Fudzilla: Intel 45nm Havendale MCM replaced by 32nm+45nm MCM

Monday 09th February 2009, 02:15:00 PM, written by Arun

Following similar rumours from other sources, Fudo now claims that Intel is replacing the Havendale (the 45nm CPU+GPU MCM) by Clarkdale, which keeps the same IGP but replaces the CPU by a 32nm dual-core chip. It is slated to still be on time for the 2010 Notebook refresh cycle in Q1 2010.

This seems to imply that the GPU part of Havendale was what caused the delays (from Q2 2009 to Q1 2010 in practice), and so Intel decided to just skip that platform and use their 32nm silicon which seems to be ahead of schedule. This appears very credible to us. Especially important in this context is Intel's decision to increase capital investment for 32nm in 2009 despite the recession, and so this chip will probably be their highest-volume part early in the life of the process.

The IGP remains a derivative of Intel's G35/G45 architecture, and its performance can be estimated from two contradictory graphs (one from here & two from here). The first graph implies a performance improvement of ~2.7x over G45, while the latter implies one of 1.7x. Pretty big difference, but either way it's an incremental improvement - and it'd actually remain slower than NVIDIA's MCP79 or AMD's RS780 for most purposes.

It would be impossible for Intel to release Clarkdale without the IGP part (unsurprisingly, given their desire to keep that market for themselves) as the memory controller is actually on that chip too; the CPU communicates with it via QuickPath, just like the current high-end Nehalem communicates with the X58 northbridge. It would need its own socket, and rumours indicated a long time ago that this would actually be the plan - but presumably because SiS decided to give up on the IGP market and relations with NVIDIA soured, Intel decided to cancel that part of its line-up.


Discuss on the forums

Tagging

intel ± havendale, clarkdale, igp, mcm, nvidia, ion


Latest Thread Comments (6 total)
Posted by AnarchX on Tuesday, 10-Feb-09 22:00:11 UTC
Image: http://img264.imageshack.us/img264/2834/clarkdaleigplolly3.jpg
http://download.intel.com/pressroom/kits/32nm/westmere/32nm_WSM_Press.pdf

If you consider the scaling between booth packages is equal (penryn could be smaller - 775 vs 1156 pins), Ironlake (the IGP-die) should be >=100mm² @ 45nm. :shock:

Even with Intels architecture should be a significant performance bump over GMA X4500 possible, but nicer would be indeed some SGX-tech in this space.:lol:

Posted by Davros on Tuesday, 10-Feb-09 22:37:56 UTC
they are making a single chip using 2 process's ?

Posted by 3dilettante on Tuesday, 10-Feb-09 22:41:10 UTC
It's two chips on the same package.

Posted by AlStrong on Tuesday, 10-Feb-09 22:44:00 UTC
Quoting Davros
they are making a single chip using 2 process's ?
They are taking chips from two differently processed wafers and sticking them onto the same package.edit: beaten. :p

Posted by Davros on Tuesday, 10-Feb-09 23:19:02 UTC
tnx everyone apart from AlStrong who was too slow :D

Posted by Arun on Wednesday, 11-Feb-09 00:34:48 UTC
Quoting AnarchX
Ironlake (the IGP-die) should be >=100mm² @ 45nm. :shock:Even with Intels architecture should be a significant performance bump over GMA X4500 possible, but nicer would be indeed some SGX-tech in this space.:lol:
You're too optimistic. Intel's IGPs just [censored]. G45's current die size is 98mm² on 65nm (see Page 11 of: http://www.intel.com/Assets/PDF/designguide/319972.pdf ), and ICH10 is way bigger than SB600/700 too. People always think Intel's IGPs are so weak because they're cheapskates. No, sorry; they just can't design that stuff to save their lives.


Add your comment in the forums

Related intel News

32nm sixsome over at RealWorldTech
Intel Core i3 and i5 processors launched
Analysis: Intel-TSMC announcement more complex than reported
Intel and TSMC join forces to further Atom
Intel announce Core i7 processors, reviews show up
Intel's Aaron Coday talks to Develop about Larrabee
Larrabee to also be presented at Hot Chips
Intel Larrabee @ SIGGRAPH 2008
Larrabee's Rasterisation Focus Confirmed
Nehalem Article @ RWT + 3.2GHz samples(?)