SemiWiki – Page 787 – The Open Forum for Semiconductor Professionals

RVN! 26 Banner revised (800 x 100 px) (600 x 100 px)

October 7, 2014

Bouncing Between iPhone and Android

Bouncing Between iPhone and Android
by mbriggs on 10-07-2014 at 6:00 pm
Categories: Mobile

I’ve now been the proud owner of a Motorola Droid, HTC Incredible, iPhone 5, Samsung S4 and iPhone 6. I just can’t decide if I’m in the iPhone or Android camp!

If you don’t feel like reading the entire article, the iPhone 6 is a superior phone experience (judgement made on my 3 weeks of ownership). If you are on the fence, buy one. You might even get sucked in by the Verizon Edge plan, so it becomes moderately cost effective to upgrade every 12-18 mo, vs 2 years.

Verizon Edge is a way to purchase a device. The full retail price is broken down into 20 easy monthly payments, which appear on your bill. After 30 days, you are then able to upgrade to a new device once 60% of your current Edge device is paid off.

Also Read: Samsung Profits Fall 60%

The iPhone 6 hasn’t been perfect. The second time I used the Starbucks app it wouldn’t let me hit Pay. The people behind me in line were getting very annoyed as I fumbled with my phone. Maybe phones, circa 2020 will be glitch free, but I don’t think we are there, yet.

Everything else works great. It’s fast, responsive and intuitive. The camera takes good low light pictures as advertised. I miss the Aviate launcher, but I’ll get over it.

I switched from the HTC Incredible (which I liked far better than my Droid and Samsung S4) to the iPhone 5. This transition was difficult and motivated me to write the article 8 Reasons Why I Hate My iPhone 5 in October of 2012. I came to like the iPhone 5, but didn’t love it, which is why I bounced back to the S4. Let’s see if the 8 reasons are still valid:

Why I Initially Hated My iPhone 5

1. Voice recognition. Siri is much better now. When Dan and I have our “ask the phone a question” contest while dog walking, even his iPhone 5s is roughly equivalent to the Samsung S4.

2. Camera. The camera is startlingly good in the iPhone 6, enough so that my Nikon D70 is gathering dust.

3. Data Plan. I still hate Verizon, but what can I say? I threaten to move to T-Mobile on a monthly basis, but don’t want to hear the squawking from my wife if the transition doesn’t go smoothly.

4. Photo Sharing. I’ve become an avid Dropbox user. Did you know that Dropbox will automatically upload your photos to the cloud, for free, from Android or iPhone?

5. Maps. Google maps is available on the iPhone. It’s a constant companion of mine. I don’t know how I found my way, anywhere, before it was available. It’s a little bothersome that I can’t tell Siri to navigate somewhere via google maps, but c’est la vie.

6. Cost. See #3. Verizon successfully squeezes the maximum dollars out of me per month.

7. App Installation: Works fine, and I appreciate the vetting that Apple does with the App store. Most of the apps work well. The apps in the Google Play store are more of a crap shoot. I find I need to find the apps with a Google Search because search in the app store is lame.

8. Password: I’ve gotten used to typing my “secure” password. Does everyone agree that the Ukranians probably have ALL our information, including bank accounts, passwords, and SSNs? Just make sure your bank won’t do a transfer to a new location without your approval.

In the large, I’d say the 8 reasons are no longer valid. Phew!

Is it Android or Samsung that breaks?
I don’t know if my usage patterns tend to confuse Android, or I’ve been unlucky and ran into strange hardware errors. You’d think that podcasts, music via iTunes, google maps, and lots of email would be a fairly common use case. On the Samsung S4 my GPS hiking app, as well as some Wifi hotspots would really mess things up, and more often than not, require a restart. Google maps seemed to work well 2 out of 3 times. It seemed to get worse the longer I had the phone. One time I did a factory reset and reloaded all the apps, but it didn’t seem to help much.

My 20 year old son had a similar experience with his Samsung S3. For the first couple of months it was a joy. After that, lots of restarts.

The Earpods Make a Difference
I listen to podcasts while walking and I like the Apple earpods, vs in ear, buds. The Samsung phones only support the in ear buds, which I really dislike. The earpods kinda/sorta work on the Samsung phone, but only enough to keep you using them, albeit with annoyance. Having podcasts pause and start again properly is a relief. Having a decent microphone as part of the earpods is welcome.

Summary
So far, so good with the iPhone 6. The only Android features I miss are Google Now and the ability to have a custom home screen. I hope I don’t decide I want an iPhone 6+ as I don’t think it will fit easily in my pocket.

October 7, 2014

WTL Leverage FDSOI to Achieve Both Low Power AND High Speed

WTL Leverage FDSOI to Achieve Both Low Power AND High Speed
by Eric Esteve on 10-07-2014 at 11:46 am
Categories: Foundries, STMicroelectronics

In fact, this is the title of a presentation given by Pete Foley during FD-SOI Forum 2014 held in Shanghai, a couple of weeks ago. What is nice with clever people like Pete Foley is that they get the point, and get it quickly. Getting the point is to insert AND in capital in the title, as using FD-SOI technology allows to benefit from low-power (thanks to the SOI technology) and performance, thanks to Forward Body Bias (FBB). I didn’t know who is Pete Foley and a quick look at Linkedin give me a hint: Pete’s DNA is innovation. In 1983 Pete was the “Third member of the original Apple VLSI team”, in 1987 he “Led the Newton hardware architecture development before it was redefined from a slate into a “PDA” form factor”… and so on. He is since 2009 the CEO of Waves Semiconductor, a “fabless semiconductor startup that is commercializing a programmable solution addressing power, concurrency, design time, design cost, and deep submicron challenges facing the semiconductor market.” This article is inspired from his presentation that you can find it here FD-SOI Forum 2014 along with other presentation from ST-Microelectronic, Synopsys, IBS and more.

Waves Semiconductor is developing real products based on Waves Threshold Logic (WTL). If we want to make it short, WTL is not based on standard Boolean logic, is asynchronous and “unifies Data and Control into a single “Logically Determined” paradigm. Simple words are not enough, that’s why the above picture should help! As you can see, there is no clock, no register, just DATA and NULL pipeline stages, and Acknowledge signals generated by the data. Alternating DATA/NULL provide data behavior, Pete Foley claim that WTL is “Data Driven Completion”. Asynchronous means that no clock is used, with several immediate consequences: no clock means no glitch (lower risk), no necessity to provision time for set-up and hold or for clock skew and budget for worst case PVT (better performance), and no need for register and CDM (40% real estate saving). Last but not least, much less dynamic power, as the pipeline only consumes power when the data change. So WTL is fast, fully standard CMOS compatible, but is also very low-power, especially when used on FD-SOI technology.

If you remember this blog in Semiwiki, a device designed on FD-SOI will offer better performance than the same design on Bulk Silicon (assuming the same node) if you use Forward Body Bias (FBB), or will exhibit much lower power consumption… but not at the same time. According with Pete Foley, WTL can use both types of transistor simultaneously to achieve speed of LVT devices under FBB but leakage of RVT devices under Reverse BB (RBB): the best of both worlds! You could argue that WTL is a nice but theoretical concept… In fact Waves Semi is designing products based on WTL like Azure pictured below:

If you are scared about WTL, and prefer to discuss about an existing SoC product initially designed on 32/28 LP Bulk and ported in FD-SOI 28nm, this example of an ARM Cortex A9 will be appropriate. You can find it in the presentation “The FD-SOI Technology for Energy Efficient SoCs” from Giorgio Cesana. The total power consumption, on the Y axis, is increasing with the CPU frequency (as expected), but the power clearly grow faster and get higher with the A9 on Bulk (in vertical blue) than the same design on FD-SOI (in vertical yellow). And you can see that using the FD-SOI Forward Body Bias effect helps further decreasing power consumption (in vertical green). If we try to quantify the gain in power, we have to look at the horizontal yellow and green horizontal curves. This gives a 50% lower power consumption for the A9 on FD-SOI FBB and 40% on FD-SOI, both compared with Bulk.

As a conclusion, just watch this slide (also from ST): “FD-SOI, Ecosystem available in the whole chain” extracted from this presentation. SOI wafers can be found at multiple sources, design solutions are enabled by EDA and IP vendors (or from IP porting) and the designed SoC can be fabed in production at ST facility and starting in 2015 at Samsung…

From Eric Esteve from IPNEST

Samsung Profits Fall 60%

Samsung Profits Fall 60%
by Paul McLellan on 10-07-2014 at 11:32 am
Categories: Mobile

Samsung’s profits fell to $3.8B which is 60% down on where they were a year ago. They said this is primarily due to a shortfall in smartphone sales. Samsung’s problem is not that it competes with Apple, their phones seem close enough that the Galaxy line has been doing very well in places like the US. It’s problem is that a lot of its market share in in China where it has to compete with Xiaomi (the market leader even though you’ve probably never even seen a Xiaomi phone), Lenovo, Huawei and others.

The Android operating system is great in some ways for smartphone makers. Apple spends billions on developing iOS. Samsung and others just get Android for free. Google is spending the billions on development, recouping it (or not, who knows?) through search advertising, which despite its broad product portfolio is the only place it makes real money. But it is an Achilles’ heel too, since it means that a Samsung smartphone just isn’t well differentiated. Samsung has been working on their own operating system, Tizen, that I expected they would aggressively use to set them apart from all their Android rivals but every month it seems to slip and get less significant. The first Samsung Gear watch used Tizen but now it is Android, for example. I guessed wrongly that Samsung was big enough that they could use their muscle to establish a third ecosystem of applications around Tizen but now I don’t believe that is going to happen.

Also read: Intel Invades China Mobile Market!

But here is something that is even more amazing to think about. Maybe Samsung will be driven from the smartphone market completely. Look at what happened to Palm. Blackberry. Nokia. Motorola. As Ernest Hemingway said, when asked how he went bankrupt, “Gradually, and then suddenly.” Could it really happen to Samsung too? First their sales fall, their profits go away, their competition gets stronger. The carriers desert them for cheaper suppliers. I don’t know the latest numbers for Q3 yet, since the quarter is not over (well it is, but nothing much has been announced) but here was my lookback on Q2 earlier this summer. Samsung is still #1 but then at one point so were both Motorola and then Nokia.

What about Apple? Its newest phone is underwhelming, adding features that its competition has had for years. But they are an aspirational brand and they are so wildly profitable that they can do well even on declining and not all that large market share. They are currently #2 (behind Samsung). Someone in a comment on one of my blogs said that eventually they would be forced out of the market, just look at PCs. So I looked up the numbers. Apparently the Mac product line makes 40% of all the profit in the PC market, despite relatively puny market share. But they certainly have the hipster demographic, just walk into any coffee shop. It is just in corporate where they are weak, outside the graphic design department. Most internet development is done on Mac, music, graphics and so on.

Xiaomi is certainly one company to watch. They only sell in a handful of countries. They release their phones in batches and only sell them online but they sell out in a few minutes. They are the #1 brand in China already by unit volume. They sell their phones for pretty much cost, with the plan to make it up on services that they have not yet launched. So they could yet crash and burn since they are making a loss on each phone but making it up in volume, as the old joke goes.

But back to Samsung. The explosive growth phase of high end smartphones is over. The growth is in the low end, where Samsung do have product. The challenge, though, is that it is hard to replace high-end smartphones. After a house and a car, a TV and a top-of-the-line smartphone are probably the most expensive possessions. Even if Samsung is successful in the internet of things (IoT) it is not clear it will generate anything like the same revenue or the same semiconductor demand. I have always said the same thing about Intel in mobile: they can’t afford to be successful in mobile, it would crater their margins. Samsung is in a similar place.

Intel Invades China Mobile Market!

Intel Invades China Mobile Market!
by Daniel Nenni on 10-07-2014 at 7:00 am
Categories: Foundries, Intel Foundry, Mobile

As you may have heard, Intel is making some interesting moves into the China mobile market. At first it had me a bit puzzled but I had some lengthy discussions about it during ARM TechCon last week so I’m ready to start writing. Spoiler alert: This will be pro Intel so please sit down and take some deep breaths before reading.

The first interesting news was the Intel investment in Rockchip back in May of this year. We have written quite a bit about Rockchip on SemiWiki and I have heard even more during private conversations. Great company but they face stiff competition from a dozen different SoC companies big and small including Spreadtrum. If not for the Intel deal my understanding is that Rockchip would have faced an asset sale. With Intel’s backing however, I see differentiation and a huge upside, absolutely.

“The strategic agreement with Rockchip is an example of Intel’s commitment to take pragmatic and different approaches to grow our presence in the global mobile market by more quickly delivering a broader portfolio of Intel architecture and communications technology solutions.” Intel’s CEO Brian Krzanich.

The Spreadtrum deal is even more interesting. Intel took an equity position in the holding company (Tsinghua Unigroup) which holds Spreadtrum Communications and RDA Microelectronics.

“China is now the largest consumption market for smartphones and has the largest number of Internet users in the world,” said Brian Krzanich, Intel CEO. “These agreements with Tsinghua Unigroup underscore Intel’s 29-year-long history of investing in and working in China. This partnership will also enhance our ability to support a wider range of mobile customers in China and the rest of the world by more quickly delivering a broader portfolio of Intel architecture and communications technology solutions.”

Intel is reportedly investing $1.5 billion into Tsinghua Unigroup for a 20% stake. Interestingly enough Spreadtrum and RDA were taken private by Tsinghua just last year for $1.78 billion so the math doesn’t really work here. Dealing with the Chinese Government has always perplexed me so I’m interested to see how this plays out. And what about the Rockchip relationship? Rockchip and Spreadtrum are fierce competitors. Hopefully I can get more information during my next trip to Shanghai. Face-2-face communication is much more reliable than Googling around and making things up.

Also read: The Apple Samsung TSMC Intel 14nm Mashup!

A while back I gave Intel Custom Foundry some very good advice: Invest in the fabless semiconductor ecosystem because unless Intel is an integral (non-threatening) part of the ecosystem they will not succeed. One suggestion I made was to buy fabless companies to fill Intel fabs (I believe I mentioned Mediatek and Broadcom). In this regards the Spreadtrum investment is brilliant considering the China market has the biggest upside for Intel Custom Foundry.

Granted, TSMC manufactures the majority of mobile chips today and the Chinese Government does have a significant stake in SMIC, but based on my experience with SMIC they have no chance at 28nm, 20nm, or 14nm. From what I was told, SMIC licensed the IBM gate-first HKMG 28nm process only to change it to gate-last to better accommodate TSMC designs for second sourcing. This failed miserably and now SMIC is working on 28nm FD-SOI which means they will have to change back to gate-first. So the gates of China are open to Intel Custom Foundry for mobile devices and I hope they succeed for the greater good of the fabless semiconductor industry.

My next trip to Asia is at the end of this month so please let’s discuss this in the comment section so I can get clarification during my trip:

[LIST=1]

Why did Intel really make these investments?

Is Intel an IP company now?

Will Intel manufacture these chips in the US?

Will Intel manufacture these chips in their 300mm China Fab?

Can Intel succeed in the China mobile market?

Silicon Does NOT Lie!

Silicon Does NOT Lie!
by Daniel Nenni on 10-06-2014 at 7:00 am
Categories: General

In the 30+ years that I have worked in Silicon Valley I have seen many great products fail and even more mediocre products succeed, the difference being how the companies communicate to the outside world. In the semiconductor industry, presenting the value proposition of your company or product is under even more scrutiny now that independent benchmarking and product tear downs have become prevalent. So why is it some of our marketing brethren still communicate like it was 1999?

Do you remember Osborne Computer and the “Osborne Effect”? Osborne computer pre-announced a new product only to have it delayed for more than a year leading the company to bankruptcy. People like myself did not buy the original Osbourne1 because we wanted the newer version that was pre-announced. Instead, I bought my first IBM PC and the rest is history. You should also be aware of the term “self-defeating prophecy”:

A self-defeating prophecy can be the result of rebellion to the prediction. If the audience of a prediction has an interest in seeing it falsified, and its fulfillment depends on their actions or inaction, their actions upon hearing it will make the prediction less plausible.

So yes, no matter how good your product or service is, how you communicate to the outside world is critical. And now with New Media (Twitter, Blogospehere, SemiWiki, etc…) and the resulting transparency the old guard semiconductor communications strategies just do not work like they used to. Invite me over for lunch sometime and I will share my SemiWiki New Media experience with your marketing communications people. It’s a 30 minute presentation with another 30 minutes for Q&A.

One of the services we offer to companies that subscribe to SemiWiki is strategic marketing communication. Pre-brief us on your communication plan or your product announcement and we will give you some very blunt feedback based on SemiWiki big data and our collective professional experiences. Having a pair of outside eyes looking in is ALWAYS a good idea whether you act on the advice or not, absolutely. Remember, we are not reporters, journalists, editors, or parrots. We are semiconductor professionals who trade on our reputations.

The infamous Altera slide is my new favorite example of bad marketing communication. Leveraging their intimate knowledge of the TSMC 20nm process, someone created this slide to not only defame Xilinx but to discredit TSMC and the fabless semiconductor ecosystem. Just my opinion of course but when the Xilinx and Altera FinFET based silicon becomes available there will be a tear down and the truth will be told. My bet is that a silicon correlated version of this slide will be an embarrassment to not only Altera but Intel as well. If I was pre-briefed on this slide no way would it have seen the light of day.

On another note, it amuses me when IDMs compare themselves to a fabless company or a foundry or even an IP company using revenues, employees, CAPEX, or whatever number suits their purpose. When in fact you need to take a company like Qualcomm, add in ARM, TSMC, Synopsys, Cadence, Mentor, and a whole host of other collaborators. There are hundreds of thousands of us in the fabless semiconductor ecosystem and we spend many billions of dollars every year on R&D and CAPEX. If you think David and Goliath is an interesting story, try hundreds of thousands of Davids and one Goliath.

10nm, the View from IBM

10nm, the View from IBM
by Paul McLellan on 10-05-2014 at 7:01 am
Categories: Cadence, EDA

On the Cadence booth at DAC, Lars Liebmann of IBM presented on the challenges of 10nm. As he put it, how the lithography folks are keeping things very interesting for the EDA tool development engineers. Although 14nm/16nm hasn’t yet ramped into HVM, the advanced work for tools and IP has all moved to 10nm. Although Lars gave the presentation, it is also credited to Vassilios Gerousis, Mike Zhang and Paul Gutwin of Cadence, Geng Han of IBM and Brian Cline of ARM. I talked to Lars this week and got a copy of the presentation but there is a video of the whole thing on the Cadence website. See below for a link.

I’m going to take it as given that you know all about double patterning, and I’m going to take it as given that EUV is not going to be available at 10nm. We have to live with 193nm light.

At 10nm, the first change is that even though we are still in the double patterning resolution window, we need to add a third color to resolve odd cycles, carefully optimized design rules are not enough to prevent cell to cell color interactions. When two cells are placed next to each other that conflict, there are three approaches to fixing it in the placer: flip the offending cell so it no longer offends, spread the offending cells further apart, or swap the offending color (since the power lines are one color that we can swap from cell to cell, even with 3 colors there is actually only a single choice of what colors to swap).

Vias are another problem. Even at 14nm we can’t put two vias next to each other in the obvious way, they are too close. The nearest alternatives are diagonally adjacent. But at 10nm the area of interaction is so large that not just neighboring vias but the two vias over need to be colored appropriately. This cannot be done after the routing has been completed, since it would often turn out to be impossible. Instead explicit odd cycle via prevention is required in the router.

The type of double patterning used at 14nm is known as LELE for litho-etch-litho-etch. But there are resolution limits to this limited by overlay and also dielectric breakdown. So for 10nm we will need what is called either SIT (sidewall image transfer) or SADP (self-aligned double patterning). In this a mandrel is created, sidewalls are created on the edge of the mandrel, and then a cut/block mask is used to trim the design and remove unwanted parts. But arbitrary layout can lead to errors (non-manufacturability) on the block mask. A whole new lot of line end stagger rules come into play. The block mask patterning becomes the resolution limiter.

The solution is to go to a sea of wires with a highly constrained cut mask. This means lengthening many wires (which will affect performance), creating dummy floating wires and worrying about line-end stagger control.

All of this has been implemented in Cadence’s Encounter and with a compliant ARM 9-track standard cell library.

The video of Lars’s presentation is here. It is the presentation at 11am on Tuesday.

The Apple Samsung TSMC Intel 14nm Mashup!

The Apple Samsung TSMC Intel 14nm Mashup!
by Daniel Nenni on 10-04-2014 at 1:00 pm
Categories: Foundries, Intel Foundry, Mobile, TSMC

One of the strengths of the fabless semiconductor ecosystem is competition since it keeps innovation high and prices low. One of the challenges of fostering competition is that you have to make good on a threat of using a competing product during a pricing negotiation. Well, in my opinion, for the next version of the iPhone, Apple did just that. Apple put Samsung and TSMC against each other and as a result will use both Samsung 14nm and the better performing TSMC 16nm FF+ for the 2015 iProducts. Since Samsung is a quarter or two ahead of TSMC on FinFETs, Samsung will get the iPhone business in Q3 2015 and TSMC will get the iPad and maybe a MAC Book in Q4 2015. Qualcomm, NVIDIA, AMD, Broadcom, and the other fabless heavyweights will follow suit. It’s all about wafer price negotiations and that is what keeps us strong.

Also Read:Where will Apple Manufacture the next iPhone Brain?

Common wisdom suggests that Apple would not do business with Intel or Samsung as they both compete with Apple on some level. Samsung aggressively sells competing phones/tablets and Intel gives free SOCs to companies that compete with Apple. But when you are buying billions of dollars of wafers, price tends to trump all. Now let’s talk about the rumors overheard at ARM TechCon last week:

Out of the total volume of A8 chips, Samsung is producing around 30 percent, while TSMC is making 70 percent, sources familiar with the matter said.

The first rumor is that Samsung will get 30-40% of the Apple A8 business. Apple will use the same A8 SoC for the iPhone 6 and iPad products. I’m hoping the A8 will be clocked up for the iPad because I know it can go faster than 1.4GHz! Bottom line: Samsung is NOT supplying 20nm A8 parts to Apple. Show me an iPhone or iPad tear down that proves otherwise and a new iPad is yours. Also according to ZDNet Korea’s Cho Mu-hyun:

Kim Ki-nam, president of the Korean electronic giant’s semiconductor business and head of System LSI business, told reporters at Samsung’s headquarters in Seoul that once the company begins to supply Apple with chips using its latest technology, profits “will improve positively”.

Samsung is expected to start producing application processors (APs) for clients such as Apple, Qualcomm, and AMD, using its 14-nanometre process around the end of the year. Kim declined to comment on when Samsung will start mass producing said chips for clients.

The second rumor is that Samsung won the Apple business for 14nm. As I mentioned above, Apple will use both Samsung and TSMC for the next generation of iProducts in 2015 so this is a half-truth. To me a half-truth is also a half-lie and I have a serious problem with that especially coming from a publicly traded company. Samsung should publicly clarify that it is NOT supplying 20nm wafers to Apple and they are NOT the only vendor supplying Apple FinFET wafers in 2015. Transparency inspires trust, right?

In regards to Intel Custom Foundry, being stuck between Samsung and TSMC in a wafer price war is no way to start your day. Given that TSMC 10nm is on track with Intel Foundry 10nm (remember Intel Foundry has yet to ship production 14nm) and Samsung is a close third, 10nm will be another serious pricing challenge. And let’s not forget that GOLBALFOUNDRIES and IBM are also in the10nm hunt. Competition is for the greater good of the fabless semiconductor ecosystem, absolutely.

Key Collaboration to Enable Designs at Advanced Nodes

Key Collaboration to Enable Designs at Advanced Nodes
by Pawan Fangaria on 10-03-2014 at 10:00 pm
Categories: Cadence, EDA, Foundries, GlobalFoundries

In the semiconductor ecosystem, several partners (or better to say stakeholders) join together in the overall value chain to finally output the most coveted chip, err I should say SoC these days. It becomes really interesting when we start analyzing the real value added by each of them, none appears to be less. Well, then to whom can we give the trophy if we have one to offer? Okay, then ask a simple question, what’s it without which the design couldn’t have proceeded in the first place in real physical terms and you get the instant answer – it’s PDK. The Process Design Kits may be lesser known in the world of ESL and RTL, but ask a back-end designer and she will tell you can’t do without PDK no matter how powerful EDA tools you have, and it’s utmost important for your PDK to be of highest quality. But then the EDA tools need to drive into the design what PDK has in the most effective manner.

It was a very pleasant moment when I came across a DACpresentation offered by Maq Mannan of GLOBALFOUNDRIES. Although it was a recorded presentation, it was pleasure listening to Maq; it revived my old memories of Cadence. Maq is a veteran in PDK domain; he founded DSM Technologies (which dealt in PDK development automation systems) in late 1990 – early 2000 which got acquired by Cadenceand that formed the PDK group at Cadence headed by Maq. Now Maq is at GLOBALFOUNDRIES, leading and delivering the best PDKs which are of highest quality, flexible and easy-to-use, enabling GF’s leading edge process technologies into SoC designs.

Starting from say 130nm to 20nm and below, added with FinFETs, the number of design rules have increased tremendously (approximately doubled with each node change), and along with it increased number of operations with each rule in the same or higher proportion; multi-patterning added further major complexities. Designers cannot keep track of rules while working. Keeping these in mind, the PDKs from GF have in-built, easy-to-use methods for designers to carry out their work without any loss of productivity. Recently GF and Samsungpartnered for PDKs at 14nm (14LPP and 14LPE) which can go to both the fabs with same design rules, models, DRM etc. thus improving productivity. The 14LPE V1.0 PDK is already released. As the complexities have increased with rapidly advancing technologies, it’s essential for fabs and EDA vendors to collaborate and leverage their capabilities to keep the tools in sync with technology to drive the designs in best possible manner. GF has built strong collaboration with multiple EDA and IP partners. Let’s look at what’s cooking up between GF and Cadence.

GF has major collaboration with Cadence with most of their customers using Virtuoso and Cadence libraries which is very significant and critical to their success. The area is spread across from digital routers to device models, simulation, extraction and verification to DFM kits, and the entire infrastructure relating to those is covered into PDK.

Cadence introduced IPVS which provides in-memory verification (with real-time DRC on edits) of layout within the Virtuoso layout design environment eliminating time consuming StreamIn/Out which was earlier needed by Standalone PVS. This capability is very significant in boosting designers’ productivity. GF supports PVS decks for virtually all technology nodes. This is a capability close to my heart as I always aspired for it in Virtuoso while I managed the physical design part of Virtuoso at Cadence.

Read one my older article on this – 28nm Layout Needs Signoff Quality at Design Time.

At advanced nodes (28nm and below), layout dependent effects (LDE) becomes significant causing significant deviation in device characteristics from their normal behaviour. Cadence tools along with wide range of support in GF’s PDKs to take these effects into consideration provides a versatile layout aware design flow which enables LDE estimation at schematic, detects LDE problems in layout, checks against constraints provided by circuit designers and verifies using LVS the final extracted layout for both placement as well as interconnect effects.

GF also provides decomposition options which are very critical for customers at advanced nodes, a key requirement for multi-patterning.

The designs can be done in colored layers (M1_E1, M1_E2) which is appropriate for standard cells, colorless layer (M1) appropriate for p-cell design and P&R which can be decomposed later by specific tool in PDK or mixed (hybrid) way. GF offers all three methodologies at 20nm; designs can be done in any of these combinations. For 14nm, they are working with Samsung to align on a single approach.

GF provides complete modeling environment with versatile Spice model library containing MOSFETs (RVT, HVT, LVT, SRAM, EG, ZG) including any special memories and devices and Non_FETS (Resistors, Capacitors, Diodes, Fuse and others). Models for variation (MC, Mismatch, and Analog), aging (Reliability) and stress (DSL, STI, WPE etc.) are accounted for.

It’s a great presentation to go through and know about PDKs and Cadence – GF collaboration in more detail. It’s freely available here. Click the link against “Collaboration Key to Enablement at Advanced Nodes”.

GlobalFoundries and Samsung at ARM

GlobalFoundries and Samsung at ARM
by Paul McLellan on 10-03-2014 at 3:01 pm
Categories: Arm, Foundries, GlobalFoundries

It was back in April that GlobalFoundries and Samsung announced that GF would license Samsung’s 14nm process to run in their Fab8 in upstate New York. Since then there has not really been any news and of course those of us that follow the foundry industry wondered to what extent there was real substance to the agreement or if it was just a feel-good press release. Well, this week a lot of information was presented. It was at ARM TechCon this week. There was a panel with Kelvin Low of Samsung, Shubhankar Basu of GlobalFoundries and Wolfgang Helfricht of ARM. Kelvin went first and emphasized that this is true multisourcing with true 14nm compatibility and it covers both 14LPE (E stands for ‘early’, since it is the early access version of the process) and 14KPP (P stands for ‘plus’, apparently, since I guess ‘late’ wouldn’t sound good). The wafer demand is, of course, largely driven by mobile but also by high-end compute functions in server farms. 14nm helps to bridge the performance gap that exists due to the limited power budgets available. FinFETs are especially good at very low voltages, and lowering the voltage is always good since it is squared in the dynamic power equation. The two processes LPE and LPP have the same design rules for fast migration. LPP has more performance and even lower power. Kelvin claims that the process enables designs that are 10% smaller than “other foundries” (no prizes for guessing who that might be) due to aggressive gate pitch, smallest SRAM and innovative layout. There have been 30 test chip tapeouts since 2012, and multiple product tapeouts completed (the first last year). Shubhanker revealed that fab8 is on-track with 100% physical module spec matching demonstrated. Currently matches fin, gate, eSiGe, eSiP, RMG, MoL. There is steady progress on SRAM yield and volume ramp in 2015. Multiple product and testchips have taped out. PDKs are available. MPW shuttles are available. IP support, reference flows, tech files. Yes, you can do a design. Wolfgang, from the Artisan part of ARM, talked about the availability of standard cell and other libraries. At 14nm there are new sources of variation and ARM have Artisan Signoff Architect. There are also complex EM and IR limitations which are addressed with Artisan Power Grid Architect. There are multiple standard cell libraries:

SC10MCP with 10.5 tracks for 14LPP
SC9MCP with 9 tracks for 14LPP
SC9MC also with 9 tracks and fast availability for 14LPE.

All products support 0.7V, 0.8V and 0.9V operation. There are memories with multi-voltage peripheries. The 14LPP GPIO system is at 1.8V with a single programmable I/O cell, and compatibile with mobile DDR. High volume designs are in progress using these libraries. Serdes performance is in the 25GHz range. So there you have it. This is true multisource manufacturing. It covers 4 of the anticipated 7 fabs that will run a 14nm process (TSMC has the other 3 but it calls the process 16nm), two in Samsung Korea, Samsung Austin and GF in New York, so there is assurance of supply. The IP ecosystem is portable between all the fabs. See also GlobalFoundries Gets a 14nm Process
More articles by Paul McLellan…

October 3, 2014

A de-parallel universe for Windows 10

A de-parallel universe for Windows 10
by Don Dingee on 10-03-2014 at 7:00 am
Categories: Foundries, Intel Foundry, Mobile

It was CES 2011 when Steve Ballmer sweatered up and pitched the coming universe according to Microsoft, where the same Windows base would run on everything – PC, phone, tablet, and game console. Getting from that visionary statement to Windows 10 hasn’t been a smooth ride. Continue reading “A de-parallel universe for Windows 10”