Friday, February 21, 2014

Movidius works with Google to bring visual awareness to next wave of intelligent mobile apps

SAN MATEO, USA: Movidius, the leader in vision processor technology for mobile and portable devices, announced that it is working with Google to create intelligent vision-based applications beyond the capability of today’s mobile devices.

As part of the agreement, Movidius’ Myriad 1 vision processor platform will power Project Tango, an Android-based smartphone product/development kit within Google’s Advanced Technology and Projects (ATAP) group.

Myriad 1 is the first generation of the Movidius Vision Processor Platform, a new ultralow power, high-performance and programmable architecture of computational chips, software and development tools that enables a range of devices to intelligently understand and contextualize their surroundings.

From smartphone, tablets, wearable and even robotics – Movidius can deliver up to 10 times more flexibility, processing speed and power efficiencies compared to traditional processors.

“Project Tango strives to give mobile devices a human-like understanding of space and motion through advanced sensor fusion and computer vision, enabling new and enhanced types of user experiences – including 3D scanning, indoor navigation and immersive gaming,” said Johnny Lee, Technical Program lead, Advanced Technology & Projects at Google.
“Movidius has provided a key component towards enabling access to these features in a small mobile platform with a chip designed with visual sensing and battery life in mind. We look forward to continuing our collaboration with Movidius as these new applications evolve and new device designs come to market.”

By mapping space and motion in real time with detailed precision and accuracy, Myriad 1 aims to mirror human vision with a newfound level of depth, clarity and realism on mobile and portable connected devices. Myriad 1 will serve as the architecture for application developers and device designers looking to power the next wave of intelligent applications that leverage mobile sight and perception for an engaging user experience.

“Google has paved the future direction for smart mobile vision systems and we’re excited to be working with a company that shares our vision to usher in the next wave of applications that fundamentally alter how a mobile device is used to experience the world around us,” said Remi El-Ouazzane CEO of Movidius. “Project Tango is truly a groundbreaking platform, and we look forward to seeing the innovation the developer community achieves with unprecedented access to the Movidius Vision Processor.”

Thursday, June 20, 2013

ASML, KLA-Tencor, and Lam Research to benefit from 450mm wafer transition

Dr. Robert N. Castellano, The Information Network

USA: The semiconductor industry has been increasing the size of wafers about every 10 years. In 2000, semiconductor device manufacturers migrated from 200 mm substrates to 300 mm substrates due to technological advances and manufacturing cost advantages.

300 mm wafers provide more than 2.25 times as many die per wafer, and offer significant economies of scale in the manufacturing process. Approximately 96 percent of 2012 net orders for wafer fabrication equipment were for 300 mm manufacturing systems. The next planned wafer size increase is expected to be up to 450 mm substrates.

Increases in wafer sizes have been a natural evolution and manufacturing efficiency improvement for the semiconductor industry for decades, as shown in the figure below.
The table below shows the impact of the conversion from 300mm to 450mm technology on several processing parameters.
Shown in the figure below, this illustrates the special dual “S-curve” timing required when a new wafer generation is being introduced, which was modeled after the experience with the 300 mm wafer generation ramp around 2000.
The ITRS has proposed a revision in the timing targets for the 450 mm generation semiconductor manufacturing and foundry pilot lines, which are now delayed about 2 years (2015–16, versus the previous 2013-14 target made by the ITRS in 2009).

By 2015-16, Intel Corp (INTC), Globalfoundries Inc., Samsung Electronics Co. Ltd. and Taiwan Semiconductor Manufacturing Co. Ltd. (TSM) will have 450mm wafer fabs constructed and equipment installed.  While these are only a handful of fabs so far, they represented 34% of the revenues of the entire semiconductor industry in 2012, as shown in the table below.
What’s in store for equipment suppliers?
Market share is a key determinant in what semiconductor equipment supplier will reap the benefits of sales to customers that represent a third of the total industry.  And this is important.  A vendor will not have a 450mm position unless it has a 300mm position.  In other words, the slate will not be clean going from 300mm to 450mm as far as purchases are determined.

Semiconductor manufacturers will purchase 450mm equipment from the same equipment supplier that sold them 3000mm equipment.  For an equipment company to become the supplier of record to a customer, it needs to execute well, providing best-of-breed equipment and technology.  Unless that company self-destructs, the execution they focused on to sell 300mm equipment will carry over to 450mm.

So who’s to gain in this transition?

ASML (ASML) is an easy one. Besides the fact that they dominate the lithography market with an 84 percent share (see chart below), they also dominate the EUV (extended UV) market with a 100 percent share. ASML announced they would port their EUV to 450mm only.  Intel invested $4.1 billion in ASML, as part of an effort to raise the overall confidence level in EUV and the 450mm wafer transition. TSMC and Samsung also recently invested in the Dutch-based lithography giant for similar reasons.
KLA Tencor (KLAC) dominates the process control market with tools for semiconductor inspection and metrology (measurement) and will be another winner in the 450mm transition. KLA-Tencor holds a 50 percent market share, and that share has been growing worldwide, as shown in the chart below.
Hitachi High-Tec’s process control revenue grew 16.6 percent in 2012 and will be a winner in the 450mm transition.  Hitachi purchased SII Technologies in 2013, which recorder revenues of $90 million in 2012, further escalating Hitachi’s revenues in 2013.

Applied Materials (AMAT) made claim that they gained share against KLAC in 2012, but this is not the case according to our analysis.  Applied has been hyping its process control sector of late, perhaps in response, according to rumors mind you, to the fact the now president, Gary Dickerson, was turned down for the post of CEO of KLAC when he worked there, and then resigned to become CEO of Varian Semiconductor, before being hired by Applied last year.

Based on analyzing the semiconductor equipment market for 27 years, I see erosion in Applied’s execution.  Many of the top executives left the company last year and the vacuum was filled by ex-Varian Semiconductor employees from their misguided $4.7 billion acquisition of the company in 2011.

Lam Research (LRCX) dominates the plasma etch market with nearly a 50 percent share, as shown in the chart below. If we go back to 2000, Applied Materials had led the market with a 36.6 percent share and Lam held only a 26.3 percent share. In the following 12 years, Applied’s share dropped to 27.6 percent while Lam’s increased to 48.3 percent.
Lam Research will be another winner in the 450mm transition. The company is a leading capital equipment supplier to the largest memory manufacturers including Samsung, Toshiba and Micron (MU) and also the key supplier to TSMC.

More importantly Lam recently secured a key etch win with Intel, long a stronghold for AMAT.  Lam acquired Novellus last year. Novellus went head-to-head in competition with Applied Materials in the deposition market, and there had been constant speculation that the re-badging of Novellus hardware by Lam will find its way into Intel.  It appears this may now be the case, Lam’s etch and deposition wools will share a common platform, and it will drive strong sales in Intel’s 14nm build-out.

The semiconductor equipment industry is dynamic and things change. For example, in the metal CVD market in 2000 for example, Applied led the market with a 70.6 percent share compared to 17.1 percent for Novellus. In 2011, Novellus claimed a 64 percent share to Applied’s 17.2 percent.

Wednesday, June 19, 2013

ST reveals new technology platform for RF front end of wireless devices

SWITZERLAND: Booming demand for faster wireless broadband connections is calling for increasingly complex circuitry inside devices such as smartphones and tablets.

STMicroelectronics is addressing that demand by perfecting an advanced component process technology specifically optimized to increase performance and reduce the size of the RF (radio frequency) front-end of mobile devices.

In wireless devices, the RF front-end circuit is typically built using individual amplifiers, switches and tuners. As new high-speed standards such as 4G mobile and Wi-Fi (IEEE 802.11ac) use multiple frequency bands to increase data throughput, the latest equipment requires additional front-end circuitry.

While current 3G phones use up to five frequency bands, the 3GPP standards for next-generation 4G LTE support up to 40 bands. Conventional separate components dramatically increase overall size whereas ST's new manufacturing process, known as H9SOI_FEM, allows production of complete integrated front-end modules.

This process is an evolution of the H9SOI Silicon-on-Insulator process; a groundbreaking technology introduced by ST in 2008 and subsequently used by customers to produce more than 400 million RF switches for mobile phones and Wi-Fi applications.

Building on that experience, ST has optimized H9SOI for creating integrated front-end modules, resulting in today's announcement of H9SOI_FEM offering the industry's best figure of merit for antenna switch and antenna tuning devices with Ron x Coff at 207fs. ST has also invested to ensure suitable manufacturing capacity for even the most demanding of customers.

From a commercial point of view, smartphones featuring high-speed multi-band wireless are driving booming demand for RF front-end components, particularly as integrated modules. The number of RF devices in a smartphone is roughly three times the number in an entry-level 2G/3G phone, while smartphone shipments are currently over one billion units annually and growing at around 30 percent according to analysis by Prismark.

Additionally OEMs require suppliers to provide smaller, thinner components with higher power efficiency. ST sees opportunities for discrete components, as well as integrated power-amplifier/switch and power-amplifier/switch/tuner modules based on its new best-in-class H9SOI_FEM process.

"The H9SOI_FEM dedicated process enables our customers to develop state of the art front-end modules that are half the size or smaller compared to today's front-end solutions," said Flavio Benetti, GM of the Mixed Process Division of STMicroelectronics. "Moreover, we have achieved a simplified process flow to enable extremely short overall lead-times and supply flexibility, which are crucial for end customers in this market."

ST is now ready to start working with customers on new designs using H9SOI_FEM. Volume ramp-up is expected by the end of this year.

ZEISS launches photomask repair tool MeRiT HR II

GERMANY: The new generation system MeRiT HR II from ZEISS offers a significantly higher degree of automation and increased flexibility in the repair of transparent and opaque defects.

Any defect geometry on all known as well as new mask types and materials can be processed: Binary, OMOG, HD MoSi and EUV. The technical specifications of the MeRiT® HR II match to the industry roadmap and sharp market requirements.

The redesigned tool includes a sound automation package which enables a new level of automation with optimized repair flow. After a short setup routine the processes are able to seamlessly repair numerous defects without the need for human interaction. A higher throughput due to lower idling time is also offered. In case of any doubt a security interlock consults the operator who can intervene when necessary.

“The significantly improved automation level of the new MeRiT HR II frees up precious operator time. Furthermore the combination of automation and manual repair provides an increased flexibility. Altogether the new system has its finger on the pulse as it includes the most relevant requirements mask shops have for repairing all kinds of photomask types in a very short time.” states Dr. Oliver Kienzle, head of the ZEISS strategic business unit Semiconductor Metrology Systems who develops and sells the MeRiT HR II.

DRAM market grows up; industry’s newfound maturity yields growth amid adversity

USA: It’s said that adversity breeds character—and that certainly seems to be the case for the global market for DRAM (dynamic random access memory). This market has achieved some maturity in the face of daunting challenges, allowing the industry to achieve a balance between supply and demand this year.

After DRAM wafer output peaked in 2008 at 16.4 million 300-millimeter-equivalent wafers, production is expected to decline by 24 percent to 13.0 million this year, according to an IHS DRAM Dynamics Market Brief.

The projected cut will be the second straight year of deliberate downsizing following an 8 percent drop-off last year. This year’s output is expected to be slashed by 5 percent compared to 2012, as shown in the attached figure.

Curtailing DRAM capacity is a positive move for the industry, resulting in a gradual normalization between supply and demand for DRAM. The industry is now believed to be perhaps slightly undersized relative to demand moving forward because of the intentional slash in output, and DRAM pricing can continue to remain firm if production remains slightly behind demand.

DRAM revenue in the first quarter rose to its highest level in nearly two years, thanks to a jump in commodity prices spurred by demand from the server PC and mobile PC segments. Pricing for the bellwether 4-gigabyte DDR3 module rose to $23 in March, up from $16 in December, an unusually large increase.

“The DRAM industry has struggled with major challenges in recent years, including chronic oversupply and slowing demand from its main market, the PC business,” said Mike Howard, senior principal analyst for DRAM and memory at IHS. “This has led to continued weak pricing, financial losses and market revenue declines. However, the DRAM industry has entered a more mature state, enacting structural changes that will allow it to grow even in challenging market conditions.”

DRAM market enters the post-PC era
In one major change, the DRAM market is adjusting to the fact that demand is diversifying away from PCs alone to servers and mobile devices.

Nearly 65 percent of all DRAM bit shipments went to a desktop or laptop 10 years ago, but that figure is less than 50 percent today and will fall further to south of 40 percent by the end of next year.

Meanwhile, servers and mobile gadgets like smartphones and tablets command an increasing share of DRAM bit shipments.

The overall result is that the travails of one segment—like the embattled PC space—won’t be able to disrupt the entire market, lacking the size and critical mass to do so. The server and mobile segments also help by using more specialized products that require a more involved design-in process, thereby reducing the commodity nature of the DRAM that the segments consume.

DRAM downsizing
In another change that has benefited the hypercompetitive industry, a number of DRAM suppliers in the past few years have either reduced their presence or have altogether exited the market.

The Taiwanese are no longer the powerhouse suppliers they used to be, while notable DRAM makers Qimonda of Germany and Elpida Memory of Japan have gone bankrupt and have been bought out by other players.

By the end of this year, only three DRAM manufacturers will remain—Samsung and SK Hynix of South Korea, and U.S.-based Micron Technology. With fewer entities to influence the market, a more conservative approach toward capacity expansion is expected, and more stable growth can follow.

Process deceleration
A final factor helping the global DRAM business is the slower pace of advancement in DRAM manufacturing processes. Each new generation of DRAM manufacturing technology is now taking longer to arrive.

The engineering challenges associated with shrinking DRAM size smaller than 30 nanometer— and eventually below 20 nanometer—are considerable.

The slowing cadence in manufacturing process evolution is resulting in slower bit growth, which is keeping supply in better balance with demand.

Challenge of constant undersupply
While the current state of intermittent undersupply is favorable to the industry, a state of persistent undersupply could backfire and prove harmful. Large, obstinate supply shortfalls will result in broader adoption of competing technologies as devices seek alternatives besides DRAM, and possible regulatory intervention could occur over perceived anticompetitive concerns.

Clearly then, it is in the best interest of the industry to manage supply so that it more closely matches demand—and thereby control its own future. Next year, manufacturers will need to seriously look at options for expanding manufacturing capacity to accommodate demand. But properly managed, DRAM prospects can remain healthy, IHS believes.

Mentor Graphics accelerates verification with emulation-ready verification IP for MIPI products

USA: Mentor Graphics Corp. announced MIPI-protocol verification IP (VIP) for use with its latest-generation Veloce hardware emulation platform.

This allows engineers to exhaustively stress test a device-under-test (DUT) that includes one or more MIPI protocol interfaces on their SoC, and run verification cycles at orders of magnitude faster than simulation.

The MIPI VIP enables the use of stimuli generated by modern simulation testbenches, including SystemVerilog/UVM, and SystemC-based environments, and applies them to a MIPI-based, design-under-test (DUT) running in the Veloce emulator. Since the connection between the testbench and the VIP is at a transaction-level rather than signal interface, a high level of performance is delivered.

The new VIP supports users of the MIPI camera and display-based protocols, CSI and DSI, and is compatible with the Questa® functional verification platform as well as the Veloce hardware emulation platform.

OCZ achieves first-pass silicon success with DesignWare IP and Synopsys professional services

USA: Synopsys Inc. announced that OCZ Technology Group Inc. (OCZ) has achieved first-pass silicon success for its NAND flash Vector SSD using Synopsys' DesignWare DDR2/3-Lite PHY, Embedded Memories, STAR Memory System solution and Professional Services.

Choosing Synopsys technology and services enabled OCZ to accelerate the completion of their project, meet their performance requirements and achieve first-pass silicon success with a production-ready design.

Targeting the high-end consumer market, the OCZ Vector SSD was designed to deliver superior sustained performance through its new, high-performance Indilinx Barefoot 3 flash controller supporting the SATA-3 protocol. For OCZ's design team to effectively focus on developing the Barefoot 3 controller while meeting their broader design goals, they needed reliable, silicon-proven IP, an integrated design flow and design expertise to augment their own. OCZ selected Synopsys' high-performance and area-optimized DDR2/3 PHY, embedded SRAMs and integrated memory test and repair solution.

To mitigate risks and accelerate the implementation process, OCZ deployed the Lynx Design System, an integrated RTL-to-GDSII block- and chip-level design environment with unique visualization capabilities for efficient design exploration and reporting of design status and trends.  In addition, Synopsys' design consultants worked closely with OCZ's engineers throughout the implementation of their chip, delivering expertise and advanced methodologies in IP integration, physical design, and physical verification that enabled OCZ to complete their implementation in less than six months.

"We were very focused on reducing schedule risk, which made it imperative that we partner with an established vendor like Synopsys that had the proven portfolio of IP, design flows and services that we needed," said Brian McMath, technical director at OCZ. "Of particular value to us was the DDR3 PHY integration expertise Synopsys Professional Services brought to the project. The combination of Synopsys' proven DesignWare IP, experienced design consultants and tapeout-proven Lynx Design System made it possible for us to achieve our performance goals while saving an estimated two months of schedule time.  The net result of our collaboration was that we got our product to market much faster than if we had done it on our own."

"Companies such as OCZ who develop innovative electronic products need design solutions they can rely on to mitigate project risks and meet tight time-to-market windows," said John Koeter, VP of marketing for IP and systems at Synopsys. "Synopsys invests heavily in developing high-quality DesignWare IP as well as providing tightly correlated tool flows and experienced design consultants so our customers can focus on the aspects of their design that differentiate them from the competition."

Micron to enjoy growth in three major product lines thanks to merger with Elpida and industry restructuring

TAIWAN: According to DRAMeXchange , a research division of Trendforce , thanks in large part to the rising spot prices of DRAM (which has grown twofold since last November), the spot market price of DDR3 2Gb has a good chance of approaching $2 in the upcoming periods.

The profits shown on Micron’s recent June 19th financial report are, in many ways, a reasonable indication of the positive trends that are currently emerging in the DRAM industry.

Observing from the market perspective, with Nanya officially transitioning into a specialty DRAM company and handing over its manufacturing responsibilities of Inotera to Micron, the US based chip maker can be said to be in a relatively strong position to increase its entire PC and server DRAM production. In the past, PC DRAM had been known to be largely unprofitable due to the intense competitions that took place within the market.

The losses incurred in the industry had been so severe that, at one point, various DRAM manufacturers felt the urge to switch to the more profitable product lines. The DRAM industry’s financial woes would not change for the better until early 2013, when the production of PC DRAM shrunk and led to a continuously declining supply bit growth.
At present, both Samsung and SK Hynix are actively lowering their PC DRAM output in favor of manufacturing mobile and server DRAM. The decision is in many ways largely beneficial to Micron, whose PC DRAM production ratio is almost 50 percent. With regards to the second half of the year, given that Micron's and Elpida's merger process is set to complete during Q3 (a process which would in effect help to increase the wafer production volumes from 190K to 370K), it is reasonable to expect the new Micron group's presence in the mobile DRAM market to increase.

As the DRAM market moves further towards an oligopoly, the company is expected to be able to pose legitimate challenges to both of the aforementioned Korean DRAM manufacturers.

Looking at the NAND Flash industry, given that the supplies in the NAND Flash market are being regulated more tightly this year, and considering how more and more manufacturers are transitioning to system products as a means to offset the weak sales effects in retail market, the contract prices in the NAND Flash market have, for the most part, remained flat throughout Q2 (from April to June).

The severe oversupply situation in 2H12, it should be noted, has also experienced a notable improvement this year. With the manufacturing costs being lowered, Micron has continued its efforts to migrate from the 25nm processes to the 20nm processes, and is beginning to place a greater amount of emphasis on the more profitable 20nm-class NAND Flash products.

In the second quarter, Micron's 20nm production proportion has officially surpassed 50 percent. As the manufacturing costs of making the 20nm products drop below those of the 25nm products, there is a chance for the US based chip maker to begin emerging as a more effective cost leader within the industry. Micron's NAND Flash production capacity, as of this moment, is at 180 thousand. Considering that the company's Tech plant in Singapore has been allocating production capacity from DRAM to NAND Flash, it is likely that in Q4, the number of the NAND Flash output will increase to 200 thousand.  

In the event that Micron sticks to its original plan of manufacturing embedded-type products, it is predicted that the chip maker will not only be able to secure the market share it already owns for SSD and specialty Flash products, but that it will also exert a noticeably greater impact on the eMMC and eMCP markets during 2H13. Because of this, we believe that the revenue of the NSG department will remain stable, and that the performances of the ESG and WSG departments will help provide a further boost to Micron's future NAND Flash revenue (given the increased use of NAND Flash in embedded applications as well as mobile devices).

Thanks to the merger with Elpida, PC DRAM, mobile DRAM, and NAND Flash are all expected to be Micron's major product lines. As the use of the 2Xnm processes intended for DRAM products increases in the next few quarters, and as the migration towards the 1Xnm NAND Flash process gradually intensifies, there is a good chance that Micron will experience at least a few more stable quarters with good operating margins (whether for DRAM or NAND Flash products).

As long as the market gradually rebounds and Micron continues developing its core technologies, the company is poised to become a big winner in both the DRAM and NAND Flash industry.

TowerJazz to service high volume, commercial infrared markets

USA: TowerJazz, the global specialty foundry leader, will be the wafer manufacturer for infrared sensing and camera devices using its proven specialty process technologies.

In addition to traditional infrared applications, TowerJazz will facilitate expansion into other consumer markets such as gaming, personal security, and application driven platforms, market segments which are already well served by the company.

The advanced CMOS-based process at TowerJazz’s US location is a viable, commercially sustainable foundry offering to support its commercial as well as its aerospace and defense customers.

“This development is a natural fit for TowerJazz. Our leading edge CMOS for custom imaging products and our expertise in bringing specialty processing and MEMS to volume manufacturing fits extremely well with the proven capabilities of our customer,” said David Howard, executive director and fellow, TowerJazz.

Atrenta announces design service alliance network in Israel

ISRAEL: Atrenta Inc. announced an alliance program with several design service providers in Israel.

The goal of the program is to bring the benefits of Atrenta’s SpyGlass RTL platform to a wider range of customers. The terms of the alliance will be managed as part of Atrenta’s SpyLinks partner program.

Design service providers that specialize in everything from FPGAs to leading edge SoCs are part of the program. Organizations that are part of the program include Inomize, Rachip and Veriest Venture. Under the terms of the program, these design service providers will collaborate with Atrenta to develop SpyGlass expertise that can be deployed as part of their service offerings.

Eliminate complicated wireless set up with TI’s new low-cost, easy-to-use NFC solutions in the Internet of Things

USA: Texas Instruments Inc. (TI) continues to deliver new near field communications (NFC) innovations in its broad portfolio of solutions that make entry to NFC development easier and lower cost.

TI introduced the Dynamic NFC Transponder RF430CL330H hardware to enable uncomplicated and inexpensive wireless set up. Also announced today, the NFCLink software is a standard NFC library for the TRF79xx NFC transceiver family to ease NFC development on TI embedded processors.

The new Dynamic NFC Transponder Interface RF430CL330H is low cost, bringing a secure, simplified pairing process for Bluetooth® and Wi-Fi connections to products, such as printers, speakers, headsets, and remote controls, as well as wireless keyboards, mice, switches and sensors. It is the only dynamic NFC tag device designed specifically for NFC connection handover and service interface functions, including host diagnostics and software upgrades.

Streamlining NFC development across TI’s entire embedded processing portfolio, the NFCLink software firmware library, in partnership with Stollman E+V GmbH and Kronegger GmbH, enables developers to quickly and easily create NFC applications for TRF79xx NFC transceivers using TI’s ultra-low-power MSP430 microcontrollers (MCUs), Tiva C Series ARM MCUs and OMAP processors.

Additional TI embedded processor platforms will be supported in the future. NFC applications running on operating systems include point-of-service devices, routers, set-top boxes, automotive infotainment and other various consumer devices.

AMD unveils server strategy and roadmap

USA:  AMD publicly disclosed its strategy and roadmap to recapture market share in enterprise and data center servers by unveiling innovative products that address key technologies and meet the requirements of the fastest-growing data center and cloud computing workloads.

Additionally, AMD revealed details of its 2014 server portfolio including best-in-class Accelerated Processing Units (APUs), two- and four-socket CPUs, and details on what it expects to be the industry's premier ARM server processor. This is on the heels of announcing the general availability of the AMD Opteron X-Series processor, code named "Kyoto," which dominates the small-core server market on every performance benchmark.

These forthcoming AMD Opteron processors bring important innovations to the rapidly changing compute market, including integrated CPU and GPU compute (APU); high core-count ARM servers for high-density compute in the data center; and substantial improvements in compute per-watt per-dollar and total cost of ownership.

In 2014, AMD will set the bar in power-efficient server compute with the industry's premier ARM server CPU. The 64-bit CPU, code named "Seattle," is based on ARM Cortex-A57 cores and is expected to provide category-leading throughput as well as setting the bar in performance-per-watt.

AMD will also deliver a best-in-class APU, code named "Berlin." "Berlin" is an x86 CPU and APU, based on a new generation of cores named "Steamroller." Designed to double the performance of the recently available "Kyoto" part, "Berlin" will offer extraordinary compute-per-watt that will enable massive rack density.

The third processor announced is code named "Warsaw," AMD's next-generation 2P/4P offering. It is optimized to handle the heavily virtualized workloads found in enterprise environments including the more complex compute needs of data analytics, xSQL and traditional databases. "Warsaw" will provide significantly improved performance-per-watt over today's AMD Opteron 6300 family.