MOUNTAIN VIEW, Calif.--Advanced Micro Devices Inc.'s (AMD) roadmap is about to see some significant changes according to sources at the company.
With online reports suggesting AMD could scrap its 28-nm APU production at Globalfoundries in order to move the manufacturing to TSMC, EE Times has confirmed with sources at the firm that roadmap changes are “imminent” and that “28-nm schedules aren’t what we wanted.” Whether those changes pertain to a switch of manufacturing partners, however, has yet to be fully substantiated.
“We will have a roadmap update in the very near term,” said a source at AMD though he stressed that the firm had always kept its manufacturing options open.
“We did not get specific about our manufacturing partner for 28-nm APUs for 2012, we have always said that both GlobalFoundries and TSMC are strong manufacturing partners and we reserve rights to keep our options open,” he said. All of AMD’s 28-nm discrete GPUs are already being made at TSMC.
Another source at the company told EETimes he was not surprised to hear the murmurs of manufacturing changes surface, but said any public announcements about such a change in the roadmap would be considered highly sensitive and would likely be broken to financial analysts first.
EETimes has discovered that several analysts have been invited to a recently scheduled meeting on Dec. 5th.
“Changes are coming,” said In-Stat’s Jim McGregor, saying a change in manufacturing partners would not surprise him in the slightest, especially with relationships between Globalfoundries and AMD growing increasingly strained.
“AMD has had so many problems with Globalfoundries this year, from low yields to low ramps and high cost,” he said adding, “The relationship between the two does not seem to be providing benefit to either company at the moment.”
An agreement which expires on January 1, has meant that AMD only pays Globalfoundries for viable 32-nm dies, which has a financial loss for Globalfoundries based on the difficulties it has had achieving high yields. After January 1, AMD would have to go back to paying for every wafer, successful or not, something the firm may not be very keen to do.
McGregor said that should the 28-nm Globalfoundries made APUs really be scrapped, it would cause a “world of hurt” for AMD, which would be left a generation behind its competition for most of 2012, making AMD even less significant in the market.
“AMD just doesn’t matter. Even Intel doesn’t care about AMD anymore, the firm doesn’t even register on Intel’s radar, it’s that far behind,” McGregor noted.
AMD sources, however, were quick to defend the firm’s chances, even in the case 28-nm should change course over the coming weeks.
“Even if we were forced to stick with 40nm Brazos, it would not be the end of the world,” an AMD insider told us. The firm feels it has a successful product with Brazos, which is aimed at the $299-$400 notebook market. 20 million have shipped so far through Q3.
“We’ll be ok because of the price point Brazos is currently at in the market,” said the AMD insider, adding that the firm was a lot more optimistic about its upcoming Trinity products, which are the next generation to the 32-nm Bulldozer-based Llano.
“Trinity is in good shape,” he said, though some believe the firm may have trouble matching Llano’s performance-per-watt ratio and would leave a gaping hole in AMD’s ultra-low power product line at a time when tablets and lower powered notebooks are all the rage.
Meanwhile, Globalfoundries said it would not comment on its customer's foundry selection process or on their products unless they did so first. The spokesman also said problems with Llano had been specific to that product and that yields for AMD's 32/28nm Bulldozer products were on target and not affecting AMD's ability to meet customer commitments.
“We are still the only foundry producing HKMG products that can be purchased in stores now,” the Globalfoundries spokesman said, noting that the fab expected to ship “far more” HKMG volume in 2011 than all other foundries combined.
“In a recent call with financial analysts, AMD confirmed that the Llano ramp has essentially matched the Brazos ramp at TSMC. Given that Brazos is a very small die on the mature 40nm node process, this is a significant achievement. Llano is a much larger die, far more complex, and it is on 32nm with HKMG,” he said.
If AMD were to change manufacturing partners for its 28-nm APUs, it could take up to 18 months for chips similar to the planned --but failed-- Krishna/Wichita designs.
An official AMD Financial Analyst Day is to be held in February, with more updates and roadmap changes scheduled for that time frame, but in the meanwhile, we can expect public changes to be announced in December.
The story is partly wrong and exaggerated. Already months ago taiwanese media such as digitimes and commercial times reported that the chips are to be produced by TSMC and not Glofo. While Glofo is already busy with Llano and Trinity chips, TSMC also has not so much capacaties in 28nm for all its customers. Therefore, AMD is obiously shifting to more profitable graphic gpus. All in all, the story has little to do with Glofo or problems at AMD but with high demand for hardly available 28 nm products from TSMC.
AMD simply decided to let produce TSMC more Radeon GPUs with high margins instead of low margin APUs because
the 40nm Bobacats are still selling very well. It has nothing to do with Glofo. Please, get it!
Here is the link. And if you read the articles from Semiaccurate and Extremetech than you will see that it was no more than a (wrong) assumption that Glofo is intended to produce the chip. But this was not the case and now all this allegations have no foundation any more. The truth is totally different.
The wrong story at the wrong time with the wrong actors, wrong locations and wrong backgrounds. It was in summer when taiwanese media reported that TSMC will produce. There had never been any confirmation that Glofo would produce them, only considerations. Semiaccurate only refers into its own hearsay in an earlier article written by himself: http://semiaccurate.com/2010/11/18/glofo-will-fab-amds-krishna/
This is from a year ago and even earlier than the news from Taiwan from summer and when no decision had been made!
Jakobus, why are you complaining about the accuracy of a story from a site called "semi-accurate"?? Surely the name speaks for itself. As for the story written above, if you read it, I referred to the "reports" in passing and then said I had no confirmation on such a manufacturing move taking place, only confirmation of a roadmap change. You're right that there is absolutely no confirmation, and that is made abundantly clear in the above article, so I really don't see why you are so upset about it. It's an opinion piece, alluding to the reports, not confirming them, and CONFIRMING that the only thing we know for sure is that roadmap changes are coming in December. If you have issues with Semi-Accurate, that is fortunately not our problem, as the site has nothing to do with EETimes. Best!
Thank you for your answer. I am even more upset by the article on Extremetech and the fact that the official press release a few months about manufacturing issues earlier caused serious troubles. Very nice that you wrote that these issues are almost outworked already! Yes, but I do not really like to see false rumors to spread and to cause damage and I know that EEtimes is reliable in this regard and this should be verified. There is so much AMD-bashing these days and we must only read all these comments here.
Sorry and Thanks
Here you can compare it by your own:
There are already articles critisizing SA at ATI-Forum.de and planet3dnow.de
There seems to be some confusion about the article and the validity of the statements. The foundry business is not as easy as most would think. Building wafers for many customers and meeting their requirements for performance and cost is very difficult. As we move to smaller geometries, the problems increase. Having a low yielding process is not always the foundries' fault. It can also be caused by poor design techniques. The problems have to be worked out by the companies as a joint project supported by both or the consequences are dire.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.