hmm why is AMD following intel's mistakes?

[endyen]

<A HREF="http://www.overclockers.com/articles1236/index03.asp" target="_new">Reallah?</A>

 

[endyen]

[/url] [/quote] What's the dif, when you have friends.

 

[P4Man]

Just my 2 cents on this subject: AMD isn't making any mistakes with thier multicore strategy.

First, they are pushing it where it matters most: servers. Typical server loads are highly parallel by nature, so having more cores is almost always better. Current dual core opterons are really THE bargain of the moment, and Groo (Charlie from the INQ) hints at quad core opterons as early as 1Q06 even though official roadmaps point to 2007.

Secondly, for desktops, dual core is a usefull feature. For a given process node, you may loose a speedgrade or two due to higher power consumption, but you gain a potential doubling of performance, especially when running more than one app. Thats not a bad trade off. Keep in mind that power consumption goes up exponentially with clockfrequency and only lineary with # of cores. Of course, this requires apps that can make use of multicores, but running 2 cpu intensive apps isn't THAT exceptional, so its an okay trade off I think. More than 2 cores on the desktop will only make sense if ever software really makes use of this, which will be quite a while for most things I believe.

In the mobile market, I honestly fail to see the point of dual core, for all but the most exceptional circumstances. I can only guess intel had trouble clocking up Dothan, so instead they used the thermal headroom and small diesize of Dothan to make something no one really needs: a dualcore mobile chip. AMD seems to feel obliged to follow, but again, *I* don't see the point.

If you put it all together, if anyone is f*ing up their multicore stategy for the moment, is Intel. They will have better multicore mobile chips than server chips early next year, and no matter how you look at that, thats plain silly. It would be like having EMT64 on Dothan, but 32 bit only Xeons and Itaniums.

As for your concern about thermals of multicore chips; don't fear. This newly discovered holy grail of computing is actually born from the fact that power consumption is becoming such a problem, and that cpu's just don't keep scaling their frequency with process shrinks as they used to. IBM said it boldly: clockscaling is dead. Not quite true of course, but its obvious that single threaded performance increases are slowing down at an alarming rate. There just isn't all that much anymore left to improve IPC, and clocking is slowing down as well if you want to keep sub 150W cpu's. Multicore is just a way (soon the only way ?) to make some good use of Moore's Law when the other tricks start failing.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[mpasternak]

isnt multicore old technology?

arent PowerPC's and a few other CPU's for other types of architecture using it to some extent (whats sony's Ceell processor?).

it can b e doen.. will be done...

 

[apesoccer]

OMG...

Lets see... Scottchens dad come from China and married a Canadian, mast3rbate came from China and lives in US, and now you from China and live in Canada. Oh and Night_L is like me from Taiwan, but live in other country.

I always knew that the Canooks were planning on taking over the U.S....but i didn't realize the Asians were the driving force behind the Canadians!!!!

OMG! It's all becoming clear now! [The increase in destroyers, the second and third non-propeller powered attack planes (ie jets in laymen's terms)] Yea, I'm on to you crazy Asians! Don't think I don't know whats going on. I'm going to the top with this...Anybody know if Hilary is still in NY?

F@H:
AMD: [64 3000+][2500+][2000+ down][1.3x2][366]
Intel: [X 3.0x3][P4 3.0x2][P4 2.4x5 down][P4 1.4]

"...and i'm not gay" RX8 -Greatest Quote of ALL Time

 

[P4Man]

>isnt multicore old technology?

It certainly isn't a great invention, that is for sure.

>arent PowerPC's and a few other CPU's for other types of
>architecture using it to some extent

Depends where you draw the line. PowerPC certainly not, but Power (its high end brother) has been available for a while as MCMs, where several cores as well as huge L3 caches are combined into a single piece of silicon. Not too different from a Pentium D actually. See a pic of such a beast here:
<A HREF="http://www.theinquirer.net/?article=12145" target="_new">http://www.theinquirer.net/?article=12145</A>

"true" multicore however, isn't new either. I'm sure it has been around for decades in specialized chips, and most certainly for several years by Sun (UltraSparc IV) and HP (PA Risc- 8 thousand somehting, not sure 8800A ?).

At best you could say Athlon X2 and Pentium D are the first mass produced general purpose dual core CPU's . Big deal.

>(whats sony's Ceell processor?).

Nah, hmm.. well, not really. Cell is more an extremely wide cpu than a true multicore.. it depends on your definition. Either way, Cell isn't here yet, so it couldn't claim the crown anyhow. I do believe Cell is the way forward though; not that chip specifically, but a more traditional and powerfull x86 core surrounded by more specialized units for things like FP and DSP, not unlike the Cell. Assymetrical multicore would be a good way to descibe that I guess.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

 

[HansGruber]

in the future amd will try to get like 8 cores or 16 cores out while dumming down each core's performance to acheive the mutiple cores...)

Latency from CPU to CPU in multicore systems appears to be lower than in multisocket systems.

Also you should look at the prices of 8 to 16-way systems, really expensive.
But multicore is affordable, it costs a fraction of multisocket systems.

i mean 2 and 4 core is okay but if ya go any further heat and complexity will make the thing bad

With your experiences with 8-16 core CPU,s i must believe you. :lol:

<font color=red>"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein</font color=red>

 

[theholylancer]

In reply to:

in the future amd will try to get like 8 cores or 16 cores out while dumming down each core's performance to acheive the mutiple cores...)

Latency from CPU to CPU in multicore systems appears to be lower than in multisocket systems.

Also you should look at the prices of 8 to 16-way systems, really expensive.
But multicore is affordable, it costs a fraction of multisocket systems.

In reply to:

i mean 2 and 4 core is okay but if ya go any further heat and complexity will make the thing bad

With your experiences with 8-16 core CPU,s i must believe you.

"We can't solve problems by using the same kind of thinking we used when we created them."
- Albert Einstein

well this thread is all based on hypothoical ( sititutationsreally bad spelling) and that this is merly my point of view, becides usually when something gets more features or performance by going more complex then it will have more problems, think about it how much did dos offer and how much trouble it had? not many. while we are all cursing at windows, it does thing that dos never can (without some massive programming) which no one will do....... while windows DO have a lot of crap...... so that is where i get by belif that more complex=more trouble
if say the performance per core was increased then maybe only say 4 cores can acheive the same performance of 8 cores and w/o the complexity, i'm saying that AMD should develop more performance per core when they already have a 2 core or 4 core CPU developed and not go any further than that (except for server operations where i'm sure that no one will be playing games on a server so the higher performance per core is not needed but the average performance is more needed)

like your sig said eventually these computers will become obslite (quantium computer, bio-computer, etc... will replace these slicon chipped ones) because we have to rethink the way that computers work as that these slicon chips have to have a litmit or when something else is gona be easier to advance eventually, i beleive we would eventuall make the computer understand more than 0101010101010101 and that would really improve its performance , but how much trouble will it bring(terminator(movie series)?)?<P ID="edit"><FONT SIZE=-1><EM>Edited by TheHolyLancer on 06/15/05 05:57 PM.</EM></FONT></P>

 

[theholylancer]

lol we are the kind of people who put survial first as that there are manay people in china competing for youyr lunch.... that is one of the main reason we moved............... i mean look at china, people who are really good comes to the western worlds, the rest college/unversity grads are like wiping floors at restruants.......

 

[Ned_Flanders]

<A HREF="http://www.tomshardware.com/hardnews/20050615_184304.html" target="_new">What was I saying about multcore compilers / languages etc?</A> :lol:

______________
<font color=green>NED AND MOZZARTUSM - REAL (P)RESIDENTIAL CANDIDATES</font color=green>

 

[theholylancer]

hmmmm i see... maybe i should not go for the routh i'm going at aiming for game AI programming and go for multi-thread mangement programming?

 

[ofer987]

Yea, once i get a mutlicore processor myself I'll start experimenting with parallel computing.. hehehe .

 

[theholylancer]

man i think will need to learn all throse programming language first b4 i experiment with programing lol