Bookmark and Share


Advanced Micro Devices seems to be planning extremely rapid transition of its desktop microprocessors to 32nm silicon-on-insulator fabrication process, a person with knowledge of the matter said. By the second quarter of 2012 virtually all AMD processors will be made using 32nm SOI technology and be based on new designs.

At present the absolute majority of desktop microprocessors shipped by Advanced Micro Devices are AMD Athlon II and AMD Phenom II chips in AM3 form-factor. But already in a little bit more than a year from now all the desktop chips that AMD will sell will be made using 32nm SOI process technology and will rely on code-named Bulldozer, Brazos and Llano designs, the source indicated.

Approximately 20% of AMD's desktop microprocessors in Q2 2012 will be in AM3+ form-factor and will belong to the FX family of chips, which is based on Bulldozer micro-architecture. About 10% of desktop chips will come in FT1 ball-grid-array form-factor and will therefore be based on Brazos design. The remaining 70% will be Llano accelerated processing units with two or four x86 cores as well as a Radeon HD 6000-class graphics engine in  FM1 form-factor.

It is remarkable that after a number of delays AMD seems to have rather ambitious plans for the 32nm ramp. By contrast, it usually takes years for Intel Corp., the much larger rival of AMD, to transit to one micro-architecture to another or from one technology process for another.

AMD did not comment on the news-story.

Tags: AMD, Bulldozer, Llano, Phenom, Deneb, Athlon, 32nm, Brazos, Ontario, zacate


Comments currently: 22
Discussion started: 02/11/11 12:03:36 PM
Latest comment: 02/15/11 05:42:20 PM
Expand all threads | Collapse all threads


"It is remarkable that after a number of delays AMD seems to have rather ambitious plans for the 32nm ramp. By contrast, it usually takes years for Intel Corp., the much larger rival of AMD, to transit to one micro-architecture to another or from one technology process for another."

The key word here is ambitious! So far as AMD's track record is concern, this is a far fetch dream.
0 0 [Posted by: dudde  | Date: 02/11/11 12:03:36 PM]
- collapse thread

That's tottaly untrue. AMD transition since 130nm SOI was always rapid. 90nm line quickly replaced od 130nm designs simply because it was cost effective. 65nm budget Am2+ G1/G2s also quickly populated market and push out 90nm only in enthusiastic segment wile avaiting troubled K10 architecture.

45nm was best example of quick rampup in less tnah 6mont no Athlon based on 65nm Kuma cores werent available.

65nm was neede because of price cuts, and 45nm Athlon II replacements were higly anticipated because of cost effectivenes and finally job done properly over previosly buggy B2 and poor performing 65nm K10 overall.

0 0 [Posted by: OmegaHuman  | Date: 02/11/11 08:53:12 PM]
not really. they are always quick with the transitions.
they start later (because of their smaller size and the fact they only have 1 factory to recover the costs of developing a new node with, they need to be later. pure economics. the later you develop the cheaper it will be, (it will be more off the shelf by then)
but once they do start they move quickly.
0 0 [Posted by: Countess  | Date: 02/12/11 09:54:08 AM]
nah.. they do start later... however, they still have to go through the phase of producing a good yield... like any kind of manufacturing, you got to be able to consistently produce a product. pure economics!
0 0 [Posted by: dudde  | Date: 02/12/11 07:12:46 PM]
not the issue. they only have fab to redo. intel needs to rebuild at least 3 fabs to take over all production and unlike AMD they dont/can't remodel for a new production process it on the fly.

because AMD has only had 1 fab for decades they needed to develop the tech to do it. and once converted, it makes no sense what so ever to make the old products in a converted factory (baring any big yield issue ofcourse)
0 0 [Posted by: Countess  | Date: 02/13/11 06:48:08 AM]
Although Dresden was the only fab, it used to have suffixes a,b,c,d, and AMD would update 2 out of 4 at the same time, then 6 months later update the other 2. This was probably timed with the slow part of year for sales, and a large stockpiling before hand. Dresden is very large, so thinking of it as 1 fab is dumb.

Additionally AMD now has 3 fabs with NY supposed to come online. I think the real reason the 32nm is projected to scale that fast, is because AMD signed contracts with GF to take certain amounts of shipments by certain dates. We already know they did this for their ATI gpu's but since GF canceled 32nm bulk, were released form those shipment quotas.
0 0 [Posted by: cashkennedy  | Date: 02/15/11 09:42:26 AM]

"it usually takes years for Intel Corp" because they don't really have any reason to move faster. They make more money like that by moving slowly. Also they keep AMD close enough so that they can use it as a viable example of strong competition. Without AMD they are monopoly. A second... VIA is not going to convince anyone that they are not a monopoly. So they hit the brakes, they make more money by doing so, they control the speed that technology is moving ahead and they don't worry about FTC.
0 0 [Posted by: john_gre  | Date: 02/11/11 02:41:55 PM]
- collapse thread

And intel is bad and worse competitor just by they're ramping up slowly, and yet have twenty four month head start. :rofl:

You describe it your way, but there's one, or more, different but reasonable explanations.

Processing node transition costs a lot. Even Glofo ex.AMd with just one factory which is going under transition not a whole set or more fabs. And intel as biggest player certainly doest want to spill out their profits and pushing out better processing nodes while 9mo last time (on 45nm transition) gave them much more than just a breathing room to split ahead from competition.
They get opportunity to produce cost effective products and simultaneously pollute market with twice smaller chips than competition which are faster and more lucrative based on marketing. That at the same time allowed them to deliver twice as many chips for higher price.
Then after polluting one market they could afford to focus themselves to even more improve ROI by delivering crappy, but extremely well marketized product like Atoms that really hadn't any competition. At least not valuable one, as it was 60mm2 chip while MAD could only offer some cutdown A64 X2 (G2) which were almost 3x larger and not really profitable chips. Nor really attractive to the market even if they could be mass produced, and not just being some vaporware -- piled stocks of old product pushed onto market under newly branded *Athlon Neo* name.

0 0 [Posted by: OmegaHuman  | Date: 02/11/11 09:28:12 PM]

If this is really going to happen, this will be the first time we see such a transition backed by a contract manufacturer (GF).
0 0 [Posted by: Dresdenboy  | Date: 02/11/11 06:25:33 PM]

What is really interesting in this hype news is that is deliberately neglected to mention that in time while AMD will fully ramp up it's 32nm node intel will already start producing 22nm chips and polluting market with less expensive and more efficient chips.

LRB aka. Larabee incarnations could be first intels 22nm success, following not too much anticipated IvyBridge dumb shrink, which will push AMD in even deeper troubles they never experienced before. Because not even 28nm high end TSMC process node couldn't be more cost and probably power efficient than almost 50% smaller 22nm intels quasi-gpus.

Nobody even tried to explain how node life cycle was usually TWO YEARS. And AMD lagged in processing node transition behind intel for 9mo during on 45nm node, and now it's more than obvious that 32nm transition will follow intel's 32nm node for more than 18mo. And this describes market situation much better than some overhyped statements about Q2 2012 and AMD ramping up 32nm litography node that intel already delivered in Q1 2010!!
0 0 [Posted by: OmegaHuman  | Date: 02/11/11 08:58:11 PM]
- collapse thread

I agree with you, and I am an AMD fan. I think Intel has done a superb job lately despite of the X58 chipset bug. AMD is not moving fast enough. All the speculation about why Dirk Meyer was let go (after successfully turning AMD around), I think part of it has to do with this. Afterall, AMD is fighting an up-hill battle against Intel, who is ten times bigger.

I hope AMD comes back strong with BD and in the near future. We will all benefit from it.
0 0 [Posted by: gamoniac  | Date: 02/11/11 09:32:23 PM]
Larabee is a great GPGPU but it is a bad bad idea for as a GPU. not even a 22 vs 28nm node advance will compensate for that. and of course their driver team will again make it completely useless as a gpu too.

and while intel might be sooner with 22nm it will again, as always, take more then a year to switch even half their cpu's to the new node.
and the bulldozer module design, fitting just about 2 cores little more then the space of one if AMD is to be believed that would compensate plenty on the cost front.

and the 32nm delay for AMD will be 14months, not 18. it being the first SOI+HKMG process in history, a little extra delay is justified. (along side releasing 3 completely different core designs in the space of 6 months)

i see plenty of opportunities for AMD.
0 0 [Posted by: Countess  | Date: 02/12/11 09:39:31 AM]
As far as i remember Larabee has been canceled as a GPU product. Knights Corner / Ferry are the only products coming from the Larabee research and are intended to just be used for GPgpu computing.

I also agree that larabee even 1 process node ahead will not be competitive, which is why intel scrapped the project.
0 0 [Posted by: cashkennedy  | Date: 02/15/11 09:48:24 AM]
"Larabee is a great GPGPU but it is a bad bad idea for as a GPU. not even a 22 vs 28nm node advance will compensate for that."

I think you underestimate intel's marketing yet again. There's only less than 15% that goes into stock earnings. And do you think the rest really goes into rnd :grin:

Newer procesing nodes not something intel needs to do a fully year to ramp up in just one plant. They at least have several "smaller ones" that went under transition. And 32nm node as it seems was just dumb shrink node over 45nm which might be the reasoned thru ecocrysis and lack of demand growth. As it now seems intels 22nm node will be purely different story.

Yep it would be 14month where? In server market. AMD ramp up 45nm already in Q4 2008 (human years) and still no announcement of 32nm desktop chips in Q1 2011. Of course server market is something different but server platforms arent reachable for all of us even if we could afford them.

I'm glad you see plenty opportunities. But DAMN is under reconstruction (read: for sale) ever since they underwent on ATi spill out. And recent AMD brand consolidation just confirms that

0 0 [Posted by: OmegaHuman  | Date: 02/15/11 05:29:32 PM]

About that "news" i can only say: "Blessed who believes".

Or downside of the story, in one year none will be interested in current 45nm offering from AMD and no matter how good or bad 32nm yields will be, AMD won't have chances to sell 45nm cpus.
0 0 [Posted by: Azazel  | Date: 02/11/11 09:41:14 PM]
- collapse thread

the mix of bulldozer vs brazo tells you plenty about the mix of products. the vast majority of cpu's are cheap and 45nm cpu's will be that and if still available will sell all right.
OEM buyers dont care how the cpu performs, just look at how success blatantly awfull cpu's like the pentium-D where and still sold great through OEM's.
AMD now has access to that same market now that intel's anti-trust practices have been put a end too.

and i have no reason not to believe this news. AMD has always been faster in completing the full transition.
0 0 [Posted by: Countess  | Date: 02/12/11 09:48:55 AM]
I bought plenty of Pentium D's, it was dual core and was way cheaper then AMD x2's. If a product being inferior to another product means it shouldn't sell, then AMD would have no customers right now. Price means something, which i think is what you were perhaps realizing is why the OEM's sold pentium D's, because moving a PC with a 500 dollar x2 processor was a lot harder.
0 0 [Posted by: cashkennedy  | Date: 02/15/11 09:51:04 AM]
Well you're one of the geeks that still hadnt heard of
and blurred compilation

Of course they werent fined with some reasonably high sum, they spent more on bribery than on fines
0 0 [Posted by: OmegaHuman  | Date: 02/15/11 05:42:20 PM]
I concur "Blessed who believes" (edited books always needed grazing flock in main cameo roles)
0 0 [Posted by: OmegaHuman  | Date: 02/15/11 05:32:57 PM]

Right now AMD is a full node behind Intel (45nm vs 32nm), but while Intel transitions to 22nm they will both be on 32nm and AMD will have a more competitive architecture. I look for competition to heat up this year.
0 0 [Posted by: bbo320  | Date: 02/12/11 11:34:57 AM]
- collapse thread

Competition lagging behind two years, well if at least these two years cold pass without unmerciful taxation.
0 0 [Posted by: OmegaHuman  | Date: 02/15/11 05:36:18 PM]

Coach to players:

Last year we were overconfident.
This year we're really going to get them.

Don't you feel inspired by this?
0 0 [Posted by: etudiant  | Date: 02/13/11 08:23:49 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