Globalfoundries and Samsung Electronics, two members of the so-called IBM fab club, said that they did not experience any issues with the gate-first 32nm and 28nm process technologies. The companies denied the claims of a financial analyst that was indirectly confirmed by Advanced Micro Devices, which delayed the roll-out of its future chips because of process tech-related problems.
Andrew Lu, an analyst with Barclays Bank, said in a recent note to investors that that the adopters of the high-K metal gate gate-first technology adopters had encountered some problems that were related to thermal instability, threshold voltage shifts and re-growth in the gate stack, which is serious for pMOS (positive type metal oxide semiconductor) at scaled electrical oxide thickness. Separately, AMD said that it would delay the release of its code-named Llano processor that sports x86 and graphics cores and is made using 32nm SOI process technology to mid-2011 from early next year.
''There have been some misconceptions about the gate-first approach to high-k/metal-gate. [...] There are no Vt stability issues and gate-first offers comparable or superior performance when compared to gate-last approaches. Our 32-nm high-k/metal-gate ramp is in early production at Fab 1 and we are confident in our ability to deliver for our customer and maintain our time-to-volume leadership position in the foundry industry. We are currently accepting designs for all of our 28-nm technologies. Multiple customer designs have already been silicon-validated, and many more test chips are in prototyping at Fab 1 on the way to early risk production late this year," a statement from Globalfoundries reads.
By contrast to Globalfoundries and IBM's fab club, TSMC and Intel use gate-last approach. The latter is shipping second-generation HKMG-based products these days.
''As you may recall, Samsung announced in early June the full qualification of 32-nm low power high-k metal gate in our S Line. That qual included a full 1000 hr high temp operating life (HTOL) and have experienced no such problems," said a spokesperson for Samsung.
Tags: IBM, Globalfoundries, Samsung, AMD, 32nm, 28nm, TSMC, Semiconductor
Comments currently: 4
Discussion started: 09/29/10 05:20:37 PM
Latest comment: 10/01/10 11:03:18 AM
I was kinda surprised Xbit didn't report on the GTC speech from GlobalFoundries earlier. It was pretty interesting. While indeed they did address the claims made by the analyst last year, they also mentioned gate-first has allowed a 10-20% smaller die than TSMC's gate-last 28nm approach, although I don't know if that's actually validated by test masks from a company like Qualcomm that may use both TSMC and GF. While TSMC says that designing towards their gate-last process can eliminate the size deficit, there's likely 'some' truth to it, as there has appeared to be all along. It's likely not 20%, but 5-10% sounds reasonable comparing both companies quoted numbers compared to TSMC's 40nm.
In this article about the GTC event, the publisher of The Microprocessor Report has a quote agreeing with what my speculation has been for a while; Gate-last (TSMC/Intel) will net you a little bit more performance(/watt), but the gate-last (GF) approach looks to be an ideal fit for System-On-a-Chip (ie ARM designs from Qualcomm/Texas Instruments) and foundry customers (like 'ATi' and nVIDIA) because of smaller die size and allowance of designing a chip similar to old processes without having to learn a whole new set of tricks. It's worth a read:
I stand by my former guesstimate: All things equal, TSMC will probably be ~5% faster clock/watt and GF ~5% smaller die size for a given design, give or take a couple/few percentage points. What is 'better' is purely up to trade-offs made at a design level or their given market. A reasonable guess continues to be that nVIDIA seems like a TSMC company, AMD a GF company. I don't think either will have major problems once the process grows out of it's infancy. GF will get yields up after mastering voltage thresholds and what-not for different designs, and partners using TSMC will learn how to use the orthogonal design style to keep die size in check.
Who gets their first, or how first chips turn out though, is anyone's guess.
09/29/10 05:20:37 PM]
"my speculation has been for a while; Gate-last (TSMC/Intel) will net you a little bit more performance(/watt), but the gate-last (GF) approach looks to be an ideal fit for System-On-a-Chip (ie ARM designs from Qualcomm/Texas Instruments) and foundry customers (like 'ATi' and nVIDIA"
Is it a typo? Then it should say: "gate-first (GF)approach looks to be an ideal fit for System-On-a-Chip".
On the global (uninformed) view, this manufacturing factor seems to be the real bottleneck behind the high prices and general unavailability of watt-efficient CPUs, RAM and the wailing-behind-transistor-tech that (being still big & unefficient by today's standards) sustain the entire world of home, professional or industrial appliances, electronic tools and machines, excepting -of course- motor engines (electric & the rest) and batteries...
So it's indeed good news the 28-32 nm vanguard is coming to fruition.
09/30/10 03:28:25 PM]
Despite the denials, GF is obviously having teething problems with their 32nm process. Such problems are the likeliest explanation for the slippage in the Llano release and the continuing absence of a firmer release date for Bulldozer.
AMD/GF have admitted that their 32nm yields are still not where they have to be.
Meanwhile, Intel has beeen selling 32nm Clarkdale since early 2010 and seems to be politely holding back 32nm Sandy Bridge until january 2, 2011.
Enthusiasts don't care about GF vs GL. They want 32nm AMD processors ASAP, to compete with the 32nm Intel CPU lineup.
The way things are constantly slipping for AMD, by the time Bulldozer is readily available, Intel will be about ready to launch 22nm CPUs
10/01/10 07:40:09 AM]
A virtual Doupoly where one is (not)having manufacturing problems, but still their renewal operation lines are somewhat being posponed, together with the other actor "politely holding back" their matching coming tech...
All this months after their few main product buyers allegedly asked for delay time so they could sell their existing stock.
It's just either real problems/build uncertanties or telepathic concord behaviour for the short-term benefit of a few...
///EU inspectors should be attentively watching this.
10/01/10 11:03:18 AM]
Add your Comment
Enter your username and e-mail address. Password will be sent to you.