All the latest quantum computer articles

See the latest stories on quantum computing from eeNews Europe
Showing posts with label memory. Show all posts
Showing posts with label memory. Show all posts

Monday, May 23, 2016

Xilinx expands its 16nm UltraScale+ FPGA roadmap for new data centre technology

By Nick Flaherty www.flaherty.co.uk

Xilinx has added new acceleration technologies for the data centre to its 16nm FinFET-based UltraScale+ product roadmap.

The integrated High-Bandwidth Memory (HBM) is combined with the recently announced Cache Coherent Interconnect for Acceleration technology (CCIX). CCIX is initially driven by a group of seven companies to enable an acceleration framework that works with multiple processor architectures. These acceleration enhanced technologies will enable efficient heterogeneous computing for the most demanding data center workloads. The new products will also be highly leveraged in many other compute intensive applications requiring high memory bandwidth.

CCIX is backed by AMD, ARM, Huawei, IBM, Mellanox and Qualcomm alongside Xilinx to provide a high-performance open acceleration framework to data centers. For the first time in the industry, a single interconnect technology specification will ensure that processors using different instruction set architectures (ISA) can coherently share data with accelerators and enable efficient heterogeneous computing – significantly improving compute efficiency for servers running data center workloads.

Applications such as big data analytics, search, machine learning, NFV, wireless 4G/5G, in-memory database processing, video analytics, and network processing, benefit from acceleration engines that need to move data seamlessly among the various system components. CCIX will allow these components to access and process data irrespective of where it resides, without the need for complex programming environments. This will enable both off-load and bump-in-the-wire inline application acceleration while leveraging existing server ecosystems and form factors, thereby lowering software barriers and improving total cost of ownership (TCO) of accelerated systems.

“AMD strongly supports development of open standards to make heterogeneous computing more pervasive,” said Gerry Talbot, AMD corporate fellow and vice president of I/O and circuit technologies. “By joining with others in the industry to develop new interconnect specifications to accelerate performance, AMD continues its commitment to open, heterogeneous computing.”

“A ‘one size fits all architecture’ approach to data center workloads does not deliver the required performance and efficiency,” said Lakshmi Mandyam, director server systems and ecosystems, ARM. “CCIX enables more optimized solutions by simplifying software development and deployment of applications that benefit from specialized processing and hardware off-load, delivering higher performance and value to data center customers.”

"IBM Power Systems have recently demonstrated a total commitment to openness as a catalyst for industry innovation, creating benefits in cost and performance to clients in a post-Moore's Law era," said Brad McCredie, IBM Fellow and Vice President of POWER Development. "IBM is committed to working with like-minded industry leaders to expand our efforts around open coherency to help meet our clients' growing cognitive needs."

Built on the TSMCs proven CoWoS process, the HBM-enabled FPGAs will improve acceleration capabilities by offering 10X higher memory bandwidth relative to discrete memory channels for the lowest possible latency. The CCIX technology promotes efficient heterogeneous computing by allowing processors with different instruction-set architectures to coherently share data with accelerators such as the HBM-enabled FPGAs.

“Having already delivered 19 billion transistors on a chip at 20nm leveraging our second generation 3D IC technology, we are creating a third generation 3D IC breakthrough for data center acceleration and other compute intensive designs,” said Victor Peng, executive vice president and general manager, Programmable Products at Xilinx. “When combined with next generation CCIX acceleration framework and our software defined SDAccel development environment, this technology will enable a new breed of high-density, flexible platforms for accelerating compute, storage and networking applications.”

“CCIX enables greater performance and connectivity capabilities over existing interconnects, and actually paves the road to the next generation CPU – Accelerator – Network standard interface,” said Gilad Shainer, vice president of Marketing at Mellanox. “With an anticipated broad eco-system support of the CCIX standard, data centers will now be able to optimize their data usage, thereby achieving world-leading applications efficiency and scale.”

“Qualcomm Technologies is excited about the development of a new technology enabling efficient, high-performance architectures in an open, ISA-agnostic platform,” said Vinay Ravuri, vice president of product management, Qualcomm Technologies, Inc. “The data center of the future demands open architectures enabling choice of compute, acceleration and interconnect technologies, and this is a significant step forward in delivering on that goal.”

“CCIX will leverage existing server interconnect infrastructure and deliver higher bandwidth, lower latency, and cache coherent access to shared memory,” said Gaurav Singh, vice president of Architecture at Xilinx. “This will result in a significant improvement in the usability of accelerators and overall performance and efficiency of data center platforms.”
Xilinx is already collaborating with leading hyperscale data center customers to create optimized configurations and products.

Tuesday, February 23, 2016

Dedicated USB Type-C flash drive marks maturing of connector

By Nick Flaherty www.flaherty.co.uk

SanDisk has launched its first dedicated USB Type-C flash drive, pushing up read speeds to 150Mbit/s.
The company had already launched a dual drive that combined a USB Type-C and the existing Type-A connector, but the dedicated Ultra USB Type-C (USB-C) flash drive is designed specifically for next-generation devices and marks the maturing of the Type-C connector.
The need for a dual device came from the difference between the incompatible Type-A and Type-C connectors, while the underlying embedded controller for USB3.1 is the same for all the drives.
The first dedicated USB Type-C flash drive from SanDisk 
“We expect to see many new USB Type-C supported devices released in 2016,” Dinesh Bahal, vice president of product marketing at SanDisk. “As more consumers purchase ultra-mobile PCs, smartphones and tablets that feature this new, more advanced USB standard, it is critical to offer a complete ecosystem of compatible products. New offerings, like the USB Type-C Flash Drive, will give consumers the capabilities they’ve come to expect from traditional Type A ports, but with the added benefit of better performance.” 

 The new SanDisk Ultra USB Type-C Flash Drive is available in up to 128GB and delivers USB 3.1 performance of up to 150MB/s read speed, making it SanDisk’s fastest and highest capacity USB Type-C offering. The USB drive features a slim, retractable design to protect the reversible connector and is compatible with the SanDisk Memory Zone app for Android. The app, which is available for free through the Google Play Store, auto-launches upon insertion of the drive into a compatible device, allowing immediate access to the user’s content and easy file management3.
“USB Type-C connectors could be a game-changer, as they offer significant benefits to consumers, such as a slim design and reversible plug orientation," said IDC research VP Shane Rau. "We’ve already seen a considerable increase over the past year when the connector began showing up on new products from major devices manufacturers. As demand increases, we expect more and more solutions to hit the market, such as the SanDisk Ultra USB Type-C Flash Drive.”
The drives have a five-year warranty and come in capacities of 16GB, 32GB, 64GB and 128GB with U.S. MSRPs of $19.99, $29.99, $49.99 and $79.99, respectively.
Sony has launched a dual Type-A and Type-C drive that ships this month
Several flash drive makers, including SanDisk, already have dual drives that combine the Typee-A and Type-C connectors. Sony last month launched its USB CA1-series with 130MB/s read transfer speeds and densities up to 64Gbytes. The sleek connector is tailored to fit multiple devices such as your smartphone, yet robust enough to also fit a tablet and laptop for quick sharing and easy back up. The drive is available this month.
Lexar also launched a dual JumpDrive, the M20c, last year. This has read speeds up to 150Mbit/s and write of 60Mbit/s for its highest density 64GByte drive.

SanDisk already had a dual Type-C and Type-A drive
Kingston Digital, the Flash memory affiliate of Kingston Technology, last year also launched its DataTraveler microDuo 3C USB Flash drive that combined both USB Type-A and Type-C connectors. This provides up to 64GB of extra storage with read speeds up to 100MB/s and write speeds up to 15MB/s. 
Kingston Digital's dual Type-A and Type-C drive

Tuesday, December 02, 2014

Cypress and Spansion in $4bn merger

By Nick Flaherty www.flaherty.co.uk

Blimey! Cypress and Spansion are to merge in a $4bn deal that will change the landscape of the embedded market and potentially mark more consolidation. The key area for this deal is of course memory - Spansion was the spin off of AMD's flash memory business, while Cypress started out making SRAMs and has since moved into other controller-based devices. 
As a result the combined company will have annual turnover of $2bn and be the leader in SRAM and in NOR flash, and be a major player in microcontrollers - Cypress is well established in ARM-based microcontroller designs alongside its PSoC capacitative touch screen controller, while Spansion has recently been adding ARM cores to its memory devices. 
Although the deal is pitched as 50/50 - which is itself often a problem for deciding the direction going forward - it actually looks like a Cypress takeover.  Cypress founder and long term CEO TJ Rodgers will be CEO of the merged company while Ray Bingham of Spansion will be non-executive chairman and the company will be called - wait for it - Cypress Semiconductor. 
“This merger represents the combination of two smart, profitable, passionately entrepreneurial companies that are No. 1 in their respective memory markets and have successfully diversified into embedded processing,” said Rodgers, Cypress’s founding president and CEO. “Our combined company will be a leading provider of embedded MCUs and specialized memories. We will also have extraordinary opportunities for EPS accretion due to the synergy in virtually every area of our enterprises.”
The merger is expected to achieve more than $135 million in cost synergies on an annualized basis within three years and to be accretive to non-GAAP earnings within the first full year after the transaction closes. The combined company will continue to pay $0.11 per share in quarterly dividends to shareholders.
“Bringing together these high-performing organizations creates operating efficiencies and economies of scale, and will deliver maximum value for our shareholders, new opportunities for employees and an improved experience for our customers,” said John Kispert, CEO of Spansion. “With unparalleled expertise, global reach in markets like Japan and market-leading products for automotive, IoT, industrial and communications markets, the new company is well positioned to deliver best-of-breed solutions and execute on our long-term vision of adding value through embedded system-on-chip solutions.”
The closing of the transaction is subject to customary conditions, including approval by Cypress and Spansion stockholders and review by regulators in the U.S., Germany and China. The transaction has been unanimously approved by the boards of directors of both companies. Cypress and Spansion expect the deal to close in the first half of 2015. 

What other consolidation will happen remains to be seen but memory companies particularly will be vulnerable - the most notable recent deal was Global Foundries being paid to take on IBM's semiconductor business. I expect to see some other 'mergers' based around the Internet of Things as companies position themselves for growth i the next few years.

Wednesday, August 10, 2011

New SATA µSSD Specification for Embedded Solid State Drives

By Nick Flaherty www.flaherty.co.uk

Connector-free, high-capacity, embedded SSDs for a new generation of tablets and ultrathin notebooks

Memory maker SanDisk has implemented the new SATA µSSD specification into its SanDisk iSSD product line of postage stamp-sized embedded SSDs. SATA-IO, an industry consortium dedicated to sustaining the quality, integrity and dissemination of serial ATA (SATA) technology, introduced the standard today.
The number of media tablets shipped worldwide is expected to grow from 17.8 million units in 2010 to 53.5 million units in 2011 according to IDC.
The Serial ATA International Organization (SATA-IO) is also developing SATA Express, a new specification that combines SATA software infrastructure with the PCI Express (PCIe) interface. SATA Express enables the development of new devices that use the PCIe interface and maintain compatibility with existing SATA applications. The technology will provide a cost-effective means to increase device interface speeds to 8Gb/s and 16Gb/s.
Embedded SSDs offer fast performance in a tiny footprint, making them an attractive solution for all categories of ultrathin devices. The SATA µSSD specification (see below) eliminates the module connector from the traditional SATA interface, enabling developers to produce a single-chip SATA implementation for embedded storage applications. Among the first products to implement the new standard, the SanDisk iSSD series is an ideal storage solution for OEMs developing the next generation of thin, powerful mobile computing platforms.
“To widely adopt a new component technology, manufacturers need to have confidence in its performance, longevity and cost-effectiveness,” said Jeff Janukowicz, research manager, solid state storage technology at IDC. “Today’s announcement of an industry-wide standardization for embedded SSDs, combined with OEM requirements for size and performance gains in storage, should help propel the market for these tiny, versatile drives.”
“Connector-free embedded SSDs allow OEMs to develop a new generation of thin yet powerful tablets and ultrathin notebooks,” said Kevin Conley, senior vice president, client storage solutions, SanDisk. “Initiatives such as the µSSD specification promote the development of new standards for storage solutions that help manufacturers continually refine their end consumer product and drive new industry sectors.”
The SATA µSSD standard-conforming SanDisk iSSD series utilizes a new electrical pin-out that allows SATA delivery using a single ball grid array (BGA) package. The BGA package sits directly on the motherboard, allowing for form factors as small as 16mm x 20mm x 1.2mm (up to 32GB)/1.4mm (for 64GB) and 16mm x 20mm x 1.85mm (for 128GB). The SanDisk iSSD i100 SSD is available in 8 gigabyte (GB)1 to 128GB capacities, offering OEMs a flexible range of storage options.
“The market for tablets and ultrathin computing devices continues to grow along with the need for small form factor storage solutions,” said Mladen Luksic, president, SATA-IO. “We are excited to have industry-wide support for the µSSD specification and look forward to seeing many µSSD-based products available in the near future.”
SATA Express will provide a low-cost solution to fully use the performance of SSDs, while storage devices not requiring the speed of SATA Express will continue to be served by existing SATA technology. The specification will define new device and motherboard connectors that will support both new SATA Express and current SATA devices.
“The SATA Express specification provides SSD and hybrid drive manufacturers the advantages of performance and scalability enabled by PCIe 3.0 – which is available now – and the ubiquity of SATA,” said Mladen Luksic, SATA-IO president. “We expect the SATA Express specification to be completed by the end of 2011.”

The Complete SATA Ecosystem
SATA Express is one of several specifications from SATA-IO designed to offer low-cost, high performance storage solutions optimized for specific device segments. Since its introduction in 2001, SATA technology has penetrated 99% of the PC market and evolved to provide options for a number of applications beyond traditional hard disk storage. SATA is now implemented in a variety of applications including solid state and optical drives, embedded mobile devices, consumer electronics products and enterprise storage. Key specifications for implementing SATA technology beyond the HDD device segment include:
  • mSATA: A low-profile solution for mobile computing devices and other small form factor applications.
  • SATA Universal Storage Module (USM): An integrated SATA interface for providing portable, volume storage I/O to consumer electronics devices, as well as PC applications.
  • SATA µSSD: an embedded, single-chip solution that connects directly to the motherboard to enable ultra-thin form factor devices.
More information on these and other SATA specifications is available at www.sata-io.org.


Enhanced by Zemanta

Monday, April 26, 2010

AMD brings DDR3 and quadcore to embedded with two new platforms

Brings DDR3 high speed memory, quad core processors and 1080p HD video into embedded.
AMD has launched two new platforms that bring key functions from the desktop to the embedded market but with long term support and fanless operation.
The compact ASB2 platform and the high-performance AM3 platform offer combinations of power and performance with up to 74% (apparently, rather than 75%) improvement in performance-per-watt over previous generations. The embedded platforms consist of chipset and graphics solutions along with high-performance CPUs as low as 8W.
“The barriers to widespread adoption of x86 in the broad embedded market have traditionally been a combination of power, price and the physical footprint of the silicon,” said Buddy Broeker, director of the Embedded Solutions Division at AMD. “AMD’s embedded solutions have been steadily driving down those barriers while adding enterprise-class performance and features that may not have been readily available to designers in the past.

New Platform Features
  • Faster memory with support for 2 channels DDR3
  • Improved I/O for high-throughput and real-time applications with available HyperTransport 3.0 technology
  • ECC for high-reliability applications like SMB/SOHO storage systems
  • Multiple CPU options that offer wide choice of performance and power


    • Power envelopes in 8, 12, 15, 25, 45, and 65 watts TDP
    • Single-, dual- and quad-core
    • Up to 2.8 GHz
  • AMD 785E chipset with support for PCI Express 2.0
  • New ATI Radeon HD 4200 graphics with DirectX 10.1, full 1080p display resolution support, HDMI, and power-savings capability with options to support multiple displays
  • Socket AM3 package, compatible with socket AM2 when using DDR2 memory for increased design flexibility and scalability
  • Lidless BGA package offers low-cost to manufacture, high reliability, and low z-height for small form factors and fan-less designs
Reblog this post [with Zemanta]

Tuesday, December 22, 2009

Arizona State stacks nanoscale memory chips

Researchers improve chip memory by stacking cells

Researchers at Arizona State University have developed an elegant method for significantly improving the memory capacity of electronic chips.
The researchers have shown that they can build stackable memory based on 'ionic memory technology' which could make them ideal candidates for storage cells in high-density memory. Best of all, the new method uses well-known electronics materials.
“This opens the door to inexpensive, high-density data storage by ‘stacking’ memory layers on top one another inside a single chip,” said Michael Kozicki, an ASU electrical engineering professor and director of the Center for Applied Nanoionics. “This could lead to hard drive data storage capacity on a chip, which enables portable systems that are smaller, more rugged and able to go longer between battery charges.”
“This is a significant improvement on the technology we developed two years ago where we made a new type of memory that could replace Flash, using materials common to the semiconductor industry (copper-doped silicon dioxide). What we have done now is add some critical functionality to the memory cell merely by involving another common materia – silicon.”
Kozicki outlined the new memory device in a technical presentation he made in November at the 2009 International Electron Devices and Materials Symposia in Taiwan. Kozicki said that given current technology, electronics researchers are fast reaching the physical limits of device memory. This fact has spurred research into new types of memory that can store more information into less and less physical space. One way of doing this is to stack memory cells.
Kozicki said stacking memory cells has not been achieved before because the cells could not be isolated. Each memory cell has a storage element and an access device; the latter allowing you to read, write or erase each storage cell individually.
“Before, if you joined several memory cells together you wouldn’t be able to access one without accessing all of the others because they were all wired together,” Kozicki said. “What we did was put in an access, or isolation device, that electrically splits all of them into individual cells.”
Up until now, people built these access elements into the silicon substrate.
“But if you do that for one layer of memory and then you build another layer, where will you put the access device,” Kozicki asked. “You already used up the silicon on the first layer and it’s a single crystal, it is very difficult to have multiple layers of single crystal material.”
The new approach does use silicon, but not single crystal silicon, which can be deposited in layers as part of the three-dimensional memory fabrication process. Kozicki said his team was wrestling with how to find a way to build a diode into the memory cell. The diode would isolate the cells.
Kozicki said this idea usually involves several additional layers and processing steps when making the circuit, but his team found an elegant way of achieving diode capability by substituting one known material for another, in this case replacing a layer of metal with doped silicon.
“We can actually use a number of different types of silicon that can be layered,” he said. “We get away from using the substrate altogether for controlling the memory cells and put these access devices in the layers of memory above the silicon substrate.”
“Rather than having one transistor in the substrate controlling each memory cell, we have a memory cell with a built-in diode (access device) and since it is built into the cell, it will allow us to put in as many layers as we can squeeze in there,” Kozicki said. “We’ve shown that by replacing the bottom electrode with silicon it is feasible to go any number of layers above it.
With each layer applied, memory capacity significantly expands.
“Stackable memory is thought to be the only way of reaching the densities necessary for the type of solid state memory that can compete with hard drives on cost as well as information storage capacity,” Kozicki said. “If you had eight layers of memory in a single chip, this would give you almost eight times the density without increasing the area.”
Reblog this post [with Zemanta]

Wednesday, November 25, 2009

USB3 solid state disk from OCZ

New names in the industry are emerging with USB3.0 coming through (despite the problems at Intel) with OCZ Technology of San Jose teaming up with Symwave to show an external high speed solid state disk (SSD) drive.



Symwave is a fabless chip supplier of SuperSpeed USB3.0 devices with headquarters in Orange County, Calif., and design centers in Shenzhen, China, and San Diego, Calif. “Thanks to Symwave’s industry leading USB 3.0 storage controller, our external SSD device delivers 10x the transfer rate of USB 2.0 at 5Gb/s, as well as several ‘green’ improvements including superior power management and lower CPU utilization,” said Eugene Chang, Vice President of Product Management at the OCZ Technology Group. “OCZ’s products are synonymous with high performance. Combining our USB 3.0 controller with their SSD technology, both architected for performance without compromise, is a perfect match,” said John O’Neill, Vice President of Marketing at Symwave. “CES will be a showcase for new and innovative products that provide consumers with a dramatically improved user experience. Symwave is committed to delivering silicon and software solutions that will transform the way consumers view and use USB storage devices.”
Interestingly, Symwave has seen an investment by networking leader SMSC alongside venture capital firms including Kodiak Venture Partners and CMEA Ventures.

OCZ has just launched the Colossus 3.5” Solid State Drive (SSD) Series, designed to meet the performance and storage demands of high-end desktop users and enterprise clients. Available in up to one terabyte with RAID0 configurations, the Colossus Series marks a milestone in solid state drive technology, making it possible to have both high performance and high capacity in one solution.
“The new Colossus Series is designed to boost desktop and workstation performance and is for high power users that put a premium on speed, reliability and maximum storage capacity,” said Chang. “The Colossus core-architecture is also available to enterprise clients with locked BOMs (build of materials) and customized firmware to match their unique applications.”
The Colossus SSD is available in 128GB, 256GB, 512GB, and 1TB with 1.5 million hour mean time between failures (MTBF).

Monday, November 02, 2009

DDR3 memory - what happend?


I've just had to buy a new laptop, and wanted something that would last for a good few years, so naturally wanted high speed, low power DDR3 memory. Could I find much? No, really just Dell and Acer.
So this analysis of what happened from Malcolm Penn at Future Horizons is well timed, and an excellent description of how technology does, or doesn't get adopted, and the implications for the embedded market! It looks like my timing is good though as now I have a lovely DDR3 system at a reasonable, not bleeding edge, cost.

Semiconductor Spotlight – DDR3

The DRAM manufacturers as a whole made a combined loss during 2008 of US$20 billion. At the moment memory pricing is firming but semiconductor memories are still plagued by a regular cyclicality of over-capacity and declining ASPs countered by tightening capacity and rising ASPs. However, although improvements are evident in the short term, the longer-term trends are still up for debate. One of the major trends currently under way is the trend to DDR3 memory from the current mainstream DDR2.
DDR, DDR2 and DDR3 are types of SDRAM (Synchronous Dynamic Random Access Memory) and use a clock signal for synchronization. DDR stands for Double Data Rate, meaning that memories from this category transfer two data chunks per clock cycle.
One of the main differences between DDR, DDR2 and DDR3 is the highest transfer rate each generation can achieve. For example, higher speed DDR2-1066 memory operates at a clock rate of 533 MHz. Whilst there is an equivalent DDR3 memory that also operates at this speed, the DDR3-1600 operates with a clock rate of 800MHz. In addition, the internal data path of DDR2 memory is 4 bits whilst DDR3 is 8 bits. DDR3 memories also operate at lower voltages compared to DDR2 memories.
This lower voltage means that DDR3 consumes less power than DDR2 for a given clock rate and lower power is becoming increasingly important for laptops and the general trend of electronic equipment to become ‘greener’.
On the minus side, DDR3 memories have higher latencies than DDR2 memories. This means that DDR3 memories take more clock cycles initially before delivering data compared to DDR2 memories.
The primary benefit of DDR3 is the ability to transfer at twice the data rate of DDR2. DDR3 can operate at higher clock rates than DDR2 and DDR3 memory uses less power for the same clock rate. An emphasis on low power and higher data rates make DDR3 memory a more attractive option provided there is a low premium on price and the motherboards are ‘DDR3 ready’.
The first DDR2's replaced DDR in PCs by Dell during the summer of 2004 and were DDR2-533 specification. Dell has been one of the leaders in new system development and it was expected that that the transition to DDR2 would start getting underway very soon after. However, history showed that the uptake of DDR2 was not a smooth transition with memory manufacturers reverting to DDR and then going back to DDR2. The changeover eventually occurred during early 2006 when DDR2-667 had both a performance and price advantage.
It looks as though history will repeat itself with the transition to DDR3 from DDR2 but perhaps over a shorter time-span. DDR3 has been in gestation since 2005 and has appeared on some motherboards during 2007. Optimistic predictions were made for adoption during early 2008 with predictions of 30 percent or more penetration of DDR3 memory with less than ten percent price difference in price over DDR2 by the end of 2008.
By mid-year 2008 this prediction was looking increasingly unlikely and eventually did not happen. However, the pricing has been altering in favour of DDR3 in the last four months and the percentage of DDR3 has been creeping up to over 20 percent of total PC memory in the first half of this year.
PC OEMs are inexorably price driven and will tend to use the cheapest available memory given comparable performance in mainstream machines. Another factor is, of course, the type of memory the motherboard can take. DDR3 has become more popular since July and, as a consequence, the price for this memory rose as demand increased.
As the DDR3 prices increased, PC manufacturers started to go back to using DDR2 as these were lower in price. However, increased DDR2 demand, in turn, drove up prices. With prices now higher for DDR2, PC OEMs will switch back to DDR3 parts. This cycle will likely continue for the next six months until DDR3 memory becomes the mainstream memory for PCs probably during mid-2010.
The adoption of any new technology is always beset by problems in its introduction and the transition from DDR2 to DDR3 is promising to be no exception. It has been clear for some time that DDR3 memory has had the potential to become the mainstream PC memory but market adoption depends on a careful mix of price and performance with a lot of emphasis on price.
As such we believe DDR3 will really come into its own when the new generation 50nm designs hit the market. This advance will also allow a large increase in die-per-wafer and either an improvement in profitability or a tool to increase market share depending on the individual memory manufacture’s strategy.

Monday, September 28, 2009

SATA flash drives move to 62GBytes for embedded


Toshiba Electronics Europe (TEE) has moved to 32nm NAND flash technology for a series of solid state drive (SSD) modules, increasing the capacity and lowering the cost for embedded systems. The SG2 modules come in two types, one based on the new low-profile mini-SATA (mSATA) interface standard and the other a Half-Slim type, which uses a SATA connector. The drives are available in 30GB and 62GB modules. Volume production will start in October.
The two types of modules, each smaller than a business card, provide greater design flexibility and save space and cost compared to SSDs with hard disk drive form factors and cases. The 62GB module is only one seventh the volume and one eighth the weight of a 2.5-inch form factor SSD, and consumes half the power. With interface speeds up to 3Gbit/s, a maximum sequential read speed of 180MByte/s and a maximum sequential write speed of 70MByte/s,the modules will help bring the performance advantages of SSDs to notebooks, portable electronics and embedded systems. An advanced controller features a translation mode, which enables any drive configuration, and the drive supports 28-bit LBA (Logical Block Address) mode commands and 48-bit LBA mode commands. Multi-word DMA, Ultra-DMA modes and Advanced PIO commands are supported. The drives have an optional capability for secure Full Disk Encryption (FDE) backup that prevents unauthorized data access.
Embedded designers can either use the newer mobile mSATA interface where power is a key requirement, or existing, lower cost SATA.
The mSATA modules use the popular SATA interface in a small form factor module measuring 1.18 in. x 0.19 in. x 2 in. (30mm x 4.75mm x 50.95mm) that connects to the system board using a low profile SATA connector. Adaptive Power Mode and SATA Device Initiated Power Management typically reduce power consumption to less than half of the Read power level when the device is in idle, standby or sleep mode.
The Half-Slim SATA II modules feature a SATA connector, measure 1.18 in. x 0.19 in. x 2 in. (54mm x 4 mm x 39mm). The Half-Slim SATA II modules have the same SATA connector used on 2.5-inch HDD and SDD form factors. Thus, this smaller form factor can easily be used in applications designed to use 2.5-inch storage form factors. The Half Slim is also compliant to JEDEC SFF 8156.

Thursday, August 30, 2007

Graceful degradation in encoding


An interesting element of the recent launch of a family of SD video codec cores from ARC is the use of 'graceful degradation' in the encoding process.
Depending on the resources available - and that can be power or memory - the cores can throw out various tools and reduce the quality of the encoding to extend the battery life or the recording time - this is all up to the designer and allows systems based on the cores to be differentiated through these engineering tradeoffs.

The five cores are based on the recently introduced VRaptor Multicore Architecture, each is programmable, encodes and decodes a wide range of popular video standards, and comes with optimized media processing elements including:
o Up to two 128-bit SIMD Media Processors
o A dual-channel media-optimized DMA engine
o Separate multi-standard encoding and decoding accelerators
o Programmable motion estimation accelerator
o SoC development tools
o Optimized video codecs:
o (encoders) H.264 BP, MPEG-4 SP/ASP, H.263 profile 0, and JPEG
o (decoders) H.264 BP, MPEG-4 SP/ASP, H.263 profile 0, VC-1 SP, MPEG-2 MP, MJPEG, JPEG, GIF, TIFF, and PNG



All this is very interesting when video becomes a key IP block for the next decade.

Tuesday, July 10, 2007

FIC launches Linux touchscreen phone

A Taiwanese consortium Openmoko, led by embedded electronics giant FIC, has launched its Linux 'open' phone, Neo 1973, named after the year FIC claims the mobile phone was invented.

The Neo 1973 uses:

* 2.8" VGA TFT color display
* Touchscreen, usable with stylus or fingers (see the Hitachi story below)
* 266MHz Samsung System on a Chip (ARM-11 based SOC)
* USB 1.1, switchable between Client and Host (unpowered) - interesting!
* Integrated Assisted GPS
* 2.5G GSM – quad band, voice, CSD, GPRS
* Bluetooth 2.0
* Micro SD slot
* High Quality audio codec

This incarnation will not work as the applications are not available that the operators demand, but this is the start of a more open generation of mobile phones using Linux that will stimulate the eco-system of mobile phone applications (and viruses of course) and eventually drive down the costs of smartphones, but it will take many years! All credit to FIC for starting the ball rolling in a serious way, and these guys know how to manufacture in volume to a price!

Wednesday, July 04, 2007

What's an iPhone? - the breakdown

Website ifixit.com to an iPhone apart to find out what was in it. Much of this has already been suggested (Marvell's wifi chip rather than Broadcom's, the Infineon S-Gold GSM/EDGE phone chip with ARM processor) but it is all in one place and confirms things like the memory and applications processor from Samsung, DAC from Wolfson in Scotland and Bluetooth from CSR in Cambridge, and power amplifier from SkyWorks.

This of course means that this stripped down variant of the OS X operating system for the iPhone (embedded OS X, if you like) has been ported to the ARM architecture for the Samsung chip, a 700MHz ARM1176JZF core, and so could run on TI's OMAP chips, for example, or many other ARM microcontrollers. This is potentially great news for embedded system designers. The bad news is that Apple as said it won't be released to the outside world. Shame.


The ifixit breakdown:

* Samsung chip underneath the metal shield on the left side of the board on the left. Ours reads K9MCGD8U5M. The 4 GB model that Think Secret took apart had K9HBG08U1M on it, which is a 4 GB chip.
* Samsung memory stacked with an ARM-architecture processor, Part number 339S0030ARM, 8900B 0719, NOD4BZ02, K4X1G153PC-XGC3, ECC457Q3 716 shows it to be a Samsung processor rather than a Mavell Xscale. The processor is likely stacked on the SDRAM, which could be two 512 Megabit chips. The processor could have H.264 and MP3 hardware decoding built in.
* The chip above the ARM is a Wolfson audio chip. Part numbers WM8758BG and 73AFMN5.
* The chip underneath the ARM is a Linear Technology 4066 USB Power Li-Ion Battery Charger, which Apple uses in the iPods as well.
* The chip on the bottom center has this text: MARVELL, W8686B13, 702AUUP. This is Marvell's 802.11b/g 18.4mm2 chip.
* The chip in the upper right is a Skyworks GSM/Edge Power amplifier (SKY77340)
* The silver chip to the left of the Skyworks chip reads CSR 41814 3A06U K715FB. This is a CSR BlueCore4-ROM WLCSP single chip radio and baseband IC for Bluetooth 2+EDR.
* The chip with the blue dot on it is rumored to be an Intel Wireless Flash 32 MB chip. Part numbers 1030W0YTQ2, 5716A673, and Z717074A. EE Times adds the part #PF38F1030W0YTQ2.
* The chip in the lower right some claim this is an Apple-branded chip, but it's purpose is currently unknown.
* The chip in the lower left is an Infineon PMB8876 S-Gold 2 multimedia engine. Part numbers: 337S3235, 60708, and EL629058S03.

Friday, June 29, 2007

iPhone leads to a lot of nonsense

Today's launch of the iPhone has given analysts a rush of blood to the head. In my analysis for Electronics Weekly, it seems that being able to watch TV on mobile phones has all the makings of a new technology bubble. It will come in time, but the initial rush and over-optimistic figures point to a flush of expectation rather than reality.
Analysts are extrapolating from the success of Internet sites such as YouTube and the iPhone's predicted One Touch YouTube access to make predictions that are pure nonsense.
Recent presentations from ABI Research have predicted a growth in mobile TV subscribers from a million this year to 250 million in three years time. Were it to happen, this is a phenomenal growth rate and reminiscent of the worst excesses of the Internet bubble.
While the growth in YouTube subscribers is certainly true, there is no guarantee that people will pay to watch these videos over their mobiles. And pay they will have to, as this will be a revenue generating activity for both the websites and the operators.

Then there are predictions of HDTV on mobiles. The OMAP3 chip from Texas Instruments is adding this capability — and it may well be needed as showing a VGA video (or, worse, quarterVGA) on a large flatscreen looks horrible. But there are problems in downloading large HDTV clips to a phone, storing it on that phone, displaying it and then transferring it to a big display.
Does this mean that the phone will have to have an expensive mini-HDMI connector to plug into the back of your large LCD flatscreen display? Or are we waiting for an UltraWideBand wireless link to transmit a clip at 480Mbit/s over the two metres to the TV? If we are, we will have to wait quite a while for the technology in a form that can be integrated into both the phone and the TV.
There is a small but growing market for mobile TV — for viewing short clips of sport or news, for live action that you can’t wait to see and for boring trips on public transport.
These huge growth predictions do no one any favours and present the electronics business as a group of credulous fools who can be taken to the cleaners all over again by the financiers.

Tuesday, June 12, 2007

Price and power war looms for ARM M3 microcontrollers


STMicroelectronics is pushing a new microcontroller based on the ARM Cortex-M3 core, the STM32 line, challenging start-up Luminary Micro which launched $1 parts last year.
The STM32F103 “Performance” line, with 72MHz clock frequency, or the STM32F101 “Access” line, with 36MHz clock frequency. Both lines offer from 32K to 128K of embedded Flash memory but differ in maximum SRAM size and peripheral combinations. At 72MHz, executing from Flash, the STM32 consumes only 36mA, which ST claims is the lowest power consumption in the 32bit market, corresponding to 0.5mA/MHz.
Luminary Micro were the first to launch an ARM Cortex microcontroller using the M3 core over a year ago, and now have 27 variants starting from $1, so there is tremendous price pressure in this market as ST's Access range starts at $1.80. Power figures for the Luminary parts are hard to come by as a lot of the figures are still pending, but from the data sheet the power is 115mW (Idd of 35mA at 3.3V for the LM3S101) or 5.75mW/MHz at 20MHz, which seems a substantial difference.
“Until now, 16bit and 32bit designers have often been faced with difficult choices, involving compromises between factors such as cost, power, and performance and whether to use industry-standard or proprietary platforms,” said Jim Nicholas, general manager of ST’s microcontroller division. “By eliminating the need to make these compromises, the STM32 will lead the obvious convergence of the 16bit and 32bit MCU markets.”
Peripherals on the ST parts include up to 128KB embedded Flash, up to 20KB of RAM, up to two ADCs (12-bit at 1 microsecond conversion time), up to three USARTs; up to two SPI (18MHz master/slave); up to two I2C; up to three 16-bit timers (4 input capture/4 output compare/4 PWM each) and a dedicated 6-PWM timer with embedded dead times for regulation and motor control vector drive applications, as well as USB, CAN, and seven DMA channels.
ST is pushing the low power applications such as medical glucose and cholesterol monitors as well as industrial PLCs, domestic appliances, domestic and industrial security, fire and HVAC systems, and consumer/PC applications such as card readers and biometrics.

Tuesday, May 22, 2007

Broadcom plays off ARM and MIPS

Broadcom is licensing processor cores from the two arch competitors ARM and MIPS Technologies.
It is licensing the ARM Cortex-M3 for wireless designs and the MIPS 74K announce today for designs over 1GHz such as digital TV, set-top boxes, next generation DVD players/recorders, broadband access, PON, residential gateways and VoIP. Broadcom was the first to license the 74K core in January as an early access customer.
“The rapid convergence of digital devices and the emergence of new consumer markets, including IPTV, HD DVD, Blu-ray Disc and 802.11n, are driving the demand for superior performance, lower system costs and low power consumption,” said Jack Browne, vice president of marketing at MIPS Technologies.
However this is a complex beast with a 17-stage pipeline employing a combination of out-of-order dispatch and asymmetric dual-issue. This gives the higher frequency and higher performance with lower area and power, with 1.7sq mm core area and 0.76 mW/MHz in TMSC's general purpose 65nm process.
Broadcom is using the ARM M3 in a multiple-use licensing agreement for wireless LAN (802.11g,b and n,presumably) and Bluetooth. The key is the Thumb2 instruction set that combines both 16bit and 32bit instructions to keep code size down.
“The superior code density of the ARM Cortex-M3 processor enables a reduced memory footprint while delivering significant performance advantages over existing designs,” said Robert Rango, Group Vice President and General Manager of Broadcom’s Wireless Connectivity Group.
“Broadcom has long been an innovator in the wireless space, creating groundbreaking solutions enabling the delivery of innovative content in the home, office and on-the-move,” said Graham Budd, executive vice president and general manager of the Processor Division at ARM. “By combining Broadcom’s industry-leading wireless expertise, with the ARM Cortex-M3 processor, Broadcom has the building blocks for a compelling solution, tailored to the specific demands and rigors of the evolving consumer electronics market.”

Tuesday, April 03, 2007

Spansion adds microcontroller to embedded memory chips


Spansion is adding a microcontroller to its non-volatile memory chip that combines multi-level NOR and NAND technology.
The MirrorBit Eclipse architecture combines multi-bit per cell NOR, NAND and flash memory on a single die. This will allow mobile phone handset makers to save up to 30 percent or more on their handset memory subsystems bill of materials costs, while giving them more flexibility in their designs.
This comes at a time when arch-rival Intel is looking to focus more on NAND mass storage (like that used in MP3 players) than NOR, where it dominates the market, particularly in mobile phone handsets.
The NOR interface of MirrorBit Eclipse gives program and data storage with XIP (execute-in- place), reducing the amount of DRAM in the system, while the ORNAND and 4 bit/cell Quad combination allows large amounts of multimedia content to be stored on the same die. It also integrates a programmable microcontroller, which replaces the conventional state machine typically used in Flash memory and also supports built-in self test (BIST).
The company expects first silicon in Q3, and plans to sample 65nm MirrorBit Eclipse solutions built from 300mm wafers at its SP1 facility later this year. Based on 2-bit per cell MirrorBit technology, these solutions will have the ability to run code at the high speed of traditional NOR. At 45nm, the solutions are planned to include both 2-bit per cell and 4-bit per cell storage, significantly increasing the density, and enabling the seamless combination of high- performance code storage with large blocks of multimedia storage capacity.
"The MirrorBit Eclipse architecture is the result of a multi-year strategy to leverage a single technology, MirrorBit, to develop a broad range of solutions including NOR, ORNAND and Quad solutions to meet customers' requirements. Now with the MirrorBit Eclipse architecture, we will bring together all three solutions on a single die to deliver a powerful combination not previously possible in the industry," said Bertrand Cambou, president and CEO of Spansion, the memory joint venture between AMD and Fujitsu. "MirrorBit Eclipse architecture's ease-of-use and compatibility with existing platforms can enable handset OEMs to rapidly bring to market innovative feature phones at the same or lower costs to hit industry standard price points."

Thursday, March 01, 2007



The latest semiconductor market analysis from Future Horizons points to the latest restatements of chip makers and the lowest month-on-month growth in ten years, which for the month of Christmas is not a good sign.
"Depending on whether you see the glass half full or half empty, the market is simply pausing or confirmation that the era of strong growth is over," said Malcolm Penn, CEO of Future Horizons. "The only real certainty in the semiconductor industry is its ability to constantly deliver uncertainty, and clearly Q4’s results were a test of vision and faith."

Wednesday, February 21, 2007

3GSM: TI looks at low power DDR2 for next generation mobile chip

Texas Instruments is finalising the architecture of its next generation chip family for mobile phones which is expected to be called OMAP4.
"Memory bandwidth is the issue," says Rick Wietfeldt, chief technical officer of the handset division at TI, so there are discussion on supporting the LP-DDR2 low power memory interface currently being developed by JEDEC with a 533MHz 32bit wide interface.
The current generation, OMAP3430 and 3420, use TI’s Imaging Video Accelerator (IVA2) to handle high definition TV encoding and decoding. From external suppliers, the chips use the ARM Cortex A8 processor core and are the first to use the SGX 3D core from Imagination Technologies and an interconnect from Sonics.

Thursday, February 01, 2007

Wireless networking boom for 2007 as silicon design teams spin out


The latest figures from leading market researcher Future Horizons see boom times for Ultrawideband (UWB), Zigbee and WiMax networking chips over the next five years, as well as DAB digital radio and silicon for robotics. But the value of chips for DVD players will almost halve, being replaced by the growth in DVD recorders, and if you are designing RFID tags for anything other than shop applications, stop!

Design teams spin out


The other area that will grow will be spin outs of design teams. As companies follow a 'fab lite' strategy (see my NXP stories in November and the recent David Manners blog on this topic) that is really a fabless strategy, Malcolm Penn predicts we will see more and more of the silicon design teams spinning out of the big chip companies as there is no need for that vertical company structure any more. Which means startup support will be even more important as time goes on. There are some big implications here, for the way companies link together, communicate and even how they buy services, that will be a running theme through the next 12 to 18 months.

Rights for robots? Not in my lifetime says leading researcher


One of the UK’s leading robotics researchers has dismissed a recent UK report that calls for robot rights (akin to human rights) to be investigated.
“Personally I think that’s absurd, not in principle but in the 20 to 25 year timeframe mentioned,” said Alan Winfield, Hewlett Packard professor of electronic engineering at the University of the West of England and head of the Bristol Robotics Lab, speaking at HP Labs in Bristol. “We won’t see robot rights in my lifetime but just possibly in the lifetime of our youngsters.”
Part of this is down to the development of artificial intelligence, which is not a factor of processing power, he says. “We already have tons of processing power. There are already computers that can simulate a similar number of neurons in the human brain but that doesn’t make them smart. We are decades if not longer from understanding the architecture of intelligence.”
The report on robot rights from Ipsos Mori was commissioned by the UK Office of Science and Innovation's Horizon Scanning Centre and released in December.
However, he is part of a project called Walking with Robots that is taking the issue of robotics out to the public with 12 other research centres around the country. “The aim is to address the questions, to take these deep and important questions about intelligent robots out to the public over the next three years,” said Winfield.
Winfield is also working on a new project to look at the evolution of a robot society by copying behaviours in large numbers of small robots. “Here we are interested in doing evolutionary anthropology, trying to see if we can set up a society of robots where there is sufficient interaction to see if there are proto-cultural artefacts that emerge,” he said. “We have a very strong team of people including an evolutionary ecologist and an artist because one of the big problems is interpreting what we see.”
The group has been working on all sorts of things, including 'self powered' robots. These use a type of fuel cell to keep them going, only the fuel for the cell is dead flies (which decompose and produce a small amount of energy, but at least there is a continual source!)