Intel wolfdale price: USD 100 - additional info

[bonkers]

http://xtreview.com/addcomment-id-2458-view-Wolfdale-processor-additional-information.html

 

[makaka]

nice find.
AMD has serious problem there.
I can wait to see the next 45NM CPU in my case. 😀 .
overclocking should be much better
I cannot see really why Intel will use 1066 system bus.
according to the same site : the DDR 3 review , show that the memory is limited by the system bus.

 

[big_cheese]

So no 45nm cpu for desktop this year? I was waiting for it, sad.

 

[makaka]

Yes for desktop. but u can go with server cpu .
Hope that intel will accelerate its plan and release 45nm cpu this summer

 

[RichPLS]

I would not be surprised if Intel did not pull in a Yorksfield launch before Christmas... a low volume, highend enthusiast part.... just for show anyway.

Not that many would buy one

😱
I would suspect within reason, that every one offered would be snatched up...

 

[lbax]

Do you expect that Yorksfield/Harpertown will be MCP or Monolithic? I have seen claims of both.

 

[bonkers]

:twisted: and now we are all confused

http://www.techarp.com/showarticle.aspx?artno=408&pgno=0

maybe they will surprise us. i hope so

 

[Mandrake_]

Do you expect that Yorksfield/Harpertown will be MCP or Monolithic? I have seen claims of both.

Yorkfield is two Wolfdale CPUs packaged together. There will be a native quad core based on "Nehalem" next year using the CSI over the old FSB.

 

[Mandrake_]

Intel would counter Barcelona with Wolfdale-DP and Harpertown, which are the 45nm shrinks of Woodcrest and Clovertown.

 

[MU_Engineer]

The Yorkfield Intel showed in their preview was an MCM made of two dual-core Wolfdale chips. Nehalem is supposed to be the first monolithic quad-core made by Intel. It would have to be because of the IMC that's going to be in at least some of them.

 

[the_vorlon]

I thought you lose performance, not gain it with die shrinks due to transitor loss :?

Anyways, wouldn't woldale combant the phenom x2 line not barcelona? btw what is intel going to combat barcelona with in the server line?

A "dumb" die shrink yiled no change in performance.

A 3.0 ghz Conroe on 65 nanao will perform identical to a 3.0 ghz Pentym on 45 nano - ALL THINGS BEING EQUAL

All things are not equal how ever.

Penrym will ~~likely~~ show up on a 1600 mhz bus (at least on EE parts)
Penryl also has a few tweeks:

16 bit instruction shuffle
6 megs L2
SSE4 instructions.
Better branch predictions.

Clock for clock Penrym will be 3=>40% faster depending on the test you look at.

The real change is that (Prescott being the huge exception) thermals usually get a lot better with a process shrink, so Penrym should clock a ton higher than Conroe.

It's all speculation, but most folks expect a 4.00 ghz part, maybe higher.

 

[Grimmy]

8O . o O (OC to past 4.0 GHz on air?)

With or without a HS? :lol: (J/K!!)

 

[pausert20]

My guess for the Wolfdale launch speed is the following:

If Intel gets the Wolfdale out before the Phenom then they will not push the core speed but wait for AMD to show their full hand.

If Phenom launches first and Intel does a full analysis of one of then they will just launch their Wolfdale at a speed grade that gives them ~10% above whatever the fastest Phenom can do. I believe they will want to leave some GHz room on the table for Nehalem.

As for the Yorkfield Quad Core. Intel could launch them anytime after the Wolfdale launches. They could completely swamp AMD in numbers of Quad core processors. D1D could completely switch to only making 45nm parts and out produce all of Opteron parts coming out of AMD's Fab 36. It all comes down to die size and process technology.

Even if AMD could get 100% die out of each wafer that goes through their Fab 36 and Intel could be only get 60% yield out of D1D 45nm they would match if not produce more usuable die than AMD. Intel has said they plan on 3 to 4 additional 45nm Fabs all at 300mm Wafers. Intel will be outproducing AMD at launch or very soon afterwards.

Then don't forget all of the new 1333FSB Conroe parts coming out of their 3 65nm factories.

283mm for the new Quad core Opteron or ~140mm for their Phenom does not equate to Intels 110mm Wolfdale or 220mm effective Yorkfield.

It all comes down to a numbers game. AMD needed to build another Fab or 2 and not buy ATI to seriously start to compete with Intel's Fab capacity and therefor market share.

Please remember these are my thoughts on what is going to happen 2nd half of 2007 beginning 2008. I hope I did not mess up the die sizes?

 

[endyen]

As for the Yorkfield Quad Core. Intel could launch them anytime after the Wolfdale launches. They could completely swamp AMD in numbers of Quad core processors. D1D could completely switch to only making 45nm parts and out produce all of Opteron parts coming out of AMD's Fab 36. It all comes down to die size and process technology.

They had better, after all, thier market share is 4* AMD's.
Just the same, it will be a chalenge. Fab 36 also uses the 300 mm wafer. Barcelona will only be a mask change, so they should come out on mature yields.
For Intel, they will be doing a node shrink, and utilizing a double mask process. Thier initial yields will be record low numbers.
I would be very suprised if D1D could come anywhere near Fab36's production.

 

[pausert20]

As for the Yorkfield Quad Core. Intel could launch them anytime after the Wolfdale launches. They could completely swamp AMD in numbers of Quad core processors. D1D could completely switch to only making 45nm parts and out produce all of Opteron parts coming out of AMD's Fab 36. It all comes down to die size and process technology.

They had better, after all, their market share is 4* AMD's.
Just the same, it will be a challenge. Fab 36 also uses the 300 mm wafer. Barcelona will only be a mask change, so they should come out on mature yields.
For Intel, they will be doing a node shrink, and utilizing a double mask process. their initial yields will be record low numbers.
I would be very surprised if D1D could come anywhere near Fab36's production.

I think you are under estimating were Intel is at in their 45nm yields right now.

The link below is from January. They now have 5 more months to tweak their process. So I believe they should be getting higher yields now. If you assume from their chart that the middle is 50% yield back in January I will leave it to your imagination how much better their yield is now.

Intel 45nm Yield

 

[pausert20]

Jack,

D1D is probably still producing 65nm parts along with the early production of 45nm and also running test wafers at 32nm levels if I were a betting man. So that might match up to what is coming out of AMD's Fab 36. The might still be producing 90nm parts there along with all of the high speed 65nm Brisbanes.

 

[pausert20]

Yeah, I know that. I was just trying to keep it simple for coparison sake. 😀

 

[endyen]

Wow Jack, you can do better than that! Look at your own link.
Fab 11 = tel
fab 28 = under construction
fab 32 = under construction

also considering that 45 nm gives roughly 40-50% more die than 65 nm

but then again, real estate wise, there is a lot more room used by cache on the intel chips. With the lower yield @ intel, they will be lucky to get the same number of chips per wafer.
Since D1D is also the main an R&D facility, it's just not set up for output quite like Chandler is.
Hopefully fab 32 will be ready to do the brunt of the work, because D1D is just not up to the number of WSPM that intel will need, to keep up (4 times as many chips needed) with Dresden.

Let me put it this way. Dresden will need ~ 15k WSPM to keep up with demand. Intel will need 45 to 60k (depending on yields) Is it even remotely possible for D1D to do that?