News
 

Bookmark and Share

(13) 

After many years of relying on silicon-on-insulator (SOI) process technology, Advanced Micro Devices plans to start using bulk process technology for its state-of-the-art Opteron processors code-named Dublin and Macau in 2013, according to a slide that resembles those from AMD's roadmaps.

AMD's forthcoming Bulldozer micro-architecture based Opteron "Interlagos" and "Valencia" microprocessors are set to be made using 32nm silicon-on-insulator process technology at Globalfoundries. The code-named Terramar and Sepang chips, which will feature enhanced Bulldozer micro-architecture (also known as Piledriver) and will have up to 20 and 10 cores, respectively, will also be made using the same manufacturing technology in 2012. But the central processing units (CPUs) for servers that will succeed them in 2013 - Dublin and Macau - will be made using 28nm fabrication process with high-k metal gate, according to a slide published by CPU World web-site that resembles those from AMD's documents.

Terramar and Dublin will use the same G2012 server platform code-named Porto, whereas Sepang and Macau will utilize C2012 server platform known under Luxembourg code-name. AMD G2012 infrastructure will support quad-channel DDR3 memory as well as PCI Express 3.0 controller. The C2012 infrastructure will sport triple-channel memory controller and will also support PCIe 3.0.

It is noteworthy that Dublin and Macau are currently not supposed to increase core count to over 10 for one die or 20 for one MCM module. Potentially, this means that the next-generation Bulldozer micro-architecture will sport a significant amount of enhancements that will allow the new chips to tangibly increase performance without boosting the number of x86 cores. According to AMD's longer-term sketch roadmap, the Opteron processors in 2013 will not start to integrate stream processors to accelerate certain tasks. 

Given the fact that at present both Dublin and Macau are in development, AMD can easily change its opinion regarding any technical specifications or design decisions.

AMD did not comment on the news-story.

Tags: AMD, Macau, Dublin, Bulldozer, Terramar, Sepang, 32nm, 28nm, Piledriver

Discussion

Comments currently: 13
Discussion started: 08/03/11 07:59:39 PM
Latest comment: 01/23/12 11:12:11 AM
Expand all threads | Collapse all threads

[1-3]

1. 
45nm SOI -> 32nm SOI -> 28nm Non-SOI

Not going to happen

45nm SOI -> 32nm SOI -> 22nm FD-SOI

more likely
3 1 [Posted by: seronx  | Date: 08/03/11 07:59:39 PM]
Reply
- collapse thread

 
sounds about right because I read somewhere that CPUs don't use half nodes (28nm is one)
0 0 [Posted by: psycho_mccrazy  | Date: 08/03/11 08:02:54 PM]
Reply
 
Nope. 28nm Non-SOI is a way to go for AMD.
AMD can't afford to have exclusive tech process for itself alone.
SOI Fab utilization is the biggest problem which AMD is facing now, 32nm and Brazos being made on TSMC will exaggerate it, and it will turn AMD in red very soon.
Thanks to Doug Grose, which was fired for that, this year WSA removes util tax off AMD. Next year will be quite different.
1 0 [Posted by: Azazel  | Date: 08/03/11 09:22:29 PM]
Reply
 
Llano and Bulldozer, then Trinity and "Piledriver?" are all on 32nm SOI HKMG they can afford it
and it's GlobalFoundries for APUs the Llano ones and the Trinity ones....not TSMC
1 0 [Posted by: seronx  | Date: 08/03/11 11:19:55 PM]
Reply
 
Look at the slide again. Dublin and Macau will be made using 28nm. AMD has chosen the path - many cores/low freq/low perf per a core. Bulk process is quite good for that path. Will this path be succesive we'll see soon.
0 0 [Posted by: Azazel  | Date: 08/04/11 02:16:57 AM]
Reply
 
Again, 2013 Plans for Next-Gen Bulldozer is 22nm

2014 is Server APU not 2013
0 0 [Posted by: seronx  | Date: 08/04/11 12:28:59 PM]
Reply
 
AMD 2013 is 28nm ET-SOI( 28nm Fully Depleted SOI with Ultra Thin Buried Oxide)

I was wrong but 2014+ is still Server APU
0 0 [Posted by: seronx  | Date: 01/23/12 11:12:11 AM]
Reply
 
The 28 nm Non-SOI is only make sense, if they integrate the GCN architecture CUs inside the CPU. They can make CPU-cores in Non-SOI easily, the only downside is the clock speed, which is may not go up to 3 GHz. On the other side ... making CUs on a SOI node is very-very hard.
0 0 [Posted by: zorg  | Date: 08/04/11 01:23:19 AM]
Reply
 
GCN (?? i need to google it --- Graphic Core Next)

omg. DO NOT DISINFORM US, IT'S NOT ARCHITECTURE but just plain old AMDs marketing gimmick As old Fusion they propagate since 2007 is pretty worn'n'torn nowadays

0 1 [Posted by: OmegaHuman  | Date: 08/06/11 04:39:27 AM]
Reply
 
it'll happen rethink it twice. they already have 28nm line for gpu if they'll test any production in GloFo, and in 2013 they'll need something to fullfill 28nm lineup. They'll probably wont ramp up 22nm so soon as they let us wait more than 30 month (FULL YEAR LONGER) just for 32nm processors hit the stores after first 45nm in Dec2008 (920/940be)
0 0 [Posted by: OmegaHuman  | Date: 08/06/11 04:29:14 AM]
Reply

2. 
Take AMD's roadmaps with a huge helping of salt. A chimp predicting the weather is more accurate than AMD with their estimates.
2 3 [Posted by: AnonymousGuy  | Date: 08/03/11 09:16:17 PM]
Reply
- collapse thread

 
This is pretty realistic estimate considering how GloFo now works, and how long they'll need to finish that new 22nm factory. It's budgeting projection, not a cocky one as it was usual in the past.
0 0 [Posted by: OmegaHuman  | Date: 08/06/11 04:31:40 AM]
Reply

3. 
GLOBALFOUNDRIES Fab 8, Module 1 in Luther Forest, Saratoga County, NY:
- "Initial production is expected to ramp at the 28nm technology node and move to volume manufacturing on the 22nm node." (GLOBALFOUNDRIES, July 24, 2009)

fab8update.com June 15, 2011:
- risk production in mid-2012
- volume production in early 2013
2 1 [Posted by: T9000  | Date: 08/04/11 05:22:42 AM]
Reply

[1-3]

Add your Comment




Related news

Latest News

Thursday, May 23, 2013

8:52 pm | AMD Officially Launches New-Generation APUs for Mobile Applications. AMD Introduces Kabini, Temash and Richland Accelerated Processing Units

6:51 pm | OCZ Reveals Vertex 450 Solid-State Drives: High-End Performance at Mainstream Prices. OCZ Introduces New SSDs Based on Indilinx Barefoot 3 Controller

3:40 pm | Nvidia Unveils GeForce GTX 780: GK110-Based Consumer Solution for $649. Nvidia’s Cut Down Titan LE Becomes GeForce GTX 780

Wednesday, May 22, 2013

11:59 pm | Be Quiet: All Current Power Supplies Are Ready for Core i “Haswell”. Be Quiet Claims Top-to-Bottom Compatibility of PSUs with New Intel Chips

11:51 pm | OCZ Partners With Netgear to Deliver Flash-Based Data Center Storage in a Box Functionality to SMBs. Leading OCZ Enterprise-Class Deneva 2 SSDs Now Qualified on Netgear's ReadyDATA 516 NAS Device

11:07 pm | Half of the World’s Population Will Be Covered by 4G/LTE Networks by 2018 - Research. More Than 1 in 2 People Will Be Covered by 4G/LTE-FDD by 2018

9:38 pm | Sony Starts Manufacturing of PlayStation 3 in Brazil. Sony Begins to Make PS3 Game Consoles in Latin America

9:11 pm | Nvidia Grid Unleashes Graphics for Virtualized Desktops. Nvidia and Citrix Commercializes Grid Technology for Virtualized Desktops

8:57 pm | MIT Scientists Mix Graphene with Hexagonal Boron Nitride to Create New Material for Computer Chips. Researchers Create New Material for Semiconductors

8:43 pm | Intel Can Enable a Successful $200 PC in the Age of the Media Tablet – Analysts. Market Observers Mull Viability of $200 PCs on Current Market

8:09 pm | Microsoft Not Worried About Xbox One’s Lack of Backwards Compatibility, Vows Big Xbox 360 Announcement at E3. Microsoft Believes Xbox One Will Not Require Games of Xbox 360

7:52 pm | Asrock’s A-Style Mainboards Set to Be Waterproof. Asrock’s New Intel 8-Series Mainboards to Feature Conformal Coating

7:35 pm | Nvidia Announces PhysX and APEX Support for Microsoft Xbox One. Microsoft Xbox One Games to Use PhysX and APEX