backtop


Print 30 comment(s) - last by DeepBlue1975.. on Dec 17 at 9:36 AM


AMD's desktop platforms detail the switch from AM2+ to AM3  (Source: AMD)

Desktop offerings from AMD will largely include dual and tri-core K10 processors  (Source: AMD)

Not willing to yield anything to Centrino, AMD's "Puma" and "Shrike" initiatives will bundle all of the company's newest mobile technology into one ubiquitous platform  (Source: AMD)
Long gone is the talk of 4x4; AMD discusses its platform programs with a mainstream approach

AMD yesterday held its 2007 Financial Analyst Day in which it revealed its updated corporate roadmap. AMD has a flurry of new platform releases planned out for 2008, 2009 and 2010 and, contrary to previous analysts days, the company had no problem detailing its initiatives in full depth.

AMD’s recent Spider launch marked AMD’s first foray into the enthusiast market with its next-generation K10 architecture. AMD, however, has two more enthusiast platforms planned out for 2008 and 2009.

Towards the second half of 2008 AMD will release its second K10-based Enthusiast platform, dubbed Leo. Leo will feature quad- and triple-core 45nm Deneb and Propus cores manufactured on a 45nm process node. The chips will support DDR2 memory, HT 3.0 and run on the same AM2+ package as current Phenom processors. The platform launch will also consist of 790FX/790/770 chipsets with SB700 south bridges, and AMD’s new R600-based graphics processors.

In the first half of 2009, AMD plans to update its Leo platform. The largest change in the "Leo refresh" will be AMD’s move to DDR3 and the new AM3 package. The platform launch will also consist of a new RD890 chipset with the SB800 south bridge. AMD will also launch its next-generation ATI R700-series graphics processors at this time.

AMD also announced its new mainstream desktop platform, dubbed Cartwheel, scheduled for quarter of 2008. The launch will consist of quad-, triple-,and dual-core K10-based Toliman and Kuma cores. The new chips will be manufactured on a 65nm process node. AMD’s RS780 DX10 integrated graphics chipset will also be launched along with optional R600-series graphics processors with Hybrid Graphics functionality and Vista Premium certification.

In short, AMD’s Hybrid Graphics technology permits users to combine integrated graphics with specific discrete graphics cards that support Hybrid Graphics to form a cost-efficient CrossFire.

In 2009, AMD will refresh its Cartwheel platform with 45nm quad-, triple-, and dual-core Popus, Heka and Regor cores. Mimicking AMD’s update schedule for its enthusiast platform, the refreshed Carthwheel platform will support DDR3 memory and come based on the AM3 interface. The platform features an RS780 chipset with a new SB800 south bridge along with optional R700-series graphics processors and Hybrid Graphics Technology.

Two is the number of the day, and that is no exception when talking about AMD’s mobile platform roadmap.

In Q1 2008 AMD will introduce its Puma notebook platform. The platform launch will sport new dual-core Griffin processors along with the AMD RS780 chipset and integrated DirectX 10 graphics via AMD’s M8x graphics chip.
In 2009, AMD will elevate its game plan in the mobile segment with the release of its first Fusion-derived mobile chips.  Consumers will begin seeing the first fruits of AMD’s Fusion efforts, which AMD has touted for over a year now.  

The Shrike platform launch will feature AMD’s third-generation Stars core dubbed Swift. Swift, manufactured on a 45nm process node, will feature three K10 microprocessor cores and one graphics processor core. The platform will also feature M9x graphics with integrated DirectX 10 support along with support for DDR3.  Like other AMD technologies, the Swift's shift to DDR3 will require a new socket, dubbed FS1.

Absent from these roadmaps is the AMD 4x4 initiative.  AMD announced it will no longer pursue the knocked-down server platform for desktop enthusiasts last month -- just one year after it announced plans for three more generations of the technology.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Nice...
By Roy2001 on 12/14/2007 6:53:39 PM , Rating: 2
That is not correct. 65->45 transition is the most difficult one.


RE: Nice...
By KristopherKubicki (blog) on 12/14/2007 7:02:17 PM , Rating: 2
I believe the 65->45 transition was supposed to be a big deal because of the lithography change, but it turned out that AMD/IBM/Intel were all able to get to 45nm without changing the lithography process dramatically.

I think going from 45 to 32 is the "hard" one


RE: Nice...
By radializer on 12/14/2007 7:44:56 PM , Rating: 5
There are multiple facets to the 65nm to 45nm process technology transition as well as the transition to 32nm.

Lithography - the options are either to stick with 193nm dry litho (as Intel has done) or move to 193nm immersion litho (as TSMC has reported). Either ways, Kris is correct in stating that the 65->45 transition has used existing technology with modifications allowing a shorter learning cycle.

Further scaling to the 32nm node may require either a transition to 157nm lithography or some other tricks. However, 157nm is still not ready for mass production and has issues that need to be ironed out. So 45->32 may be a tough one as far as Lithography is concerned.

Gate Leakage - this is a different beast and is more of a physical limitation of the current ultra-thin SiON stacks. With Intel's approach of using high-K and metal gates on 45nm, they have bought themselves scaling headroom for a few more generations. The others will eventually have to follow suit with variants of their own and I believe all the major players (IBM, TSMC, Sony, etc) also have versions cooking in their labs. This is a non-trivial integration problem that has been plaguing the process industry for over 8+ years - so it's not going to be easy.

Therefore, as far as gate engineering is concerned, Intel has performed a tough 65->45 transition but may have an easier 45->32 move - while the others may have an easier 65->45 move but may end up facing a harder 45->32 one.

The end result is; it's not a clear "hard" or "easy" call but it will depend on the choices the company makes for its next process generation. Some will be easy, and the others hard.


RE: Nice...
By JumpingJack on 12/15/2007 4:13:18 PM , Rating: 2
I would disagree here... Intel was able to stay with dry litho at 193 (3 generations of learning) and a mature equipment platform.

AMD/IBM could not get the patterning to behave for dry to hit their geometric sizes needed for 45 nm (I suspect), and had to employee a new (and relatively unproven) immersion lithography for their critical layers... not that it won't be successful, but I would think this makes it harder.


"It looks like the iPhone 4 might be their Vista, and I'm okay with that." -- Microsoft COO Kevin Turner

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki