AMD Working on MCM

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
At the anylyst day today, AMD said they are working on an MCM approach for octal core Barcelona.

Talk about eating your own words.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
If this turns out to be true then I for one applaud the managerial courage it takes to eat humble pie and set aside the ego rooted in their past (disasterous) decisions.

Intel did, and we got Core2Duo out of it. Thankfully it sounds like AMD is capable of learning from their mistakes.
 

jones377

Senior member
May 2, 2004
452
47
91
Wouldn't it be great if Nehalem (Beckton?) comes out in a native octo-core version while AMD does the MCM approach? I'd like to see the PR departments (and their msgboard lackeys) of both companies spin that!
 

SexyK

Golden Member
Jul 30, 2001
1,343
4
76
There will be native octo-core Nehalem, now it's just a matter of whether AMD really bites the bullet and goes MCM.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Originally posted by: jones377
Wouldn't it be great if Nehalem (Beckton?) comes out in a native octo-core version while AMD does the MCM approach? I'd like to see the PR departments (and their msgboard lackeys) of both companies spin that!

Well I definitely remember how AMD was quite enthusiastic about screwing the customer base with exhorbinant prices for their dual-core chips until Intel came out with Core2Duo and offered some dual-core competition...

So quite honestly I don't care how AMD gets an 8-core chip (heck, I'd be happy just with AMD able to compete in the 4-core segment in 2008) on the market so long as they do something so Intel doesn't treat the consumer to a 2006-redux in 2008 as AMD did.
 

Extelleron

Diamond Member
Dec 26, 2005
3,127
0
71
Originally posted by: SexyK
There will be native octo-core Nehalem, now it's just a matter of whether AMD really bites the bullet and goes MCM.

Nehalem will definately not be native octal-core. In fact, from what I've heard Intel won't even be going 8-core with Nehalem due to thermal/power limitations, they will wait until the 32nm refresh. A native 8-core processor would be way, way too big to even consider. A quad-core 45nm Penryn is 214mm^2, a 8-core Nehalem would surely be 500-600mm^2 in size.
 

postmortemIA

Diamond Member
Jul 11, 2006
7,721
40
91
who cares about x-cores when single core to single core performance vs C2D is abysmal.. .there are widely used programs with 8 foreground threads??
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Originally posted by: Idontcare
If this turns out to be true then I for one applaud the managerial courage it takes to eat humble pie and set aside the ego rooted in their past (disasterous) decisions.

Intel did, and we got Core2Duo out of it. Thankfully it sounds like AMD is capable of learning from their mistakes.

Actually we got core2quad and the intel D from it... the core2 duo is a single die.


Anyways, good for AMD.. I hope this will help them get their quad core out before intel gets their MCM octal core out...
 

CTho9305

Elite Member
Jul 26, 2000
9,214
1
81
Originally posted by: postmortemIA
who cares about x-cores when single core to single core performance vs C2D is abysmal.. .there are widely used programs with 8 foreground threads??

Servers, HPC, supercomputers...they care.
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Well... the only thing I use that takes advantage of multiple cores is winrar... boinc was a huge one but I stopped using it because of the cost of electricity.

And the most annoying slowdown I have is level loading in neverwinter nights 2 only uses one core (it superfluously decompresses it and then discards the previous data every time you change a "zone" or reload a saved game... regardless of how much ram you have).

The only game I know of that uses/needs 4 cores is crysis, and I am already done with it.

I really want a faster dual core right now. But my only option is to replace the mobo and get a c2d.
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Originally posted by: taltamir
I really want a faster dual core right now. But my only option is to replace the mobo and get a c2d.

Isn't a 3.2 Ghz X2 fast enough for gaming? What video card do you have?
 

Nathelion

Senior member
Jan 30, 2006
697
1
0
I remember AMD talking about how the "inter-core switch" in Barcelona being set up for octal core already... Although I doubt that will matter until 32nm, even if it's true
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
I am returning the 6400+... its not stable even with the zalman 9500...
I am back with my 2.0ghz X2
 

ViRGE

Elite Member, Moderator Emeritus
Oct 9, 1999
31,516
167
106
It'll be interesting to see how they pull it off. The FSB made Intel's MCM design fairly easy to pull off, HT and the IMC on the other hand can make it rather difficult for AMD.
 

Nemesis 1

Lifer
Dec 30, 2006
11,366
2
0
I am not going to get road maps but I believe Nehelam @ 45nm will be 6core at the highest I think its called Beckington. . Nehalem C 8x core @ 32nm in 09. Geshner 2010.

These are all native! Thanks to Quick Path.


I here a lot of guys saying so who needs 8 cores? I have a Tendency to agree. Unless theirs something we don't know. But thats not the point. With all the news thats out about future cores. I look at Nethalem and I just have to smile. You might ask whys that?

If their isn't some unknown leap appearing in near future and Nehalem as a 2core processor. This is the perfect processor. This is why I believe that.

1). IF single thread performance is 20% better than Penryn @ = CLOCKS. ALSO if intel can deliver the 4core Nehalem @ 4ghz as they say. and I have zero reason to doubtt this.
Under these circumstances 2core Nehalem would be a dream chip on the desk top.
2). Nehalem 2core could handle 4 threads If needed . But if not so what. The dual core version should O/C like Crazy and be cheap to . The Perfect Gamer chip. The perfect desk top Chip that the competition can't touch.

Now people smarter than me are creating these chips! So unless their is something going on we aren't aware of . 2 CORE NEHALEM IS A NO BRAINER ! On the desktop and forsure on the note book.

AS I said people smarter than I are behind this multi-core push. These same people wouldn't be stupid enough to put out a 2core chip that would kill there other higher margin products. Unless there is an unknown at play here. I am referring to desktop and notebook only here. I haven't seen a dual core on Nehalems roadmaps but being modular design Intel can add and subtract cores at will. SO what ever Intel has to offer at the very lowest level as far as core count on nehalem . Should be the. DO all! Be all! On the desktop . and a killer on the notebook. All this depends on single thread performance. IF SINGLE thread performance is 20%>penryn .

I have thought about this . and thought about it. Their is something we don't know.

IF Intel sells Nehalem as a 2core processor. IT would wipe out their own higher end.
It just doesn't make since. I would perfer a dual core 4 thread processor that O/C like crazy and is more efficient and could do everthing I desire as far as work load. Is a lot cheaper if I were the average pc buyer. O/Cing aside.

So to me things just aren't adding up. If using ones own logic says 2cores 4 threads cheaper platform. Great single thread performance. I would say other than servers and a few extreme users . Intels 4 core desk top just wouldn't be a big seller.
 

zsdersw

Lifer
Oct 29, 2003
10,505
2
0
It's Gesher (not Geshner) and Beckton (not Beckington).

As for the complaints about not needing 4 or 8 or more cores, aside from the usage scenarios of people other than yourself there's another thing to consider here: hardware development always drives software development and software development sometimes drives hardware development. The two go hand in hand.

Many will point to the growing number of computing tasks that can be done adequately by lower-end hardware as a sign that consumers in general don't need the latest-and-greatest hardware and, thus, shun the multi-core approach of both Intel and AMD. Progress must be made, however, on both the hardware and software fronts in order for the industry to thrive and, in turn, for that industry to remain the subject of our intense interest and devotion as enthusiasts. If technical limitations steer CPU hardware development down the multi-core path, what's the difference? Eventually the software will catch up, just as it will catch up to any specific core-level improvements.

My opinion is that the multi-core approach was a good one, as it more or less forces improvements in not only the number of cores that work together, but also in how each of those cores function. Prior to the multi-core era, improvements were generally not all that impressive and still took time for the software to be re-worked in order for improvements to be widely considered impressive. Once software catches up to multi-core (which it already has or is beginning to in some areas) this should yield impressive gains in performance.

Never look to past or present needs as a justification for hindering future needs.
 

SexyK

Golden Member
Jul 30, 2001
1,343
4
76
Originally posted by: Extelleron
Originally posted by: SexyK
There will be native octo-core Nehalem, now it's just a matter of whether AMD really bites the bullet and goes MCM.

Nehalem will definately not be native octal-core. In fact, from what I've heard Intel won't even be going 8-core with Nehalem due to thermal/power limitations, they will wait until the 32nm refresh. A native 8-core processor would be way, way too big to even consider. A quad-core 45nm Penryn is 214mm^2, a 8-core Nehalem would surely be 500-600mm^2 in size.

Surprisingly, Intel gave away quite a few details about Nehalem. Although Nehalem is still based on the 4-issue Core architecture, it takes "multithreading" to a whole new level. First of all, Nehalem can contain up to eight cores per die. Combined with 2-way Simultaneous Multi-Threading (SMT or Hyper-Threading), you'll have the ability to execute up to 16 threads on one chip!

Link
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
Originally posted by: taltamir
Originally posted by: Idontcare
If this turns out to be true then I for one applaud the managerial courage it takes to eat humble pie and set aside the ego rooted in their past (disasterous) decisions.

Intel did, and we got Core2Duo out of it. Thankfully it sounds like AMD is capable of learning from their mistakes.

Actually we got core2quad and the intel D from it... the core2 duo is a single die.


Anyways, good for AMD.. I hope this will help them get their quad core out before intel gets their MCM octal core out...

Actually I was referring to Intel's abandonment of the Netburst approach, not the adoption of MCM integration. But your point is still valid.

My point was that Intel pushed and pushed headstrong down one path, but eventually ate humble pie, shot the horse with the broken leg, and jumped full-steam onto the Core2 bandwagon.

AMD needs to do something like this, at a management level, and stop blaming the under-resourced design layout and process development engineers and lay blame squarely on the shoulders of the corporate decision makers...but if they did that then I suppose Hector Ruinz would have a hard(er) time justifying a pay increase while shareholders eat cake.
 

Extelleron

Diamond Member
Dec 26, 2005
3,127
0
71
Originally posted by: SexyK
Originally posted by: Extelleron
Originally posted by: SexyK
There will be native octo-core Nehalem, now it's just a matter of whether AMD really bites the bullet and goes MCM.

Nehalem will definately not be native octal-core. In fact, from what I've heard Intel won't even be going 8-core with Nehalem due to thermal/power limitations, they will wait until the 32nm refresh. A native 8-core processor would be way, way too big to even consider. A quad-core 45nm Penryn is 214mm^2, a 8-core Nehalem would surely be 500-600mm^2 in size.

Surprisingly, Intel gave away quite a few details about Nehalem. Although Nehalem is still based on the 4-issue Core architecture, it takes "multithreading" to a whole new level. First of all, Nehalem can contain up to eight cores per die. Combined with 2-way Simultaneous Multi-Threading (SMT or Hyper-Threading), you'll have the ability to execute up to 16 threads on one chip!

Link

http://pc.watch.impress.co.jp/...1127/kaigai402_01l.gif

Those are the three Nehalem-based cores that will be launching in 2008, AFAIK.

I know Intel demonstrated Nehalem with 16-threads, but I don't think this will be released until 32nm. It's really just common sense - how would Intel build a 8-core CPU that is remotely affordable when 4-cores on an older architecture are already 214mm^2? If you need proof that a quad-core Nehalem is already too big, just look at the amount of cache it will have - 8MB between 4 cores. If Intel had plenty of space to work with, they wouldn't make a new chip with less cache (even if it is less important because of the IMC.)

 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
Originally posted by: Idontcare
Originally posted by: taltamir
Originally posted by: Idontcare
If this turns out to be true then I for one applaud the managerial courage it takes to eat humble pie and set aside the ego rooted in their past (disasterous) decisions.

Intel did, and we got Core2Duo out of it. Thankfully it sounds like AMD is capable of learning from their mistakes.

Actually we got core2quad and the intel D from it... the core2 duo is a single die.


Anyways, good for AMD.. I hope this will help them get their quad core out before intel gets their MCM octal core out...

Actually I was referring to Intel's abandonment of the Netburst approach, not the adoption of MCM integration. But your point is still valid.

My point was that Intel pushed and pushed headstrong down one path, but eventually ate humble pie, shot the horse with the broken leg, and jumped full-steam onto the Core2 bandwagon.

AMD needs to do something like this, at a management level, and stop blaming the under-resourced design layout and process development engineers and lay blame squarely on the shoulders of the corporate decision makers...but if they did that then I suppose Hector Ruinz would have a hard(er) time justifying a pay increase while shareholders eat cake.

making him (so i hear) the highest payed CEO of a computer technology fortune 500 company... at 1.2million dollars...