Bookmark and Share


“My dear, here we must run as fast as we can, just to stay in place. And if you wish to go anywhere you must run twice as fast as that.”

– Lewis Carroll, Alice in Wonderland

Although the chief executive officer of Advanced Micro Devices insists that the company should only use two process technology nodes at a time, other executives from the company believe that just two process technologies may not be enough. Moreover, in many cases proven nodes mean lower costs. In general, AMD expects process technologies to be utilized for longer periods than they are these days.

“It is getting tougher and tougher to get to new nodes. 28nm might be with us a little longer than people [think]. [It will be a while] before they jump into 20nm node or 16nm or 14nm,” said Devinder Kumar, the corporate controller and the interim chief financial officer, at Raymond James IT supply chain conference.

Both leading foundries on the planet – Taiwan Semiconductor Manufacturing Co. and GlobalFoundries – are working hard on development of thinner process technologies and their rapid ramp up as many customers these days, primarily those who develop various solutions for mobile devices, want the lowest possible power consumption. Globalfoundries, the main foundry partner of AMD, recently introduced 14nm XM [extreme mobility] fabrication process that uses 14nm FinFET devices combined with 20nm-LPM process back-end-of-line (BEOL) interconnect flow.

AMD currently does not have plans to use 14nm XM process technology since it has no plans to make chips for smartphones. However, a recent roadmap shows that the company plans to rely on 28nm fabrication process even in late 2014, when it launches its next-generation high-performance FX-series processors for desktops and new Opteron products for servers. As it appears, AMD wants to extend lifetime of nodes to ensure maximum availability and lowest cost.

“Our vision is extending the lifetime of nodes. I think we showed with our Brazos processors that were on 40nm for a long time and we competed very effectively with other products on the market that were two nodes ahead of us. We could sell at a nice cost profile and the functionality that the consumer wants today. […] So, no nodes transition is easy. […] Earlier this year we have brought our GPU products to 28nm. You will see [Kabini and Temash] APUs in the first half of next year on the 28nm. […] We have not said when we are going to 20nm, which would be the next node,” said Ruth Cotter, vice president of investor relations at the Raymond James conference.

It is not a secret that foundries are gradually increasing production prices on new nodes, especially those that require massive capital investments to ramp up. However, there is another reason why AMD is not truly interested in rapid transitions to newer process technologies: the company’s ambidextrous strategy.

The ambidextrous strategy implies usage of similar building blocks inside completely different chips. While it clearly lowers development costs, it also means that AMD will be unable to leap forward one particular product line. Each building block needs to be developed for a particular process technology to be efficient in terms of performance. At the same time, each building block needs to be tailored for different product designs to ensure high-performance across the board. Hence, to boost the speed of one product by taking advantage of a new manufacturing tech, AMD needs to redevelop virtually all building blocks it has and tailor them for all possible designs to ensure maximum performance and power-efficiency across the product family. Obviously, this takes time and money.

While all semiconductor companies tend to be careful with transitions to newer nodes, the leading semiconductor companies – Intel Corp., Samsung Electronics and Qualcomm – tend to make products using the most advanced fabrication technologies possible in order to sell the best chips in the industry. As it turns out, such an approach has always made a lot of sense…

Tags: AMD, 28nm, 20nm, 14nm, 14nm XM, Globalfoundries, 40nm, Brazos, Kabini, Temash, Semiconductor


Comments currently: 40
Discussion started: 12/13/12 05:44:19 PM
Latest comment: 12/17/12 08:27:23 AM
Expand all threads | Collapse all threads


Its incredibly funny they think they can put a good spin on ceasing to develop better products, or stagnation of process technology.

And the efficiency of very precisely tailoring units to 1 process is quickly overcome by the efficiency of just shrinking everything 30%. > and when they have to finally make the jump to the lower process node they will still have to spend another product generation to reach the maximum efficiency for that node.
4 5 [Posted by: cashkennedy  | Date: 12/13/12 06:03:04 PM]
- collapse thread

Bear in mind FD-SOI allow for lower voltages or higher frequencies (sub 4Ghz).

AMD plan to have Kabini/Temash at ST/GF's 28nm planar FD-SOI near the end of Q12013 > Richland/Vishera 2.0 at the same process node Q32013 > Kaveri at 20nm planar FD-SOI Q12014.

They have ditched the 28nm process node with Steamroller and will focus on redesigns to attain a higher IPC at lower frequencies, a trade off related to having different Decode units within a module, so FD-SOI can shine.
20 3 [Posted by: Green lantern  | Date: 12/14/12 04:32:04 AM]
Hey if the Kaveri might be made at 20nm. Could that mean if AMD started to recover enough from the global computer slump(and if the major world economies improved), that they could adopt the 14XM for tablets and smartphones, since the 14XM is essentially a repurposed 20nm design for extreme mobility?
14 3 [Posted by: 78412  | Date: 12/14/12 01:18:08 PM]

show the post
6 20 [Posted by: tedstoy  | Date: 12/13/12 06:14:00 PM]

The smaller the nm die size on a node gets the harder it is to produce and the hotter it gets. I see it coming to a point where all semi-conductor companies are going to drasticly slow down in the die shrink process in the next 5 years, due to the fact that they have two major issues that they will end up having to face when the die size gets under the 14nm range and that is it will become a lot harder to pack in a billion + transistors in such a small die and the chips will become a lot hotter because you are packing in soo many transistors into such a small die space. An small indication of that already is ivy bridge on 22nm and the heat concerns will get greater and greater the more transistors they pack into the chip as the die continues to shrink. These are the problems that the semi-conductor industry will have to face and until they can figure out a away to fix these two major concerns then the industry will be at a stalemate with 14nm until they get those two major concerns addressed that will allow them to continue to shrink the die even smaller.
4 5 [Posted by: SteelCity1981  | Date: 12/13/12 08:34:13 PM]

So DAMN is preaching what their sharky friends in Globalfoundries are telling them to say. Glofo cheaply bought off factories around the world monopolizing the market and now they feel safe to ask for government grants "to invest" *our money*, not theirs as obliged to, into technology while the rest sinking into their private pockets

AMD if you wish to stay competitive you cant shark small customers aka. PC desktops, by favorizing big ones at the same time. And write off lawsuit winnings against Intel with 30% success in paying back lost interest just because you needed money desperately (Y2k9) And at the same time you deliberately decided to put on hold your Bulldozer release for two whole years. Hm, how convenient.

So we ended up with one uncompetitive firm while at the same time Intel paid out those settlement they pay to DAMN (is it yet paid in full at all) with insane desktop chip pricing ... And what's the best thing piplz still buying that. (Intel overprized crap) When they have no other choice. So this more and more resembles to VIA and how Intel kill them with their inferior product aka. Atom but peoples stupid blindness is unforseeable and without limits.
19 3 [Posted by: OmegaHuman  | Date: 12/14/12 12:41:27 AM]

Currently the only benefit to smaller nodes is power reduction. Since most CPUs don't have a need for ULV, there is no big advantage to a smaller node other than for ULV. 28nm can currently deliver the needs for desktop and servers without any issues.
19 4 [Posted by: beenthere  | Date: 12/14/12 07:02:25 AM]
- collapse thread

I reckon beenthere that the plan is to improve the chip design to deliver better power savings (rather than just shrinking the die) in order to guarantee supply of them at 28nm for big contracts. Hello Apple. Hello Microsoft.
18 2 [Posted by: linuxlowdown  | Date: 12/14/12 08:15:12 AM]
The benefit of smaller nodes is that you can have smaller die size or more transistors at same die size.

16 0 [Posted by: maroon1  | Date: 12/14/12 04:36:45 PM]

"However, a recent roadmap shows that the company plans to rely on 28nm fabrication process even in late 2014, when it launches its next-generation high-performance FX-series processors for desktops and new Opteron products for servers. I think we showed with our Brazos processors that were on 40nm for a long time and we competed very effectively with other products on the market that were two nodes ahead of us. We could sell at a nice cost profile and the functionality that the consumer wants today. "

Here you go for the 10th time AMD's management tells publicly they are not interested in competing with Intel in the high-end CPU race. And people still expect them to do so? They just admitted again they are targeting cost conscious consumers and plan to use 28nm for next generation FX CPUs.

Avon/789/1234/Bro et al., please don't come back in 2013 and 2014 and tell everyone how AMD is failing and can't compete when AMD admitted it's not even their strategy in 2012.

I just hope this doesn't affect their GPUs. If they still release HD9000 series on 28nm node, they'll be uncompetitive against 20nm Maxwell. Unlike NV fanboys who continued to buy slow and power hungry 40nm Fermi cards (GTX550/560/560Ti/570/580) when AMD had 28nm 7750-7870 series out already in the under $350 price level, AMD buyers aren't brainwashed enough to buy 28nm discrete GPU's in 2014 if the competitor already has 20nm ones out. In the GPU space, the node matters a heck of a lot more since GPUs are way more power hungry and pack a lot more transistors, where node shrinks are critical for exponentially increasing GPU performance by achieving higher transistor count/density and lowering leakage.
22 14 [Posted by: BestJinjo  | Date: 12/14/12 09:10:34 AM]
- collapse thread

show the post
5 13 [Posted by: Avon4Balls  | Date: 12/14/12 02:21:19 PM]
Stop dreaming, NVIntel won't be approved by regulators.

No need to change the socket since it allows people to upgrade without needing a new motherboard. That's actually a good thing AMD is doing that I wish Intel did to save me buying a Z87 mobo I don't really need next year.
14 4 [Posted by: BestJinjo  | Date: 12/14/12 05:59:35 PM]

This means we will have 28nm until GF reaches 14nm.
On the other hand the GPU-s will get whatever TSMC has best.
So there are 2.
FX CPU-s with 28 nm will reach 5Ghz stock no doubt about it.
The APU-s on 28 nm will be beasts for the money payed.
8 2 [Posted by: HeadlessBottleneck  | Date: 12/15/12 02:19:50 AM]
- collapse thread

Steamroller will be done @ 20nm in 15 months time after Richland @ 28nm. This is the next step down the die walk.
8 3 [Posted by: linuxlowdown  | Date: 12/15/12 03:16:22 AM]
Steamroller will be done @ 20nm in 15 months time after Richland @ 28nm. This is the next step down the die walk.

Where did you get that information from? No such confirmations where made AMD on any roadmaps released to the public or even on the leaked ones.

If you have a link for the source please provide me one. Thank you!
13 0 [Posted by: vanakkuty  | Date: 12/15/12 05:19:48 AM]
show the post
1 11 [Posted by: Avon4Balls  | Date: 12/15/12 06:02:48 AM]
Where did I taunt him? I was just asking a normal question.

Its a bit rich this suggestion is coming from you as your reputation in here isn't very great to say the least. Appreciate it if you don't hijack a conversation when you are not addressed to. Thank you.
12 2 [Posted by: vanakkuty  | Date: 12/15/12 06:29:39 AM]
show the post
1 12 [Posted by: Avon4Balls  | Date: 12/15/12 06:38:23 AM]
Says the person who is the most down voted and the dismissed one as a spammer. Err yeah sure.... be happy in your little bubble. You sound like OBR who is another sad case running around the internet for attention.
12 2 [Posted by: vanakkuty  | Date: 12/15/12 06:42:26 AM]

show the post
0 3 [Posted by: dudde  | Date: 12/16/12 04:01:51 PM]
- collapse thread

Everyone else is struggling with node transitions since it's becoming harder and harder to shrink transistors to lower nodes. Even Intel's world's leading manufacturing facilities had some issues at 22nm. TSMC and GloFo are behind Intel.

If you followed the news here at Xbitlabs or other tech sites, AMD's relucancy to adopt most expensive/newer nodes has been discussed a year ago by the current CEO:

"Now, let's talk about 20nm and 14nm. I think that we really flying hard in the path of subatomic environments. The price advantages as we move down nodes are starting to wane. The ability to [quickly improve] yields and ramp up our products (which have fixed amount of time) is under exceptional pressure. It costs huge amounts of money. I think we have to be strategic and think about how quickly we go down the node,[...] Just go look at the cost of wafers as you move down those technologies, they are not going down, they are going up! If the yield does not go up, how do you get your return? You have to charge bigger prices."

TSMC raises prices on AMD and Nvidia [for 28nm]

Nvidia suffers as TSMC's 28nm ramp steeper than expected

The rising wafer costs and ramp up time to justify the move to next node process has resulted in a lot more risk than before. New fabs costs more than ever and naturally when you adopt a newer nodes, your yields are worse than existing and mature node. These added costs are then passed on to fab customers like AMD/NV, etc. Why do you think NV was unable to launch GK100/110 in the consumer space in 2012? It could never meet the volumes at existing 28nm yields and it wouldn't have been able to afford selling 500mm^2 die at 2012 28nm wafer prices at just $499. 20nm wafers will be even more expensive.
1 0 [Posted by: BestJinjo  | Date: 12/17/12 08:27:23 AM]


Add your Comment

Related news

Latest News

Wednesday, November 5, 2014

10:48 pm | LG’s Unique Ultra-Wide Curved 34” Display Finally Hits the Market. LG 34UC97 Available in the U.S. and the U.K.

Wednesday, October 8, 2014

12:52 pm | Lisa Su Appointed as New CEO of Advanced Micro Devices. Rory Read Steps Down, Lisa Su Becomes New CEO of AMD

Thursday, August 28, 2014

4:22 am | AMD Has No Plans to Reconsider Recommended Prices of Radeon R9 Graphics Cards. AMD Will Not Lower Recommended Prices of Radeon R9 Graphics Solutions

Wednesday, August 27, 2014

1:09 pm | Samsung Begins to Produce 2.13GHz 64GB DDR4 Memory Modules. Samsung Uses TSV DRAMs for 64GB DDR4 RDIMMs

Tuesday, August 26, 2014

10:41 am | AMD Quietly Reveals Third Iteration of GCN Architecture with Tonga GPU. AMD Unleashes Radeon R9 285 Graphics Cards, Tonga GPU, GCN 1.2 Architecture