Sign in with
Sign up | Sign in
Your question

Dual-core K8s - 21st April!

Last response: in CPUs
Share
April 9, 2005 7:00:03 PM

Good news:
<A HREF="http://www.theinquirer.net/?article=22289" target="_new">AMD brings forward dual core Opteron launch </A>
<A HREF="http://www.theinquirer.net/?article=22404" target="_new">AMD dual processors to be available on launch day</A>

Also, Intel is making smithfield available starting in May. This probably marks the availability of dual-core products for the average buyer, with the 2.8Ghz Smithfield costing US$240.

Note the difference: Intel is introducing mainstream parts. AMD will <i>start</i> with 8xx series, then move down to 2xx and 1xx Opterons, and only then will they probably offer the more desktop-oriented FX. Heck, at least that will justify the price difference between standard A64s and AFXs... And in a more poetic justice, Intel will price their dual-core processors, which are probably pathetic if compared to AMD, as cheap processors, and AMD will ask for big bucks for a kickass processor.

In any case, I hope the news about 65nm tech (<A HREF="http://www.theinquirer.net/?article=22041" target="_new">Intel to release 65nm tech early</A>, article from Theinq) is accurate. Intel truly needs that. Heck, if they could get Yonah out before the year comes to an end, AMD would be caught completely off guard on the mobile front, for instance.


<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/09/05 06:02 PM.</EM></FONT></P>
April 9, 2005 8:14:54 PM

>Note the difference: Intel is introducing mainstream parts.
>AMD will start with 8xx series, then move down to 2xx and 1xx
> Opterons, and only then will they probably offer the more
>desktop-oriented FX.

Which of course, makes perfect sense. The real added value of multicore is primarely in servers and high end workstation, much less in desktops, and I have a hard time seeing any benefit for mobile use. For mobile having two cores means a laptop that must have its thermals specced to accomodate two cores running full speed, which means bigger, heavier, noisier laptops, where battery requirements will apparently disable one core when on battery power. net result: big heavy laptops with relative poor performance when unplugged, and questionable performance advantage when plugged in. what a step back from Banias/Dothan !

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 9, 2005 9:33:33 PM

Well Yonah is supposed to only use 31W so I dont see that as a step back.

Some people are like slinkies....
Not really good for anything but you cant help smile when you see one tumble down the stairs.
Related resources
April 9, 2005 10:48:06 PM

Well, the strongest selling point for Yonah, I suppose, would be small form factor PCs and blade servers. Heck, it would be in theory possible to make the whole Potomac structure obsolete with just a few yonah chips (these run at 667Mhz FSB, mind you) and then put some 2-socket or maybe even 4-socket Yonah systems on very dense systems. Yonah is a great chip. I'd be very happy if I had a dual-core yonah sitting silently in my desktop system, for instance... if it could be clocked high enough, mind you.

Agreed though, Intel's strategy isn't practical... But I think the idea of dual-core everywhere is not that bad... Dual-core laptops sounds like a nice enough concept for me, if indeed they have low thermals.<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/09/05 09:55 PM.</EM></FONT></P>
April 9, 2005 10:50:49 PM

Quote:
Well Yonah is supposed to only use 31W so I dont see that as a step back.

What a Yonah laptop probably has is a very, very long longevity. It would be reponsive as hell as a desktop-replacement part, and it would be quite usable for a lot of years. For instance, I have a dual-CPU P3 933Mhz system and a 1Ghz Laptop readily accessible; I can tell you, the difference is HUGE. That damned laptop gets all clobbered up with background tasks and is annoying to use... Unlike my system. And no, it's not just the video card...
April 9, 2005 11:00:12 PM

ist most likely the harddisk. Thats what makes 99% of the laptops SLOW, not matter what cpu you put in there.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 9, 2005 11:02:43 PM

>Well Yonah is supposed to only use 31W so I dont see that as
>a step back

Its a 50% increase over Dothan, is that not a step back ? further more, I fear two cores will leak quite a bit more than one, which means worse battery life. TDP doesn't mean much for battery life, unless you like to fold on batteries, power consumption in sleep mode is what counts, and where Dothan (and Banias even more) really shines. Will Yonah be as good ? I doubt it..

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 3:18:27 AM

I'm thinking you're being way too pessimistic about Yonah there. Dual-core, be it as it may, is the future. And Yonah can do that well. And it will be quite more elaborate than anything dual-core AMD can throw back for a while on the mobile front. And it will still be possible to turn one core completely off if you don't want it and keep both processors on if you're plugged in. I'm not too worried about leakage, but hey, we'll see that once reviews start showing up by the end of the year, hopefully. It will, I guess, depend on how they "turn off" the second core... We'll see.

In any case, I think you're just too cautious and pessimistic there... These are, of course, only my thoughts... but I also think you're right about notebook/laptop hard drive problems. No point having such a great processor as Yonah if you have such a performance bottleneck in the first place. But Yonah is, from where I'm standing, a step in the right direction.

<i>(BTW, I really love dual-cpu/cores, as I'm currently posting from one such system. Just to show their versatility, just yesterday, I was creating a presentation with powerpoint and it struck me that I had the following processes running: (1) 2 active instances of Prime95; (2) two firefox windows; (3) one word document; (4) Realplayer playing a few selected mp3 files; (5) Powerpoint - for the presentation; (6) Coreldraw - for creating nice images; (7) 3DSMax - for rendering 3D on top of coreldraw-created backgrounds; (8) Thunderbird checking for new mail every 5 minutes; (9) AVG Free Edition checking all incoming mail and opened files.

Not bad at all, eh? Though I must admit my hard drive had a hard time for a few moments... And I never got a single system freeze or lock-up or anything!... I didn't even have to worry about switching Prime95 off! Dual-cpu/core is definitely my thing. So I suppose I could be defending what <b>I</b> think is important, not what I think is best for everyone in the computing world, but hey, how could I do otherwise? I have to defend my own interests...)</i>

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/10/05 02:20 AM.</EM></FONT></P>
April 10, 2005 9:00:10 AM

Well Intel did say that yonah would have the same battery life as dothan and yonah should be better then dothan/banias in sleep mode as it has sleep transistors, body bias (well i'm pretty sure its got it) and other crap to lower power consumption in sleep and idle modes, seems like a step forward to me.

Some people are like slinkies....
Not really good for anything but you cant help smile when you see one tumble down the stairs.
April 10, 2005 10:17:48 AM

>I'm thinking you're being way too pessimistic about Yonah
>there.

Maybe.. time will tell.

>Dual-core, be it as it may, is the future

No its not *the* future, its only a part of it. Single threaded performance will probably matter for decades to come, in varying degrees it might often even matter a whole lot more than multithreaded performance. Yes, quote me on that even 10 years from now.

>And it will be quite more elaborate than anything dual-core
>AMD can throw back for a while on the mobile front.

What makes you think that ? Lets first see how Turion pans out, I think it could be a pleasant surprise. DC Turions are also on (unofficial ?) roadmaps, and by the time Yonah is here, AMD should be moving to 65nm and might put up a good fight.

>And it will still be possible to turn one core completely
>off

This, I'm curious about. I don't see how they will do that. Say you are running a multithreaded app while on AC, then you remove the power plug. what will happen ? I'm not even sure you can do something similar on most high end servers, just pulling a cpu while the system is on, unless you are running two mirrored cpu's.

Furthermore, I'm curious just 'how disabled' it will be on batteries. Again, its no small feat to completely shut down half the die and wake it up again on the fly. I suspect the second core will be put into some deep sleep mode instead, so still leaking. That said, maybe leakage is hardly a problem on the new 65nm process, so it might not be a big issue.

>In any case, I think you're just too cautious and
>pessimistic there...

Yeah I tend to be just like that, and, not to tooth my own horn, but I tend to be far more often right than wrong in all my "pessimism".

>BTW, I really love dual-cpu/cores, as I'm currently posting
>from one such system. Just to show their versatility, just
>yesterday, I was creating a presentation with powerpoint and
> it struck me that I had the following processes running:
>(1) 2 active instances of Prime95;

Why would anyone run prime ? just curious..

> (2) two firefox windows;

Impressive. Doesn't require any cpu resources though, just memory allocation.

>(3) one word document;

Impressive. Doesn't require any cpu resources though, just memory allocation.


> (4) Realplayer playing a few selected mp3 files;

Impressive. Doesn't require more than what 5% cpu resources though, just memory allocation.

>(5) Powerpoint - for the presentation;

Impressive. Doesn't require any cpu resources though, just memory allocation.

>(6) Coreldraw - for creating nice images;

Impressive. Doesn't require any cpu resources though, just memory allocation.


> (7) 3DSMax - for rendering 3D on top of coreldraw-created
>backgrounds;

unless you where rendering in the background (in which case, what the hell would you run prime as well for ?): impressive. Doesn't require any cpu resources though, just memory allocation.

> (8) Thunderbird checking for new mail every 5 minutes;

Wooow. That is what, 5ms cpu time every 5 minutes ?

>(9) AVG Free Edition checking all incoming mail and opened
>files.

Wooow. That is what, 5ms cpu time every 5 minutes ?

Seriously, people tend to like dual cpu machines because they do not know how to work with taskmanagers thread priorities (and windows scheduler is so f* up you need to bother in the first place). Let me tell you I can do all of the above, *plus* run a cpu intensive game with no visible slowdown on my single cpu machine. Surprised ? You only need enough RAM, and if required, assign a lower thread priority to background tasks.

Oh and one last thing; most people that rave multitaksing capability of their dual workstation, in reality just notice how much difference a SCSI drive setup makes over IDE in such scenario's. I'm fairly certain if I where to pull one cpu out of those dual workstations, most people wouldn't even notice. But replace the SCSI with an IDE, and they will start complaining about slowdowns and lockups.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 12:05:58 PM

Wow, Harsh...

So all those things that you say "Doesn't require any cpu resources though, just memory allocation"....So those require none? Absoultely NO CPU resources? So if i were to pull ym CPU out you could run all those things? I mean if it means none then i wouldn't need a CPU.

Yes, a more responsive HDD would be better in some cases, but think about it like this.....every little bit helps. and just because its not the GREATEST thing you can have doesn't mean its not a good step forward.
Just because i dont drive an Enzo doesn't mean my Corvette isn't still very quick.



What about this for dual core, which is what i actually do.

Run winamp with AVS running, Trillian, Firefox, Photoshop editing pictures for my mom who can't ever take a good picture, compressing 40-50pictures at a time to e-mail to check if she likes what i did.

Now granted not all of those require a lot of CPU cycles, but at the same time it would be nice to not have to wait so long for things to open, close, compress, compile, and filter.

You may be right. It may be overhyped about some things. But there are people out there like me who are looking forward to doing a lot fo things just a bit quicker and without the delay. And i am interested in dual core, but i still dont think will upgrade to it for maybe a year or two.

__________________________________________
April 10, 2005 12:26:08 PM

>So all those things that you say "Doesn't require any cpu
>resources though, just memory allocation"....So those require
> none? Absoultely NO CPU resources?

An iddle app will indeed consume no CPU resources, it will only consume ram (or VM). Or well, not anything more than processing mouse or keyboard input, and I have yet to see anyone claim you need 2 cpu's to handle that :D 

>Just because i dont drive an Enzo doesn't mean my Corvette
>isn't still very quick.

Heh.. only that DC processors might be more like a 4x4 in the city. Great at some things where you need 4 wheel drive, but lower speed, higher fuel consumption etc for the things most people use that car for.

>Run winamp with AVS running, Trillian, Firefox, Photoshop
>editing pictures for my mom who can't ever take a good
>picture, compressing 40-50pictures at a time to e-mail to
>check if she likes what i did.

You will get zero benefit. Well, not entirely true, photoshop has some multithreaded filters, so DC might give you a nice speedup on those. But none of your background tasks are anywhere near CPU intensive, and you are not compressing those images at the same you time you are editing them. The other apps are basically iddle, and do not require more than a couple percent CPU power when active. If you would be scanning your harddisk for virusses, it would be slightly different, but frankly it doesn't really matter if that task finishes in 10 minutes or 15, so if it runs at a low priority (which it should), this really would not justify a DC chip either since there would be no noticeable speedup for your foreground apps. If your computer does slow down when scanning for virusses, blame the harddisk first.

Don't get me wrong though, I'm not claiming there are no good uses for DC chips, its just that "multitasking" for most people consists of taskswitching between apps that are either iddle or *extremely* low on CPU resources (like IM, virus, firewall, etc) in the background. For those people, DC isn't the answer, a better disk setup would make all the difference in the world.

I'll give you one real world example of where it would help; I play IL2 flightsim quite a bit, and when I do, I run cam2pan which uses my webcam to track my headmovements and lets me pan inside my virtual cockput. cam2pan needs to analyse those images in realtime, at >30FPS, and this does cost ~10% of my cpu cycles. A second core would give me a slight benefit here, since neither IL2 or cam2pan are ever iddle, and both more or less cpu hungry, and I need realtime responses from both (unlike rendering, virusscan, and other background batch jobs).

Now at the same time, I will also run Teamspeak, which needs to compress and decompress audio in real time, a firewall, an antivirus, MSN messenger, Skype, Google notifier, Outlook, Firefox with ~10 open tabs, an antispyware, MBM, Emule, and then some apps, but none of these cause as much as a visible spike in my CPU usage using Perfmon, let alone they would require or benefit from their own core. All of them combined don't exceed 1% of my CPU cycles.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 1:24:39 PM

I see what your saying, you can only be using one program at a time and thee rest would be sitting idle.
Well thats true and not true for what i said.

The Winamp AVS sends my cpu usages instantly to 100%. also you are right, i am not compressing and editing at the same time. but what i am doing is doing a batch edit straight into another batch edit straight into a batch compress all while i am surfing the web, having winamp open, and chatting on Trillian. For 50 pictures it would take me about an hour if i am goofing around surfing but only 15min if i do nothing.


To me its not about making one task faster. I dont expect a 100% increase in speed. But i do know i will be able to run things smoother. And knowing that i can run more programs i will.

Your right, but people buy cars ALL THE TIME that do not suit the needs of the driver. Hence everyone not driving Geo Metros. How many people do you know who own SUV's that do nothing but drive to and from work and shuttle kids. Gotta love a 350HP V8 just to sit in traffic.

All i am saying is that when i know i can run more applications without feeling a lag i will. My ex-g/f is using WinXP with 128MB of RAM. Not exactly optimal. but it works for what she wants. when i use her computer i try to limit what i do because i know i can't do a lot. So as my abilities become more available i will do more things.

because as of now i can't convert DVD's to AVI's without not using my computer for the next couple hours. and it would be nice to do something like that and be able to use my computer without a huge lag in the compression or websurfing. Yes using one of those DC's for compression of a movie and one for surfing may not be as fast as this one core i have now. But like you said, web surfing and all that jive uses 1% of my cpu cycles....so i dont need a 4Ghz processor. a 1Ghz AMD with no lag while another is compressing or having it do something more instensive....that would be nice.



just one more thing....Zero benefit? are you a black and white kind of person? its either HUGE or Nothing? Granted none of them are cpu intensive but if i were using a computer that maxed out at my everyday activities i would still be using my oldass P2 300Mhz POS. I dont use the full potential of my 6800GT surfing the web either....but you know what? its nice to know that if i want to use it for select things i can. Same with DC's. Its nice to know that i can do somethings quicker if i wanted.


::Edit for spelling and rediculous typing errors::
__________________________________________
<P ID="edit"><FONT SIZE=-1><EM>Edited by pickxx on 04/10/05 06:28 AM.</EM></FONT></P>
April 10, 2005 1:49:01 PM

Quote:
because as of now i can't convert DVD's to AVI's without not using my computer for the next couple hours. and it would be nice to do something like that and be able to use my computer without a huge lag in the compression or websurfing

You can do it with single core CPU, just set the CPU usage priority of encoder to "idle". Websurfing requires very little CPU time, so you'll notice no lag in websurfing and very little lag in encoding.

Quote:
a 1Ghz AMD with no lag while another is compressing or having it do something more instensive....that would be nice.

I used to do the same thing with my 1 GHz Duron without using anything special, standard 7200 rpm HDD and 512 MB RAM. Intelligent use of process priorites really can do wonderful things.

However if the encoder has dual CPU support, then you'll be able to finish encoding faster. Otherwise, there's no benefit from dual core in this scenario.


------------
<font color=orange><b><A HREF="http://www.mozilla.org/products/firefox" target="_new">Rediscover the web</A></b></font color=orange>
April 10, 2005 1:49:39 PM

>The Winamp AVS sends my cpu usages instantly to 100%.

Then the program isnt coded properly. It won't use anywhere near 100% unless you own a 486-33.. Its possible somehow all iddle loops are being "used" by winamp as well, but rest assured 95% of the time, your cpu will be iddle and available for any other app. Just try running a game, do you get zero FPS now, or does winamp stop playing ?

I just tried with Core media player playing 320kb/s MP3 file, and I average around 4% on my XP3000. However, I also have a misbehaving app (route66, a map planner) that will display 100% cpu time even when iddle, and will also increase the temp of my cpu. But whatever other app or benchmark I start even with Route66 active, there is no performance impact whatsoever, so neither will there be one by adding a second cpu. These apps just seem to have a bug and claim any free cpu cycles for themselves; however, without monopolizing the cpu.

>i am not compressing and editing at the same time. but what
>i am doing is doing a batch edit straight into another batch
> edit straight into a batch compress all while i am surfing
>the web, having winamp open, and chatting on Trillian

Depending on what these batch edits do, a second core could help or not, since some photoshop filters or actions are multithreaded. For all the rest, really, there is not going to be ANY increase. Chatting on trillian ? surfing the web ??
Any stutters or slowdowns you might encounter are 100% certainly caused by the harddisk and/or not having enough RAM. adding a second cpu will simply not improve your trillian or websurfing experience, wether or not you have a cpu intensive background app, period.

>To me its not about making one task faster. I dont expect a
>100% increase in speed. But i do know i will be able to run
>things smoother. And knowing that i can run more programs i
>will.

If you want smoother, get more ram, or better yet, get SCSI. You may also want to ditch windows for Linux, and you'll be doing all of the above while burning DVDs and compiling openoffce with a Celeron or Duron with no slowdowns or hickups.

>Your right, but people buy cars ALL THE TIME that do not
>suit the needs of the driver.

Its true, and many people will buy into the DC hype just like you, but that won't stop me from making my point.

>because as of now i can't convert DVD's to AVI's without not
> using my computer for the next couple hours

Really ? Then either add more ram, and/or open taskmanager and assign a lower thread priority to the DVD/AVI conversion. You'll be able to do anything you want while converting in the background. Now if you'd choose to play a state of the art game at the same time, it should still work, but you might get some stuttering from HD access (get SCSI) and mostly, the conversion will take considerably longer. Adding a second core will only cure the longer conversion time, nothing else.

>so i dont need a 4Ghz processor. a 1Ghz AMD with no lag
>while another is compressing or having it do something more
>instensive....that would be nice

Get SCSI, get more RAM, and/or assign lower thread priorities to your background tasks.

>just one more thing....Zero benefit? are you a black and
>white kind of person? its either HUGE or Nothing?

I'm anything but black and white, but when its zero, its just that. In fact, depending what alternatives you consider, and what scenario you are discussing, the net result could be a slowdown. Reality is, for most people, and for most current apps, that is the case.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 2:27:14 PM

Quote:
>And it will be quite more elaborate than anything dual-core
>AMD can throw back for a while on the mobile front.

What makes you think that ? Lets first see how Turion pans out, I think it could be a pleasant surprise. DC Turions are also on (unofficial ?) roadmaps, and by the time Yonah is here, AMD should be moving to 65nm and might put up a good fight.

Several small details point to that, actually. Yonah is to have TDP of <b>31W</b> as far as I know, or at least around that, with both cores on and running. And hey, that seems to be, indeed, 15% more than Dothan, but who cares? I'm pretty sure that, leakage or no leakage, if you want to keep the damned thing only on one core, the TDP will go down by more than a measly 4W, so on one core, it will probably consume much less than Dothan.

On other fronts: It will feature a <b>shared</b> 2MB L2 cache. Does any other processor from Intel and AMD have that? Nope. It will feature improvements in SSE2 and SSE3 will be included; plus, there'll be an improvement in Dothan's low point, which is FP calculations. Add to that a 667Mhz FSB and you're probably set: this is quite a great chip. If it were to be launched preemptively in 3Q05, it would make AMD sweat... They could try rolling out low-voltage A64s like they are doing with turion right now, and we know that AMD's low-voltage DCs will only consume, say, 30-40W, but then again, 31W is for a fully-fledged Yonah, and there are low-voltage Yonah chips that run on a lot less wattage. So added IPC, shared cache, rather low power consumption and speed target of something in between 2.33 and 2.67 isn't good enough to scare AMD if released in 3Q05? I think it is. AMD won't have 65nm tech until well into 2006. And this thing will have IPC parity with A64s anytime. So, OK, you pay a +4W TDP power premium indeed, but what you get for this is a very, very robust and powerful laptop chip. And that TDP will probably get down below Dothan levels with one core off. So, where is the actual problem? I understand you're being cautious, but I'm not sure you need to be too pessimistic...
Quote:
Why would anyone run prime ? just curious...

<A HREF="http://www.mersenne.org/prize.htm" target="_new">A very small chance of earning some money</A>. Granted, it's just like 1 in 300000, but I have currently 6 systems (my dual CPU one counting as two) running that, which makes the chance go a little down to 1 in 50000. But heck, some people keep protein folding software in the background - and that doesn't have the slightest chance of earning them more money...

As for the responsiveness of the system under load, I was, as a matter of fact, constantly exporting to jpgs in corel while 3DSMAX rendered something. And the systems stays more responsive than in single core/CPU systems, that is just quite simply a fact. Prime95 didn't bother that much because it was, of course, at lowest possible priority. So I could actually render in 3DSMAX - BTW, with great improvements in speed due to multithreaded code - and the system would resume processing Prime95 after the rendering was finished.

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/10/05 01:30 PM.</EM></FONT></P>
April 10, 2005 2:28:15 PM

i know what i CAN do to speed things up and make things run smoother. And i never said i could throw one of the DC's in and be like spuer man. I said i would wait a year or two to upgrade to one. And by then i figure some of the programs i use will have a little more DC support.


Do this...Download winamp, run the AVS at full screen res (800x600 is fine for this) and dont use any pixal doubling or any half res. settings. And then try to surf the web....THEN tell me how winamp is faking it.
The batch convertions are usuallly filters to normalize the picture colors, and then to reduce the size to something easier for my mom to print. I dont know if those would be helped or not.

And saying that i should switch to linux is a non-solution for me....i haven't used it and frankly i have no interest in it. i dont like swtiching things around like that.



I am aware i can set the priority lower, but that creates longer compression times. I know what i can do to work around the limitations i currently have. but even with 2GB of RAM or whatever you want me to have, the cpu still have to crunch the numbers to compress it.

I just see this to me as something that would be nice to have every once in a while. Like a higher end gpu. I would like to set one core to compress and finish it in the same time as now without me touching anything and another core to allow me to do stupid stuff that only uses 6.287% of my cpu cycles or whatever.

I am not looking for an super speed boost but like i said, it's something that would be nice to have everyone once in a while. And i will take little to no hit at all in everything else that doesn't use it. because i already am overpowerd in the cpu for using winamp, firefox, and all that other shite.

And dont say i will get NO gains...and then tell me i will get a certain % gain sometimes. People on here spend hundreds of dollars for small %'s.

__________________________________________
April 10, 2005 2:49:32 PM

I think he's underestimating the fact that, put quite simply, DCs have twice the possible processing throughput. Also, I neglected to mention in my previous post to P4Man that I'm a physicist (I'm sure I've mentioned that a thousand times in here already) and I do a lot of number crunching and programming. So I can - and actually do - really, really use two processor/core systems for what I do. I've just started programming for multiple threads a few months ago, and I can always keep two instances of a program running in some cases - for instance, simulations in which I can compare both outcomes and/or run a series of simulations for statistics.

Now how does that added <b>+100% throughput</b> look? It just quite simply cannot be beaten.

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 04/10/05 01:51 PM.</EM></FONT></P>
April 10, 2005 3:32:15 PM

>Do this...Download winamp, run the AVS at full screen res
>(800x600 is fine for this

oh, i didn't realize you where running some visual dancing whatever stuff. Yes, that will seriously eat into your cpu cycles, no discussion there. Though just where just playing back the MP3s. But then.. what good is having that as a background app ? Well, unless you have 2 monitors, and then, yes you will see a pretty huge performance boost.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 4:01:30 PM

Yes, i have two monitors. Usually one with that dancing crap and one for Trillain, Firefox, writing papers, or whatever.

Its not something i NEED to do, but its nice to know i can do it. If i were just listening to MP3's its not a huge hit. I usually hover right around 10-15% surfing the web and having mp3's, chat programs, firefox, gmail notifier, SysMetrix, and all that jive running in the background.

Without programming my own program how much freedom will i have with DC's?
I was under the assumption, right or wrong i dont know, that i can assign 1cpu to something or somethings and the other to the rest pretty easily. Or will windows/the cpu decide how to divide out the tasks?

__________________________________________
April 10, 2005 4:17:09 PM

>Several small details point to that, actually. Yonah is to
>have TDP of 31W as far as I know, or at least around that,
>with both cores on and running. And hey, that seems to be,
>indeed, 15% more than Dothan, but who cares?

Actually, that is between 50% more and 3x (!) as much as Dothan, which has a TDP of 10-21W afaik. And you should care, if you care about thin and light.

See there are 2 issues with notebooks in this context: TDP and iddle/sleep power consumption. THe first one is going to define the cooling setup. Even if Yonah would rarely use 31W IRL, the cooling solution has to be able to accomodate that much, and 31W is just a nono for ultraportables. Its the same reason you dont see any ultralights with a mobile A64, even though it "only" boasts a TDP of 35W if I'm not mistaken. Its just too much to handle.

The second issue, iddle/sleep power consumption will for most people play a much more important role in determining battery life than TDP. Most laptops are iddling 95% of the time on batteries, using things like word, internet, email etc (Few people will actually run a game or 3DS on batteries, but those that do, should care about TDP more). Now Banias/Dothans excellent battery life is mostly due to the fact it uses incredibly little power when iddling. Its *far* better in this regard than for instance mobile A64, much more than its TDPs would lead you to believe. And most of the current still used, is, you guessed it: leakage. So you can't just say you don't care about it, because it *will* to a very large degree determine battery life. With dual cores, chances are Yonah will be (considerably) worse in this regard, unless intels 65nm provides very significant improvements here (not unlikely, but not a given either).

So combined, with Yonah compared to Dothan, you might well end up with a pretty much bigger laptop, with much more limited battery life and only sporadic advantages of using dual core (and even then, only unplugged). The first one (bigger, heavier) is pretty much a given (unless they also bring out ULV versions with sub 15W TDP), the second one (worse battery life) no more than just a guess/fear.

>On other fronts: It will feature a shared 2MB L2 cache. Does
> any other processor from Intel and AMD have that? Nope

The jury is still out if a shared cache is better than a unified cache for DC. My guess is it will depend on the app, but I would not expect a big difference either way. Further more, bigger caches result in diminishing returns. Going from banias (1MB) to Dothan (2MB) gave only a tiny performance boost per clock. Probably the reason intel is going back to 1MB per core (if I read you right).

>Nope. It will feature improvements in SSE2 and SSE3 will be
>included; plus,

SSE2 is already in there (Dothan), and SSE3.. see AMD64 venice benchmarks, the benefit is so small, its completely neglectable.

>plus, there'll be an improvement in Dothan's low point,
>which is FP calculations

Thats good, it needs a better FPU badly. It also needs better SSE2 performance, maybe that will be reworked as well ?

>Add to that a 667Mhz FSB and you're probably set: this is
>quite a great chip.

We've seen overclocked results of Dothan on 680 MHz bus, and the increases over 533 are underwhelming.

>If it were to be launched preemptively in 3Q05, it would
>make AMD sweat...

I don't think so, AMD has no marketshare in the mobile market anyway, they have nothing to lose :)  Seriously though, I'm not saying it will be a bad chip, far from, but I don't quite see it being a killer chip either. Single threaded (unplugged) performance it will have a hard time against even 90nm Turions. Multithreaded performance is a questionable requirement for most mobile uses, and it AFAIK it will still lack 64 bit support; now that is not such a big issue for a laptop either, but it still bites to market a highend laptop that is unable to run the latest version of windows. As for battery life, its anyone's guess, but I have seen no evidence it will significantly worse or better than Turion, but it looks like it might be a step back from Dothan.

>I understand you're being cautious, but I'm not sure you
>need to be too pessimistic...

I understand you're being optimistic, but I'm not sure you need to start hyping it already :)  You're making a lot of assumptions and expressing hopes of early releases and expectations of late AMD releases.. I just wouldn't bet money on either. How many release dates has AMD missed these last years ? How many did intel miss ?

One last point: I would not underestimate AMDs Turion offensive too much. They have made a big point out of reversing their priorities from desktop>notebook>server to server>notebook>desktop. We've seen what it has accomplished on the server side with Opteron, and I've always expected a similar attack on the mobile market once it got to 90nm. It got there now, and first products are looking extremely good with rather dramatic decrease in power consumption and/or clock potential, so now lets see what that will bring to the mobile front.

My bottom line: I firmly expect 90nm Turions to beat Dothan performance wise, have similar TDPs, but probably lower real life battery life. Likewise, I guess it will be very competitive performance wise with Yonah unplugged, take a beating on SMT code plugged in (until it goes dualcore on 65nm, some 6 months later) and be competitive in battery life. For many, it will be a choice between plugged in SMT performance and 64 bit abilities all other things being roughly comparable. And a dual core 65nm Turion may well end up being the better choice 12+ months from here.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 4:22:07 PM

>Without programming my own program how much freedom will i
>have with DC's?
>I was under the assumption, right or wrong i dont know, that
>i can assign 1cpu to something or somethings and the other to
> the rest pretty easily. Or will windows/the cpu decide how
>to divide out the tasks?

Leave it to the OS scheduler, and possibly use thread priorities instead. But if you want, you can set CPU affinity in the taskmanager, where you can link one process to a certain CPU. This is not a "hard" link, but where possible, the OS will try to schedule the app on the preferred cpu. AFAIK, this is only really usefull when mixing virtual cores (hyperthreading) with physical cores, to avoid cache trashing or unnecessary cache snooping over the bus. For most apps, I wouldn't touch it, the OS scheduler should know better :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 4:38:02 PM

>I think he's underestimating the fact that, put quite simply,
> DCs have twice the possible processing throughput

Oh no I'm not; first its not "twice the possible processing throughput" as a dual core chip will most likely always be clocked lower than a single core variant. Secondly, I'm just warning against people expecting this potential to translate into real world performance. Two GF6800's in SLI also have twice the potential througput of one, yet in reality, its quite a bit less isn't it ? Sometimes, in spite of 2x the price tag, and twice the performance potential, it performs even *worse* than with just one card; now why am I having a deja vu ? :) 

Lastly, I'm getting tired of people thinking or claiming DC will somehow improve anyone's ability to run lots of (iddle) background apps or switch faster between them. Thats just bogus. Same rubbish as netburst speeding up the web. Very marketable, but very much nonsense nonetheless.

All this doesn't mean I do not see the potential of DC; everyone does. Too few people see the caveats though.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 4:40:18 PM

Nice to see that dual-core is coming soon With 64 bit right behind that. I'll have to see what 2006 has before I upgrade again. Longhorn due sometime 06. But not with all features. Looks like my 2.4C OCed to 3.0 will last me for a good while. Before I do my next upgrade. Would be nice to see IF Yonah makes it's debate by end of this year.
April 10, 2005 4:44:00 PM

I figured it would automatically do it but i would have some control if i needed it.

I have used duel p3 systems and i have seen how things can be good, and also how it can be bad. I haven't owned or played with one too long to know how well it works in every situation. It makes me kinda curious because of the system resources that are shared like HDD and RAM. Would DC need low latency and some RAID0 to give it full potential?

__________________________________________
April 10, 2005 4:47:17 PM

Try running all that on a single core system!!!!! Wait for the next few years when we have 4 core and then 8 core CPUs. Now that will be fun. Plus 128 Bit.
April 10, 2005 4:58:14 PM

>It makes me kinda curious because of the system resources
>that are shared like HDD and RAM. Would DC need low latency
>and some RAID0 to give it full potential?

No, not really. Raid 0 has no relationship with DC, since there will still be a single HD controller both chips talk to. thats where the bottleneck is usually anyhow for IDE disks, meaning that with one or two cpu's, RAID 0 doesn't tend to speed anything up noticable. SCSI is a different story (as might SATA, not sure ther). But there the benefit is just a big with single as dual core. Its not like a harddisk is anywhere near fast enough to saturate one cpu, it would be a bottleneck with 1/100th of a cpu :) 

As for memory, same story. RAM is already way too slow for a single CPU, so any speed increase (be it lower latency or higher bandwith) will surely help a single cpu system as much if not more than a dual one. The only thing that in theory could apply is, with intels approach of using two cores sharing a single FSB and going over the FSB to snoop the cache, that could make for a "new" bottleneck. Having a dual FSB chipset (like IBMs new Xeon server chipset), or an integrated crossbar (AMDs ondie solution) might have some benefits here that would obivously not make any difference with just one cpu/core.

BTW, its funny that you are considering these bottlenecks now in the light of DC. They have always been there, and are most often what have held back even your single cpu system. That is basically what I've been telling you all along, dual core won't do miracles under most circumstances (though it will for some), reducing the other bottlenecks (I/O, especiall disk I/O) will provide a far greater benefit for most uses people associate with "dual core", like taskswitching.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 10, 2005 5:17:09 PM

I know the bottlenecks are already in place but i was curious as to if two cpu's reading and writing would creat a bigger bottleneck.

If one HDD can't keep up with one cpu, then how would it hold up with two?

What i am asking is this, with the known bottlenecks already in place would something like a RAID or SCSI be EVEN MORE important or having more RAM or higher Bandwidth or low latency RAM be even MORE important then now because of all the information that the cpu would want to read/write.


Granted i wouldn't be using 100% of both cores all the time, but with both of them asking for information at the same time from the HDD, would it be more important or is there something built into the chipset or on-die that would address this?


__________________________________________
April 10, 2005 8:01:47 PM

I have to agree there regarding the concerns about leakage. Is it conceivable that it will be a problem? Yes, maybe. But I do think that AMD has nothing to throw back. I mean, AMD has no specific mobile line - Turion is quite simply just a LV A64. True, A64 is great, but Turion is already the LV model. Dothan and Yonah are built for mobile and consume 27W and 31W respectively <i>on normal models.</i> Low-power dothan chips consume sub-15W power figures that AMD cannot and will not match for the time being.

This all depends on 65nm leakage and leakage control techniques and Intel's engineering prowess. I'm thinking that unless leakage skyrockets, there won't be much of a problem, but that's just me. We'll see, anyway. For me, right now, a slight power premium to pay for such a powerhorse like, say, a 2.67Ghz dual-core Yonah chip within less than 12 months from now is a valid premium to pay. But then again, dual-core isn't necessarily a great idea if they can't keep leakage within reasonable levels in the second idling core...

In any case, a 31W dual-core Yonah chip at 2.67Ghz kind of makes me scratch my head in agony... Why not on desktops? Why not try to increase clock? Why not on 800Mhz FSB on desktops? Silly, silly Intel. Forget netburst. Forget smithfield. All they needed is 64-bit tech in Yonah. Mobile chip? Sure. Limited? Not suitable? Hell yeah. But still much better than two prescotts burning 100+W for sh*tty performance.

-> About shared cache: you're right, it might be worse, but I'm thinking that it's a good idea because you might want to keep one core off for battery purposes, and the added cache would be of some benefit. Don't know, though, because, like you said, it might not help that much. We'll see, though.

-> SSE2/3: Yonah will have a vastly improved implementation of SSE2. I know it's already in there, but there have been major enhancements in runnning SSE2 code. As for SSE3, well, it has yet to flex its muscles... It's near useless now, but a few commands could be brought to good use in the next few years. Granted, it's not a plus because everyone else will have it. Yonah's FP unit will also be more powerful than Dothan's - that has been pointed out repeatedly by a number of sources.

-> 64-bit - The main problem with Yonah. That one is a very bad point indeed, and it will make a difference. Got it, Intel?... I'd be sorely disappointed if a 2005 processor launch didn't include 64-bit technology. Stupid Intel. Hope they surprise people and put 64-bit tech in there. :mad: 

In any case, it has been quite a good discussion here. You make valid points... Don't think you're wrong, but I think there are also advantages to Yonah's approach that you simply have to throw into the mix... I do also understand that hyping it too much wouldn't be appropriate because it won't be the greatest thing since sliced bread, but I was just trying to point out that yonah sounds more or less like a good idea to me... :smile:
April 10, 2005 10:10:02 PM

>I have to agree there regarding the concerns about leakage.
>Is it conceivable that it will be a problem? Yes, maybe. But
> I do think that AMD has nothing to throw back. I mean, AMD
>has no specific mobile line - Turion is quite simply just a
>LV A64. True, A64 is great, but Turion is already the LV
>model. Dothan and Yonah are built for mobile and consume 27W
> and 31W respectively on normal models. Low-power dothan
>chips consume sub-15W power figures that AMD cannot and will
> not match for the time being.

Its not the low TDP that will be so hard for AMD to match, its the extremely low iddle consumption. Consider that 90nm A64's pretty much cut real world power consumption over same clock 130nm products with ~30%. Considering mobile A64's are at 35W TDP, 90nm ones should be able to match or even undercut Dothan at similar clockspeeds IMHO. But without the smart circuitry used in Dothans cache and other parts, the iddle power consumption may well still be higher.

As for Turion versus Yonah.. on 90nm AMD will have a hard (well, likely impossible) task of matching it power and performance wise, but not necessarely one of them. On 65nm, all bets are off, but Turion (A64) has a more advanced core, capable of clocking higher, supporting 64 bit, ODMC, etc. Much will depend on how quickly AMD can ramp 65nm and how good it will be. Intel's seems to be doing excellent, so there is the challenge for AMD. I'd say the race is still on.

>In any case, a 31W dual-core Yonah chip at 2.67Ghz kind of
>makes me scratch my head in agony... Why not on desktops?

You answered it yourself: no 64 bit support. Thats going to be a hard sell next year, especially as a high end part.

>Why not try to increase clock?

What makes people think PM can clock indefinately high, just because its cool ? The entire chip was designed around low power consumption, not high clockspeeds. The process was skewed for low voltage, the cache is designed to run at (extremely) low latencies at the expensive of clockability. There are always compromises to be made, and with Banias&sons, intel consistantly chose for low power. The result, is no matter what people think, you just can't turn it into a 3 GHz part. Not without major reworkings, including performance reducing ones (like making the cache slower a la prescott). There is no free lunch.

> Why not on 800Mhz FSB on desktops? Silly, silly Intel.

Its probably not as easy as we think it is.

>Forget netburst. Forget smithfield. All they needed is
>64-bit tech in Yonah.

Is that all ? that only implies a complete overhaul and redesign of the core and all its critical paths. That is a *major* job. A 64 bit enabled Dothan or Yonah is simply a complelety new chip, and that takes years. 4 or 5 is my understanding. Assuming they started on this ~2, maybe 3 years ago, it will still be a while.


= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 11, 2005 8:28:34 AM

Today, the opteron is head and sholders the best gp server chip availabe. The dual core will take that to a whole new level. And yet, the market penetration is still weak.
Can you say IT is dumb?
April 11, 2005 8:43:53 AM

>Can you say IT is dumb?

Not necessarely. IT departments often have widely different requirements than just getting the fastest cpu for the money. These can include things like:
* I/O or memory capacity
* density (cpu's/disks/memory/IO/.. per U)
* standardisation (of oem, server types, components..)
* managability and RAS features

Opteron has not really been available in various tier one solutions offering diferent mixes of the above until pretty much the beginning of this year. Considering typically long purchase cycles, validation/testing of those systems, sales will only pick up ~6 months from there.

Still, the future for opteron looks brighter than ever. I read it already has 20% of the 4-way (x86) market, which is very impressive considering this is probably about the most conservative market segment. analysts are also predicting ~20% marketshare in the entire x86 server market this year, which is a fourfold increase over their Athlon MP marketshare. And as you pointed out, dual core chips with no competition will surely help achieve or shatter these goals. Don't worry about opteron, worry about Turion :) 

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 12, 2005 7:17:10 AM

>In any case, a 31W dual-core Yonah chip at 2.67Ghz kind of
>makes me scratch my head in agony..

Apparently intel as well:
<i>The mobile dual-core processor "Yonah" appears to be on track for a Q1 2006 introduction with low-voltage versions of the 65 nm chip also appearing on the roadmap. Yonah will debut as x20, x30, x40, and x50 models with clock speeds of 1.67, 1.83, 2.0 and 2.17 GHz.</i>
<A HREF="http://www.tomshardware.com/hardnews/20050411_132505.ht..." target="_new">http://www.tomshardware.com/hardnews/20050411_132505.ht...;/A>

Running at 2.17 GHz single core on batteries, while requires a cooling solution able to cope with 35W.. I'm not sure AMD is really panicking yet. Not too mention its on track for Q106, not Q305.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
April 12, 2005 5:02:26 PM

I saw that too.

But the early expectations were 2.5Ghz ± 1 speed grade, which would indeed put it between 2.33Ghz and 2.67Ghz. It would be rather sad if this news turned out to be true, but, who knows, it might as well be.

There are three possible causes for this...

<b>(1)</b> They are rushing Yonah out the door ASAP and anyway to beat AMD;

<b>(2)</b> The 65nm isn't as good as expected;

<b>(3)</b> They are sandbagging.

Of these three possibilities, I'd agree with <A HREF="http://www.aceshardware.com/forums/read_post.jsp?id=115..." target="_new">groo from aceshardware</A> (actually, I quoted him entirely, sorry, groo): the 3rd is the most likely, followed loosely by the first possibility. There's also the 2nd, but it's the least likely of them all. Most people on many forums on the net all agree that there is an increasing likeliness that Intel is sandbagging right now. Centrino probably has the power to make the whole pentium lineup obsolete, but they're not doing that just yet because they're choosing not to do with dothan and yonah what they could be doing. A shame. It's pretty obvious that yonah probably can do 2.5Ghz indeed, not just 2.17Ghz. Heck, 90nm dothan can do 2.26Ghz!

All in all, groo points to a good question: Maybe they could clock it higher, but why would they?

Also, Intel guaranteed that Yonah chips would <b>not</b> incur in shorter battery life, but we'll see about that. Oh, and there'll also be single-core yonah chips as well... With the obviously lower TDP associated with the 65nm process...
April 12, 2005 6:15:56 PM

Quote:
Also, Intel guaranteed that Yonah chips would not incur in shorter battery life, but we'll see about that.

Yeah, but between single-core Yonah and possibly something like tighter restrictions on display power requirements, there are plenty of ways that a Yonah-based laptop can have the same battery life and yet allow a typical Yonah to use more power.

<pre> :eek:  <font color=purple>I express to you a hex value 84 with my ten binary 'digits'. :eek:  </font color=purple></pre><p>@ 185K -> 200,000 miles or bust!
!