MS Summit shows off Barcelona Wafers!!

[BaronMatrix]

The Inq has pics of 65nm quad core wafers presented at the MS Global summit.

I guess they are on track to get samples out in Dec. If it's the beast the specs tend to imply, beast is the right word to describe it.

 

[evilr00t]

Oh noes, the INQ. Must be true. :roll: :roll: :roll: :roll: :roll: :roll:

 

[knightrous]

Pics or BAN! 8)

 

[thematrixhazuneo]

LINK

 

[knightrous]

"The Inquirer: News, reviews, facts and friction"

Shouldn't that be

"The Inquirer: News, baised reviews, fanbois and fiction" :roll:

 

[corvetteguy]

The Inq has pics of 65nm quad core wafers presented at the MS Global summit.

I guess they are on track to get samples out in Dec. If it's the beast the specs tend to imply, beast is the right word to describe it.

I know, did you see the size of those Friggin die??

I did a quick count, kinda tedious -- roughly 200 die on the wafer, this would be a die size of roughly 350 mm^2 roughly.... AMD is going to lose money on this one.

Are you sure? Maybe the cores look like more than they are because they are new quads. Wait, are these quad or dual. I forget what barcelona is lol.

 

[dasickninja]

Nice. Can't wait to see it go up against Intel's Conroe and Kentsfield.
Yay competition

 

[NMDante]

Those die are huge. Are those supposedly server CPUs or desktops? I would hate to see the size of the packaging on those things. They look as big as the Montecito die.

If that's a desktop CPU, and they are going to cram that into an AM2 socket package, that would be interesting, indeed.

 

[NMDante]

Those die are huge. Are those supposedly server CPUs or desktops? I would hate to see the size of the packaging on those things. They look as big as the Montecito die.

If that's a desktop CPU, and they are going to cram that into an AM2 socket package, that would be interesting, indeed.

They are MASSIVE aren't they --- this will never fly, especially at 65 nm.

When will AMD quit showing me pictures of printed wafers and actually produce something that boots windows?

Agreed.

If this is a desktop processor, it might mean a new socket design, which would piss off a lot of AMD folks. I don't really see how they can do the connections on an AM2 size package, but if they do, more power to them.

It looks like a server processor, imo. Not sure, but just the size makes it seem like a server processor.

 

[Heyyou27]

9-inch? It's about time AMD showed something; I thought they were hiding.

 

[BaronMatrix]

The Inq has pics of 65nm quad core wafers presented at the MS Global summit.

I guess they are on track to get samples out in Dec. If it's the beast the specs tend to imply, beast is the right word to describe it.

I know, did you see the size of those Friggin die??

I did a quick count, kinda tedious -- roughly 200 die on the wafer, this would be a die size of roughly 350 mm^2 roughly.... AMD is going to lose money on this one.

Are you sure? Maybe the cores look like more than they are because they are new quads. Wait, are these quad or dual. I forget what barcelona is lol.

Eh, I simply counted down 1/2 the wafer and multiplied by 2, then used Pi*R^2 to estimate the die size. These die are > 300 mm^2, this will not be economical unless they only make it for server and, likely the new 4x4 platform. Or then again, they may try to push it into the mainstream as Intel will have mainstream parts in Q1 07.

We shall see if the 'glued on' approach is really better overall from a yield prespective.

I have been chastised for saying Barcelona will be SERVER ONLY. BullDozer is for mobile and desktop. I also roughly counted but only to see if I could still count.
I guess those fringe parts could be counted as X2 with L3. Wow that sounds cool.

FX will be first to be quad on the desktop but that gives a good upgrade for my SOON TO COME FX+2 setup.

WHy is it so hard to believe that AMD will release when they said? They even got ahead and are sampling soon.

Beast, beast, beast I say, AKA Barcelona (usual forward-thinking disclaimer).

 

[BaronMatrix]

Oh noes, the INQ. Must be true. :roll: :roll: :roll: :roll: :roll: :roll:

I guess we can expect a lwasuit from MS soon as they are quoted as being the forum - hahaha - for the presentation. Bu tthen I guess AMD will come out and say they released no such wafer.

HMMMMM

 

[BaronMatrix]

Those die are huge. Are those supposedly server CPUs or desktops? I would hate to see the size of the packaging on those things. They look as big as the Montecito die.

If that's a desktop CPU, and they are going to cram that into an AM2 socket package, that would be interesting, indeed.

So basically you did a complex algorithm to determine actual size vs. photo size. I gess there are some specs as to the size. I would assume they are not much different than dual core at 90nm. I guess your buddy Jack can tell you the difference in size beween 90nm and 65nm.


I believe it's about half the size per transistor.(AMD buyer disclaimer)

 

[BaronMatrix]

9-inch? It's about time AMD showed something; I thought they were hiding.

I guess you'd have to feel me up to find out and I'm not into that .

 

[BaronMatrix]

Those die are huge. Are those supposedly server CPUs or desktops? I would hate to see the size of the packaging on those things. They look as big as the Montecito die.

If that's a desktop CPU, and they are going to cram that into an AM2 socket package, that would be interesting, indeed.

They are MASSIVE aren't they --- this will never fly, especially at 65 nm.

When will AMD quit showing me pictures of printed wafers and actually produce something that boots windows?

Agreed.

If this is a desktop processor, it might mean a new socket design, which would piss off a lot of AMD folks. I don't really see how they can do the connections on an AM2 size package, but if they do, more power to them.

It looks like a server processor, imo. Not sure, but just the size makes it seem like a server processor.



Ummmm, if you look at the distance betwen the the pins and the cores you will see that the core die can be larger in the same package. How do you think 65nm will fit int a 90nm AM2 socket - not that the socket determines the transistor size?

 

[BaronMatrix]

Nice. Can't wait to see it go up against Intel's Conroe and Kentsfield.
Yay competition

Yeah except that it will be going against X2 and Opteron.

 

[BaronMatrix]

The Inq has pics of 65nm quad core wafers presented at the MS Global summit.

I guess they are on track to get samples out in Dec. If it's the beast the specs tend to imply, beast is the right word to describe it.

Wow, I forgot to post the link. I guess you all have to click on some things too.

 

[knightrous]

Stop triple posting and learn to use the edit button... Your a dev and a post whore? :roll:

 

[evilr00t]

It's "Quad" everything... "Quad" posting example above!

 

[BaronMatrix]

Stop triple posting and learn to use the edit button... Your a dev and a post whore? :roll:

Those are 3 DIFFERENT POSTS SHROOM BOY.

 

[BaronMatrix]

It's "Quad" everything... "Quad" posting example above!

X people posted, X people get responses.

Good night, it 11:48 in NYC.

 

[Action_Man]

So much for not getting excited over CPU news, moron. Now get back to cleaning those toilets before they fire you.

 

[MU_Engineer]

It will be a big die, but the area of silicon of the two Conroe dies in a Kentsfield is about that number too. AMD will have a little lower yield putting all four on one die than the MCM Kentsfield, but the sheer amount of silicon is the same. This tells me we'll have to wait until 45nm to see more than 4 cores and 8MB cache onboard a CPU. Even Intel, admitted long-time lover of huge caches and MCMs will likely not put more than 8MB cache on a quad-core desktop die. Maybe the Xeon MPs will be as big as the Tulsas (well over 2cm x 2cm die!) with 8MB L3s in addition to the 8MB L2 but we'll not see that nor a MCM 8-core chip from them until the 45nm shrink.

And as far as a new socket being needed, that is not supposed to happen quite yet. This will originally debut as an Opteron (x4xx number is my guess, such as a 2.0 GHz 2412 DP Opteron) and the current Opteron Socket F has a ton of pins, enough bandwidth from dual-channel DDR2, and the socket just got released a little bit ago. It *must* have been made with things like this in mind as enterprise gear changes sockets far less often that desktop stuff. 940 lasted for >3 years- pretty long compared to the few months that Socket F has been out. On the desktop, AM2 is also a brand-new socket with enough bandwidth from dual-channel DDR2 to feed four cores and it's also a very new socket. AM3 is the successor to AM2 and it will have DDR3 support. DDR3 is not out there yet, so AMD won't need to design the new socket to support it yet. I'd think that DDR3 and AM3 will come along about the time that AMD shrinks their dies to 45nm and makes an 8-core CPU. That's a little while off yet.

 

[knightrous]

You can put multiple reply quotes in one post :roll:

I guess it's too hard for you, being a dev and all. 8)

SHROOM BOY

Gotta love those generic childish taunts :tongue:

 

[endyen]

Once you get past the cutting, those little squares look to be about 14-15mm/side. That would put them at ~ the same size as the dual core A64s. (208mm^2)