, today, not foggy speculations about what will and what might be some day.

True. But that wasn't my point. It was all about what kind of challenges Intel has to overcome before lauching a Pentium M-based desktop processor. We might not yet care about 64-bit support, but Intel does! Certainly at the time taht the new desktop processor should be launched.

now, AMD does the same with the 64bit hype, and intel, stuck with a P4 and dothan that dont do 64bit at this time, is caught on the wrong foot.

i like that

Well.... a P4 with 64-bit support will be soon available in the shops!

links, no. in german keyboards magazine there are very detailed and competent benchmarks on a regular basis. those benchmarks deal with the only thing thats important to us, and that is, how many tracks, how many plugins, what latencies with various soundcards can be run without crackles etc.

they measure this with a whole armada of standard tests they have come up with, plugins that depend on fast memory transfers, plugins that depend on raw CPU speed only, plugins known to be optimized for intel, plugins that are heavy on FPU, plus all tests are done with various latencies, and if that werent enough, under different hosts, so that you get a *real* picture of what to expect from a computer.

these are the best, most detailed and thoughful audio benchmarks i know.
they do this regularly with laptops of various kinds, custom audio pcs, standard pcs (for comparison), and of course macintoshs or all sorts too.

the dual 2 GHZ G5 was tested a while ago, and it was faster in many aspects than +3 GHZ pcs.
and i wouldnt suspect these people of being biased, because those were the same people that got themselves into the mother of all flamewars when they told the mac community that the G4 had lost the race already years ago.

Too bad my German is not too good!

im not much into macs anymore (my last mac was running macOS seven point something, that should give you an idea), but the numbers the G5s scored in these benchmarks were pretty impressive. especially under logic. note that they also tested some of the (now discontinued) single CPU G5s, which didnt have the performance to justify their price. it was impressing to see how good the dual models scaled, all the while not being that much more expensive. (relatively speaking of course)

Okay, thanks for the info!

[Durk]

on another Dothan note, aren't they basically low voltage pIII's with 2mb cache slapped on? the p3 clock for clock always did destroy the p4 anyway.

Yep that's true. Based on the PIII, but with a lot of modifications.

Actually, now I'm thinking about it, that might be also the reason why it's not scaling to amazing heights... Talking about really released processors; the PIII on 0.18 micron never reached higher clock speeds than 1 GHz. The PIII on 0.13 micron never reached anything higher than 1466 MHz.

The Pentium M on 0.13 micron was released in variations from 1.3 GHz to 1.7 GHz? And the 0.09 micron Pentium M is currently released from 1.7 GHz to 2.0 GHz. Not too big steps either. Ah well, we'll see what happens! It's at least consuming less power than the Pentium 4 Prescott..

[derek]

<<Wow, that message must have taken time!>>


dont worry, im a fast typer :-)


<<As you know, Pentium 4 Northwoods were easily overclocked as well. But those speeds never made it to the shop. 3.4 GHz was the max, but overclockers went a lot higher. Same goes for AMD Athlon XPs.
I don't expect the Dothan running higher than 2.2 GHz.>>


just so you dont misunderstand me, i really have no "platform-religious" affection for that dothan CPU in any way. if at all, id have to admit that i like to talk in favor of AMD whenever i can :-)

its just that what i read in various magazines sounded quite a bit different. they tested how much they could overclock the dothan with a regular desktop CPU cooling, and that resulted in 2,4 or 2,5 GHZ i think. and were talking about the first dothans that were released, and were released with laptops in mind.

plus, the dothan *as it is* rivals currents fastest desktop CPUs, and that is still with the slower memory bus, something that is said to be a rather easy change for intel. and not to forget, it does show this kind of performance in laptops, which by nature are less friendly to low latency audio in general

to me that doesnt sound like "its on its limit" but rather like "theres a lot of potential".


<<The problem with testing laptops is all the different components. AMD-based laptops are more often sold as budget or mid-end solution than Intel-based ones. So, which one will have better batteries?>>


there isnt all that much differences in the batteries, assuming models with comparable specs. and the point is, almost all centrino laptops ranging from rather cheap to rather expensive show this impressive running time, its more a question of "very good" vs. "awesome".

of course i would find it very cool if we had AMD laptops that at least all in all would play in the same league. but as it seems, right now thats not the case.

OTOH, if the AMD 64s are good in this regard, that means we can hope that AMD can "strike back" rather soon. i would be the first to welcome that.


<<Just wondering.. Why do I look biased when I'm saying something about numbers in specifications?>>


i didnt mean you but the guy you quoted. and the bias lies in the way you look at numbers. theres still always a lot of room for interpretation in either direction, and i just had the impression that this guy had a tendency to interpret the numbers in favour of AMD whereever he could.

again, not that i wouldnt like that. i just dont think thats whats gonna happen (for better or worse).


<<Too bad my German is not too good!>>


in this case, thats really too bad. these benchmarks are really awesome. they also always come with explanations about why this plugin would perform better on this platform and that plugin better on that platform, its all done with a lot of knowledge, comparable to your regular ultra-nerd-computer magazine, all the while keeping the focus purely on what counts for the working musician. its a great combination, and it seems they dont fear pissing of either "platform-religion", be it intel, AMD or apple.

ive thought on several occasions that it would be a good idea for the english keyboards magazine to just translate this stuff, theyve done that in either direction every now and then.

[Durk]

just so you dont misunderstand me, i really have no "platform-religious" affection for that dothan CPU in any way. if at all, id have to admit that i like to talk in favor of AMD whenever i can

Don't worry, I'm not as bad as I may look.. I just like hefty discussions, but maybe I'm getting too serious..

its just that what i read in various magazines sounded quite a bit different.

Okay, no problem. Let's wait and we'll see.

plus, the dothan *as it is* rivals currents fastest desktop CPUs

Well I rechecked the AnandTech review, it's not rivaling the fastest desktop CPUs, but only the faster desktop CPUs. At least in the area AnandTech tested. Take a look here.

, and that is still with the slower memory bus, something that is said to be a rather easy change for intel.

Yes it's easy. But they won't do this for laptops, because it takes a lot more power. And otherwise they would have done already. The Pentium M for desktops.. yes, in that case they would increase the bus with a large amount. Also note that the Pentium M will not profit as much as the Pentium 4 does with a bus speed increase.

i didnt mean you but the guy you quoted. and the bias lies in the way you look at numbers. theres still always a lot of room for interpretation in either direction, and i just had the impression that this guy had a tendency to interpret the numbers in favour of AMD whereever he could.

Yeah I know it wasn't me you were aiming at. I didn't express myself very well. I wasn't trying to put down a biased story. As for me it didn't sound very biased. Actually, I did some polishing in his story. He implicited that there would be a pretty big difference between TDPtyp and TDPmax. So I did a check and it was only 5W.

<<Too bad my German is not too good!>>

in this case, thats really too bad. these benchmarks are really awesome. they also always come with explanations about why this plugin would perform better on this platform and that plugin better on that platform,

Well I'm able to read German, but it will take some time, so the article should better be good!

Can you give me some names of those magazines?


Concluding, I think it's a very interesting move of Intel; bringing the Pentium M to the desktop. I'm very curious how this will happen! For now, I won't say it will be easy...

[Lunch Money]

Mariah Carey?

Surely Enya must hold the record for track count.

Interesting thread, BTW... sometimes a bit circular, but I'm starting to lean towards just getting an Athlon Barton 2500+ while I'm waiting to see what the future of 64-bit holds in store. It'll cost $100, and the Athlon 64 I was looking at will probably drop $60 or more by the time I get it-- making it only a $40 loss (or less!) in the long run, for the sake of more processing power TODAY.

Greg