MCM K8 quad?

nonameo

Diamond Member
Mar 13, 2006
5,902
2
76
Seriously now...

Good idea or Bad idea?

edit: since they really don't have anything to compete with intel up top... and it wouldn't hurt them too much to do so. I hate to say it, but phenom isn't going to cut it for the next half a year or so.

It could be called "affordaquad"
 

VirtualLarry

No Lifer
Aug 25, 2001
56,566
10,181
126
I think that it would be a dynamite idea. Perhaps AMD could compete with Intel in terms of affordable quad-core pricing. I would rather have a 3.0Ghz K8-quad-core rather than a 2.3Ghz K10.
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
AMD should have done that last year... I really don't know what their problem is. Too much arrogance not enough engineering quality.
 

manowar821

Diamond Member
Mar 1, 2007
6,063
0
0
I think that since they've already moved their fabs over to the K10 platform mostly, that if they can get the stupid-ass problems fixed, the K10 will be the more logical (and profitable) route.

They can't back-track, now. They need to move forward with their investment.
 

KingstonU

Golden Member
Dec 26, 2006
1,405
16
81
But K10 has proved to be a failure, even if they fix the TLB errata they are still under performing Penryn and even Conroe. Even when Shanghai comes out in H2 2008 it is only a 45nm die shrink of K10.

And then Intel will have Nehalem which has integrated memory controller and is a monolithic Quad and so will remove any and all "advantages" AMD's designs have had over Intel.

AMD doesn't have another chance to catch up until Bulldozer in 2009 replaces K10, hopefully that won't be a failure as well. *sigh*
 

Comdrpopnfresh

Golden Member
Jul 25, 2006
1,202
2
81
Ain't gonna happen. They won't be able to go anywhere close to 3.0ghz. They're just revealing 45watt tdp k8's now. They probably wouldn't use those, so the tdp would be a big drawback. Plus- they're not gonna drop money on designing crossbars and ht links into the processors for those ends. And if they did, it would be to admit that k10 quads aren't necessary and intel has the right idea of double-duals for quad, vs phenom integrated quad

::EDIT::
As intel and amd architectures become more similar, we're going to see a ramping of speed in a ghz-wars style again. If shanghai scales well in tdp performance, then amd quads will be able to clock fairly high well. I feel with more bios revisions, driver updates, and bug fixes, we'll see phenom competing fairly well with penryn- AMD just needs to perfect their design and manufacturing process.
Amd will still have an advantage when nehalem comes out- more experience with integrated memory controllers. Which will strike a cord with AMD's largest market- business and server-side customers. They like to see stability and mature platforms more than shiny new promises like enthusiasts chase
 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
You can't do MCM with K8, unless you give each dual-core it's own NB chipset, along with it's own RAM. The IMC isn't able to share either a NB or any sticks of RAM.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Originally posted by: manowar821
I think that since they've already moved their fabs over to the K10 platform mostly, that if they can get the stupid-ass problems fixed, the K10 will be the more logical (and profitable) route.

They can't back-track, now. They need to move forward with their investment.

The amount of fab capacity producing K10 as a percentage is in the single digits.

 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Originally posted by: Phynaz
Originally posted by: manowar821
I think that since they've already moved their fabs over to the K10 platform mostly, that if they can get the stupid-ass problems fixed, the K10 will be the more logical (and profitable) route.

They can't back-track, now. They need to move forward with their investment.

The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.



 

KingstonU

Golden Member
Dec 26, 2006
1,405
16
81
Originally posted by: Comdrpopnfresh

it would be to admit that k10 quads aren't necessary and intel has the right idea of double-duals for quad, vs phenom integrated quad

They already admitted to this. Don't have the link but a little while back a person high up at AMD officially said that they made a bad judgment call as monolithic has shown to have little to no benefits over MCM and that if they could go back in time they would have gone the MCM route.

I think Daily Tech wrote the article on it. Can anyone bring it up?



 

coldpower27

Golden Member
Jul 18, 2004
1,676
0
76
Originally posted by: KingstonU
Originally posted by: Comdrpopnfresh

it would be to admit that k10 quads aren't necessary and intel has the right idea of double-duals for quad, vs phenom integrated quad

They already admitted to this. Don't have the link but a little while back a person high up at AMD officially said that they made a bad judgment call as monolithic has shown to have little to no benefits over MCM and that if they could go back in time they would have gone the MCM route.

I think Daily Tech wrote the article on it. Can anyone bring it up?

http://www.theregister.co.uk/2...7/amd_rivas_barcelona/

?If I could do something different, I wish we would have immediately done a MCM - two dual cores and call it a quad-core,? said Mario Rivas, an EVP at AMD, during a recent interview in Austin, ?because, I guess, the market sucks it up.?
 

KingstonU

Golden Member
Dec 26, 2006
1,405
16
81
Thanks coldpower

Also, as far as comparing features, AMD has never (or at least for the most part) been able to compete in terms of process node or Ghz, they instead competed in terms of features like integrated memory controller.
 

coldpower27

Golden Member
Jul 18, 2004
1,676
0
76
Originally posted by: KingstonU
Thanks coldpower

Also, as far as comparing features, AMD has never (or at least for the most part) been able to compete in terms of process node or Ghz, they instead competed in terms of features like integrated memory controller.

AMD has historically never been able to compete in terms of market power as they lack Intel's exemplary reputation.

They typically could only compete on price/performance, it wasn't till the Athlon 64 and Athlon 64x2 where that started to change, but things have gone back to "the norm" with the release of Core 2, and AMD competes on price only once again.

At least this time around they can still maintain the 4P+ at somewhat decent levels, but the current Tigerton with it's sheer clockspeed is attacking that sector as well, and things will only get worse when Nehalem does arrive.
 

SexyK

Golden Member
Jul 30, 2001
1,343
4
76
Originally posted by: coldpower27
Originally posted by: KingstonU
Originally posted by: Comdrpopnfresh

it would be to admit that k10 quads aren't necessary and intel has the right idea of double-duals for quad, vs phenom integrated quad

They already admitted to this. Don't have the link but a little while back a person high up at AMD officially said that they made a bad judgment call as monolithic has shown to have little to no benefits over MCM and that if they could go back in time they would have gone the MCM route.

I think Daily Tech wrote the article on it. Can anyone bring it up?

http://www.theregister.co.uk/2...7/amd_rivas_barcelona/

?If I could do something different, I wish we would have immediately done a MCM - two dual cores and call it a quad-core,? said Mario Rivas, an EVP at AMD, during a recent interview in Austin, ?because, I guess, the market sucks it up.?

LOL. That quote cracks me up. AMD's EVP still trying to spread FUD that C2Q isn't a "true" quad core because it's not monolithic and anyone who purchases one is an idiot who just believes some giant Intel lie... lol. Ever since day 1 AMD and AMD fanboys have been trying to discredit Intel's approach to multicore and even to this day (in which Intel's design choices have been proven to be the right ones) they still try to act like Intel is somehow cheating people by using multiple dies in one package. Note to Rivas: take a look at the performance numbers then come back and say your monolithic design is superior.
 

hans007

Lifer
Feb 1, 2000
20,212
18
81
Originally posted by: myocardia
You can't do MCM with K8, unless you give each dual-core it's own NB chipset, along with it's own RAM. The IMC isn't able to share either a NB or any sticks of RAM.

well...


its sort of possible.

you can do a dual socket opteron board, with only having ram on one of the CPUs.

the other cpu can use the hypertransport link between the 2 cpus to access memory through the one that is connected to memory.

this is obviously not an ideal arrangement, but there are opteron boards that do this (only have 1 bank of memory slots)
 

taltamir

Lifer
Mar 21, 2004
13,576
6
76
in that intel is about to release a monolith "true" quad core with an integrated memory controller... all the so called advantages of AMD... And based on how things are looking intel will have it ready before the phenom is fixed... so not only did intel get quad cores to the market over a year earlier with the MCM design. But they are able to get it in the monolithic phase quicker as well... and I bet intel will then give as MCM octa core processors...
 

Viditor

Diamond Member
Oct 25, 1999
3,290
0
0
Originally posted by: taltamir
in that intel is about to release a monolith "true" quad core with an integrated memory controller... all the so called advantages of AMD... And based on how things are looking intel will have it ready before the phenom is fixed... so not only did intel get quad cores to the market over a year earlier with the MCM design. But they are able to get it in the monolithic phase quicker as well... and I bet intel will then give as MCM octa core processors...

Ummm...you're talking about Nehalem which won't be in stores till Q1 09 (though probably shipping in Q4 08). How is that "quicker"?
Also, AMD's Bulldozer will be a native 8-16 core design...due in early 2009 as well.
I doubt that Intel will release an MCM based on Nehalem...they have no reason to, and dealing with an on-die NB and MCM is quite problematical.
 

Viditor

Diamond Member
Oct 25, 1999
3,290
0
0
Originally posted by: Phynaz


The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.

Link please...or maybe a good rationale for that?
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Originally posted by: Viditor
Originally posted by: Phynaz


The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.

Link please...or maybe a good rationale for that?

AMD's own words about producing 10's of thousands of K10 per month.

What's AMD's total chip capacity? A million chips a month? Two?

BTW, why did AMD dilute their shareholder value by 9% again?



 

myocardia

Diamond Member
Jun 21, 2003
9,291
30
91
Originally posted by: hans007
its sort of possible.

you can do a dual socket opteron board, with only having ram on one of the CPUs.

the other cpu can use the hypertransport link between the 2 cpus to access memory through the one that is connected to memory.

this is obviously not an ideal arrangement, but there are opteron boards that do this (only have 1 bank of memory slots)

Oh yeah, I forgot about that cheap solution. Of course, that would obviously perform worse than a dual NB chipset, dual RAM slot motherboard, like all of the good 2P Opteron boards use.
 

hans007

Lifer
Feb 1, 2000
20,212
18
81
Originally posted by: Phynaz
Originally posted by: Viditor
Originally posted by: Phynaz


The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.

Link please...or maybe a good rationale for that?

AMD's own words about producing 10's of thousands of K10 per month.

What's AMD's total chip capacity? A million chips a month? Two?

BTW, why did AMD dilute their shareholder value by 9% again?

they diluted it to well finance continuing operations....


the UAE group that bought it has already lost 25-30% of their investnent...

honestly if AMD is sitting in the 7-8 range and they manage to break even one of these quarters it might not be a completely horrible investment at say $7 a share.
 

Phynaz

Lifer
Mar 13, 2006
10,140
819
126
Originally posted by: hans007
Originally posted by: Phynaz
Originally posted by: Viditor
Originally posted by: Phynaz


The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.

Link please...or maybe a good rationale for that?

AMD's own words about producing 10's of thousands of K10 per month.

What's AMD's total chip capacity? A million chips a month? Two?

BTW, why did AMD dilute their shareholder value by 9% again?

they diluted it to well finance continuing operations....


the UAE group that bought it has already lost 25-30% of their investnent...

honestly if AMD is sitting in the 7-8 range and they manage to break even one of these quarters it might not be a completely horrible investment at say $7 a share.

We all know why AMD did what they did, I was just poking at Viditor a little. He claims AMD doesn't have a cash problem, he says they have five years of cash on hand.

 

Viditor

Diamond Member
Oct 25, 1999
3,290
0
0
Originally posted by: Phynaz
Originally posted by: Viditor
Originally posted by: Phynaz


The amount of fab capacity producing K10 as a percentage of overall capacity is in the single digits.

Link please...or maybe a good rationale for that?

AMD's own words about producing 10's of thousands of K10 per month.

What's AMD's total chip capacity? A million chips a month? Two?

BTW, why did AMD dilute their shareholder value by 9% again?

As usual you're spinning the numbers again...the dilutions was 8.2%.
I would have thought that it was obvious why they did it...to secure cash during a credit crunch period. It's because of that deal that they have plenty of cash of course...

There's no such thing as "chip capacity" because different chips are different sizes.
It's called wafer capacity or WSPM.
Could you show me where AMD says they are only producing 10's of thousands of K10s?