Looking forward to 6 core cpu's

[Kuroimaho]

Maybe you haven't noticed the difference between the 258mm2 and 346mm2 sizes.

I wrote them so probably I did. What you seem to ignore is the quite huge price difference what you could sell these for.

~90nm2 is a pretty big difference in the semiconductor manufacturing and I haven't mentioned the yield rate which is very important.

Relative to the size not that big difference, and 5 cores could be sold as well and for more than deneb. So your economic reason is shot down.

Anyway the Deneb is a whole different story. They should have to cover the market with cheaper Deneb based CPUs.

With AMD's low margins, cheap Denebs don't sound me such a good idea maybe you haven't noticed Deneb's die size. Let's really leave this economic reason it's only getting worse.

IMHO now they can sell many Istanbul CPUs (more than Shanghai) so they don't need to hurry to sell these dies cheaper in the desktop segment.

So let's sell a slightly smaller Deneb for less than half the price, instead of making more 6 cores and sell them for twice as much. They could have better margins with the Thuban with the prices you guesstimated. Simply put they could make more money by selling these than what they make on deneb.

I sadly see that you couldn't understand my interpretation. The desktop market does not need 6 core cpu's yet. They don't have to hurry at all.

It is hard to interpret what you wrote as what you try to make it sound now.
Let me copy it here:

The other reason is that the AMD want to launch a complete new platform (Leo) and not only a CPU.

Six-core CPU + 890 series chipset + 5800 series VGA.

If that's really what you really meant I doubt anyone got what you were trying to say, this must make you even more sad.

But I agree with your new idea.
The general desktop doesn't need it, but it shouldn't stop AMD from selling it for extra profit. They didn't care much about the market's needs when they tried to sell slow native quads instead of fast native duals. Then I remember AMD was selling FX processors long time ago over a 1000$, that market must have needed those more than these Thubans.

You can't see through this whole platform question. You can buy the platform in parts you don't have to buy it in a complete system.

The platforms demonstrate the company's top products in a whole unit. It has an important marketing part.

What you seem to forget you used the platform as reason why didn't AMD launch Istanbul 9 months ago on the desktop.
The marketing reason doesn't really makes sense again, couldn't they use an older processor in the marketing, they have to delay it's launch 9 months to align it with the new motherboards ?
Strangely the last september launched 5800 gpus seem to fit into the platform. Seems to me the platform reasoning doesn't make sense either.

What kind of tweaking are you talking about? The G34 contains two dies in one package while the C32 contains only one from exactly the same die.

I am aware of that. Apparently you are sure that the IMC can't take adjustments to work more effectively in that configuration.

You're right...at last. The Istanbul is really not a dekstop only CPU. The Thuban will be an AM3 packaged Istanbul with the latest stepping (probably D2). That's all. You better to beleive it now otherwise you'll be disappointed later.

It doesn't make sense and neither your simple economic reason or the platform marketing backs it up.

If it's low clockspeed (compared to deneb) what caused they delay I would say possible.
Now with Thuban's lauch date so close to the new server processor's launch I find it more likely thuban is based on those, than something they launched last year.
The reason why I think Thuban wasn't launched last year because it didn't perform well enough against products already on the market and they were working on tweaking the 6 core die for the G34 and C32.

Your roadmap backs it up for me. If no improvements were made on the 6 core die why would they release a cut down version on the desktop ?
It should offer something what makes it superior to Deneb.

BTW it can't disappoint me, after Bulldozer's 2 year delay AMD won't make me believe their roadmaps until they start to deliver on time for a while.
Thuban is just something to have fun with until sandy and bulldozer, more interesting than socket 1156 regardless of it's performance, if Deneb worth it more that's what I am going to keep.

[Oliverda]

I wrote them so probably I did. What you seem to ignore is the quite huge price difference what you could sell these for.



Relative to the size not that big difference, and 5 cores could be sold as well and for more than deneb. So your economic reason is shot down.



With AMD's low margins, cheap Denebs don't sound me such a good idea maybe you haven't noticed Deneb's die size. Let's really leave this economic reason it's only getting worse.



So let's sell a slightly smaller Deneb for less than half the price, instead of making more 6 cores and sell them for twice as much. They could have better margins with the Thuban with the prices you guesstimated. Simply put they could make more money by selling these than what they make on deneb.



It is hard to interpret what you wrote as what you try to make it sound now.
Let me copy it here:



If that's really what you really meant I doubt anyone got what you were trying to say, this must make you even more sad.

But I agree with your new idea.
The general desktop doesn't need it, but it shouldn't stop AMD from selling it for extra profit. They didn't care much about the market's needs when they tried to sell slow native quads instead of fast native duals. Then I remember AMD was selling FX processors long time ago over a 1000$, that market must have needed those more than these Thubans.



What you seem to forget you used the platform as reason why didn't AMD launch Istanbul 9 months ago on the desktop.
The marketing reason doesn't really makes sense again, couldn't they use an older processor in the marketing, they have to delay it's launch 9 months to align it with the new motherboards ?
Strangely the last september launched 5800 gpus seem to fit into the platform. Seems to me the platform reasoning doesn't make sense either.



I am aware of that. Apparently you are sure that the IMC can't take adjustments to work more effectively in that configuration.



It doesn't make sense and neither your simple economic reason or the platform marketing backs it up.

If it's low clockspeed (compared to deneb) what caused they delay I would say possible.
Now with Thuban's lauch date so close to the new server processor's launch I find it more likely thuban is based on those, than something they launched last year.
The reason why I think Thuban wasn't launched last year because it didn't perform well enough against products already on the market and they were working on tweaking the 6 core die for the G34 and C32.

Your roadmap backs it up for me. If no improvements were made on the 6 core die why would they release a cut down version on the desktop ?
It should offer something what makes it superior to Deneb.

BTW it can't disappoint me, after Bulldozer's 2 year delay AMD won't make me believe their roadmaps until they start to deliver on time for a while.
Thuban is just something to have fun with until sandy and bulldozer, more interesting than socket 1156 regardless of it's performance, if Deneb worth it more that's what I am going to keep.

Sorry but I got tired about your ideas. The time will confirm everything.