Tuesday, July 31, 2007

Tag, you're (not) it (yet)

Proponents of near field communications want to make NFC the fifth or sixth radio in your phone, and they made a step toward that goal today by launching four tag types. NFC is mainly aimed at automating payments, which means the radios will have to be everywhere we now have debit card swipers and bar code readers.


To make this work there needs to be standards for tags that contain data the NFC devices read. The NFC Forum today essentially blessed four tag types already out in the market as part of a suite of standards that complaint devices need to support. They include tags based on the ISO 14443 Type A and B standards and Sony’s FeliCa. The Forum claims more than one billion of these tags are already deployed worldwide in public transportation, hotel and offices applications.

It's a wise move blessing what's already out there and mandating tomorrow's products need to be able to handle all these tags. That's a step toward interoperability and hopefully will make vendors think twice about launching still other tag types. But there's a world of chip, system, software infrastructure—as well as business services-- that need to be put in place to make this all work and that will take years as my colleague Junko Yoshida noted in a recent story.

When all that work is done, it is really no easier (and no more secure) to waive a radio-laden cellphone than swipe a card. The NFC folks say the technology, which can deliver about 500 Kbits/s of data, will serve other apps like sending pictures from a digital camera to a TV, but there are plenty of other wireless transports aimed at such uses.

In short, the tags are blessed now, but I am not sure the business case has been yet.

Monday, July 30, 2007

Warning SI issues ahead

In case you thought today's signal integrity issues were bad, better tune into Roy Leventhal's write up on the recent meeting of the IEEE Electromagnetic Compatibility Society in Hawaii. He showed signal integrity problems are not on any trip to an island in paradise.

"By the year 2013, silicon technology roadmaps call for 30 nm processes and 25GHz core switching rates," Leventhal wrote. "When dealing with such chips, engineers will need to design for a frequency range of DC to 80GHz. Meanwhile silicon chips are growing larger as more functionality gets packed onto them, so the potential for direct radiation off the silicon exists in the not too distant future."

No surprise then that he also reports that "at the past few symposiums, there has been a dramatic increase in the interest shown in signal integrity, power integrity, modeling and simulation." Buckle your seat belts, turbulence ahead.

Friday, July 27, 2007

HDMI in the doghouse

Pity the poor High Definition Multimedia Interface. First there are widespread reports of interoperability issues that threaten its future. Now the group trying to spearhead HDMI interoperability testing is under fire for allegedly charging too much and providing too little information about its test services, according to a story posted today from my colleague--and boss!--Junko Yoshida.


The last we left HDMI it was valiantly claiming dominance in the TV world and struggling to gain traction among computer makers who were allied with their own approach, DisplayPort. If the hard times keep knocking on HDMI's door it may drive the consumer companies right into the DisplayPort camp. There must be a reason we keep hearing disgruntled noises coming from their marketplace. Methinks for whatever reason the folks behind HDMI have failed to earn the trust of some of their would be OEM customers.

Thursday, July 26, 2007

Wireless USB noted

I almost forgot to mention both Dell and Lenovo both are putting wireless USB into at least one notebook model each shipping this year. I had a sneak peak of the Lenovo news a few weeks back, but the fact Dell is jumping in too was news to me.

The USB Implementers Forum announced Monday the Lenovo ThinkPad T61/T61p 15.4-inch Widescreen and the Dell Inspiron 1720 have both been officially certified for wireless USB and are expected to ship before the end of the year. D-Link and IOgear said they will ship hubs and dongles, too.

This doesn't make a market by a long shot, but this marks the start of what could someday become something viable.

The return of iSCSI

"The server business has come down to just three or four big vendors, and you can only make money in the I/O."

Thus spoke Kimball Brown, vice president of business development at startup ServerEngines, in an interview this week. He was describing the founder's move out of the chip set business and into 10 Gbit Ethernet and specifically iSCSI.

I tracked a number of iSCSI startups back in 2002 when iSCSI was hot—Aarohi Communications, Astute Networks, Silverback Systems, Trebia Networks and more. All have pretty much died, faded away or been acquired. But analysts say ServerEngines has the performance, relatively low power, low cost and market vision to make a go of storage over IP if anyone can.

We shall see. Brown himself told me circa 2003 as an exec at the former ServerWorks that the company's next chip sets would support iSCSI, a feature that would be dominant in servers in 2005. Maybe he is wrong again.

There are plenty of wrinkles to this story including the outlook for 10G Ethernet, the future of server blade architectures and convergence in data center networking. But for now, let's hold that one thought: I/O is the place to be today.

Wednesday, July 25, 2007

Le powerline francais

Thanks to Xavier Mongaboure for clarifying some details about the powerline technology and plans for Spidcom Technologies, a Paris-based startup I first heard about last week. It turns out Spidcom has a proprietary 200 Mbit/s powerline technology as well as plans to roll products based on the HomePlug AV standard which sports a similar data rate.

Spidcom's proprietary powerline technology uses the traditional 2-30 MHz frequency range, but it divides this spectrum into seven independent sub-bands of 4 MHz each to help it adapt to the harsh realities of the medium. The approach is protected by several patents.

The powerline approach has been deployed since 2005. Partner Shenzhen Mopnet introduced products based on it at CES this year focusing on delivering Ethernet over coax.

Going forward, Spidcom believes HomePlug AV will be the standard for in-home powerline networking. It has been a member of the HomePlug group since 2003 and plans to roll out products based on the AV spec, probably in 2008.

At that point, it will position its proprietary technology as a solution for coax nets, much as PulseLink, Tzero and now the WiMedia Alliance are doing for their UWB technology. Spidcom will help customers using its proprietary technology for in-home nets migrate to its upcoming AV chips.

Sounds to me like the business realities of this plan—repositioning a product line and competing against two other consortia for coax design wins--may be as harsh as the interference over powerline. Here's wishing luck to the brave French Resistance.

Tuesday, July 24, 2007

Multicore versus TOE

Intel's LAN division rolled out its latest 10G Ethernet controller yesterday, firing a round for CPU vendors who want to keep the job of handling high-end Ethernet tasks on their multicore processors. The Intel 82598 is its first to use its I/O Acceleration Technology (I/OAT) at 10 Gbits/s.

Even Intel admits there is a performance benefit on some high-end Ethernet controllers for the alternative, providing TCP offload engines (TOEs), and it won't rule out the possibility it may offering such a product some day. HP is so far opposed to the I/OAT approach, opting for TOEs. Intel says they are working on them and it claims design wins in all the other top sever vendors, though Intel admits server makers use a mix of I/OAT and TOE.

There are many more shoes to drop on this debate as 10G Ethernet ramps up over the next two years. At least one more new startup is about to come out of hiding on this issue. Stay tuned for more on Monday.

Meanwhile, it is interesting to note, Intel is already gearing up for the so-called Convergence Enhanced Ethernet by providing a pre-standard version of per-priority pause on its 10G controller. CEE aims to lay a foundation for running Fibre Channel protocols on Ethernet in the 2010 timeframe—or maybe sooner if this new Intel part is any indication of commercial trends.

Monday, July 23, 2007

Tiny chance for Femto

Tip 'o the hat to my colleague John Walko for his good roundup on the recent activity in 3G home basestations aka femtocells. These little boxes are gaining increasing interest among carriers who recently formed the Femto Forum to promote the boxes that compete with home Wi-Fi routers. They aim to shift more data traffic to carrier nets and fend off the rise phone calls using Wi-Fi links to the Internet.

According to Walko's article, the carriers have internal disputes about how to handle key standards in the boxes as well as how to get them down to consumer prices. I have to wonder about the business case which sounds to me like it was dreamed up by Homer Simpson.

Walko's article says carriers are concerned about whether they can get femto cells down to $70. Hello! My last Wi-Fi router was $35 which was amazing discounted to zero by a manufacturer's $35 rebate! I don't know how that worked for Linksys, but it worked for me just fine.

Also, I can see buying one of these if you are in a rare situation of having poor cellular coverage at home, but that's a femto minority of people I know. And no one will be motivated to buy the boxes so they can give up free Wi-Fi calls and start paying carriers again. Duh!

The article quotes market researchers who say this market, like so many they track, is headed up and to the right like a proverbial hockey stick. What are these guys smoking?

Friday, July 20, 2007

40 and 100G in Ethernet's future

Kudos to the engineers who toughed out a hard year at the IEEE's Higher Speed Study Group to hammer out a consensus accepted this week by the broader body that the IEEE will write standards for both 40 and 100 Gbit/s versions of Ethernet. Their work will pave a long and rich road map for years to come that should serve both data centers as well as central offices.

Debates over whether a 40 Gbit/s spec would disrupt the networking world, and prospects for compromise looked pretty dark even a week ago.

Hard as the work has been to date, the big job starts when a task force is formally convened. The effort to define a 10 Gbit/s version of Ethernet to run over copper cables was tough going and commercial products based on it still face plenty of design and cost hurdles. This week's decisions promise 100 Gbits/s over 10 meters of copper cable, and that will require some significant engineering. So congrats, and roll up your sleeves.

Thursday, July 19, 2007

Agilent may shoot IBIS Bird

The I/O Buffer Information Specification (IBIS) group has a draft spec of its Bird application programming interface for modeling interconnects at 5 Gbits/s and up. Todd Westerhoff says SiSoft already has some code for it that runs at up to a million bits a minute. See the full story here.

This is real progress—faster than I expected—on a much needed standard. Kudos, guys.

But wait, Ian Dodd, late of Mentor Graphics, tells me Agilent and a major semiconductor company—could it be IBM?--will propose an alternative at the IBIS group's September meeting in Beijing. I am still waiting for more word from Ian on the details. If anyone knows about this or any other clouds in this Bird's sky, make a comment here or drop me a note at rbmerrit@cmp.com

Wednesday, July 18, 2007

Two nods for MoCA

I talked to two senior technology executives with set-top box companies yesterday at iTV Con. Both gave a nod to the rising fortunes of Multimedia over Coax in home networking. But both also made it clear this remains a pretty wide open field.

Cable TV companies in North America are expressing support for MoCA, said Ken Morse, vice president of client architecture for Cisco's Scientific Atlanta (SA) group. However, some still want a vanilla Ethernet jack into which they can plug any networking dongle they want. And SA is also delivering set tops to AT&T using the competing approach of the Home Phoneline Networking Alliance.

Outside the US the situation is much more fragmented. Many consumers pay a fraction of what US users spend on TV, leaving carriers little budget for niceties like home networking.

As for wireless, some are exploring ultrawideband because its high bandwidth may overcome the inherent fragility of wireless, but carriers are not asking for set tops with 802.11n, Morse said. Paul Fellows, CTO with Amino Technologies, an up and coming set-top maker in England basically agreed with Morse except on the Wi-Fi point. He is integrating 802.11 with technology from startup Ruckus to bolster range and reliability.

Tuesday, July 17, 2007

HomePlug redux

Thanks to Intellon for the update on its recently announced INT6300 HomePlug AV chip I noted last week. Turns out the company migrated the 130nm part to a 90nm process. It is still bound by the AV spec's 200Mbit/s PHY rate, but Intellon claims it offers about a ten percent performance boost over the previous generation and a 25 percent reduction in the bill of materials thanks to integrating its analog front end and more memory.

Of more interest, was the report that Conexant and Spidcom—a new startup to me—are preparing their own Homeplug AV chips. The Spidcom story is not immediately clear based on a reading of their somewhat poor Web materials. They appear to have some loose affiliation with a China company-- Shenzhen Mopnet Industrial Co., Ltd—which seems to have some proprietary powerline technology that also works over coax. Paris-based Spidcom showed Mopnet's technology at CES but also has joined the HomePlug group, implying new products based on the consortium's specs may be in the offing.

Last I looked there were still three competing approaches to powerline networking—the HomePlug spec and proprietary technologies from Panasonic and Spain's DS2. That's unfortunate because as I said before powerline looks to be the lowest volume network technology in an already highly fragmented home net universe.

Monday, July 16, 2007

Atheros plays games

Kudos to Atheros for nabbing a design win in the next version of the Nintendo DS handheld game machine expected to ship in time for Xmas, according to a report from Lehman Brothers this morning. The Wall Street firm estimates the chipmaker could grab $72 million or more in revenues in 2008 from sales of its integrated ROCm chip to Nintendo.

International Data Corp. says Nintendo could ship as many as 24 million handhelds next year. That's a big bump for Atheros and Wi-Fi, too. But long term, Atheros will see more volume from its drive into cost-squeezed desktop Wi-Fi chips, an effort which will make PCs chips nearly 40 percent of its business, driving down profitability, Lehman said.

Thursday, July 12, 2007

Plugging into Infiniband

The registration statement for Infiniband switch maker Voltaire makes for some interesting reading. I did not know, for instance, the company had to defer nearly an entire quarter's revenue ($7.3 million) from late 2006 to sometime late this year because it was having trouble with its 20 Gbit/s switch which is now just about ready to be re-launched. Mellanox Technologies claims there is no problem with its switch chip used in the system.

If I were a punter, as the British say, I think I'd pass on buying Voltaire shares. The company's revenue growth is settling into something in the 'teens and it has yet to demonstrate a path to profits. By its own admission it is a third-tier player behind Cisco (Topspin) and QLogic which also have Infiniband products now.

What's more the company has a narrow support system. Three OEMs account for the majority of its sales--IBM, HP and Sun respectively. And Mellanox is the only supplier of its core silicon. And it readily admits Infiniband is still very much a niche market, a tenth the size of the emerging 10 Gbit Ethernet market.

Voltaire hopes to raise about $67 million by offering shares at about $13 each. It's not being very specific about how it will spend the money beyond saying it will help fuel R&D, overseas expansion and maybe an acquisition--though it has no concrete M&A plans. The company admits the cash will only see it through the next year or so.

I guess I should take my hat off to a startup that can build a business on a niche technology with a single chip supplier and three main customers. But I have to wonder how long they will be around to salute!

Wednesday, July 11, 2007

New spud in the stew

Intellon Corp. rolled out Monday a second-generation HomePlug AV chip, its INT6300, that it claims delivers better performance at lower cost than the its previous chip. I'm still waiting for details how much the new chip boosts performance and lowers cost, but its press release at least gives a few hints about how it does it.

The new chip uses more modern process technology and sports a faster processor and memory bus, but again no specific numbers were given. Alas! A look at the product brief (itself pretty sketchy) reveals the new device is a multi-chip module packing the analog front-end together with the media access controller and physical layer interface for the first time. I infer from the block diagram, engineers can eliminate external flash required for the previous chip, tapping into host SDRAM memory instead. Overall, I give Intellon a "D" grade for a poor job documenting this new device.

In the bigger picture, powerline communications appears to be the least favored ingredient in today's home networking stew. Cable TV companies seem to be sloooowly migrating to the MoCA technology, while IPTV carriers are opting for HPNA. Meanwhile, technically literate consumers are snapping up Wi-Fi gear at the local electronics superstore in a big way, and every once in awhile those tech savvy users buy a powerline product. Intellon's new chip won't shift those dynamics.

Tuesday, July 10, 2007

Fork in the road for Ethernet?

The more I dig into the outlook for the IEEE High Speed Study Group the cloudier it looks.

So far I have heard from a representative of computer companies that are making a big push for 100 and 40 Gbit/second standards as part of one IEEE effort going forward. The 100 G standard is something everyone wants, but computer companies are also hot for 40G for three data center applications. They want 1 meter 40 G backplanes, 10 meter short reach lengths for racks of servers and switches and 100 meter lengths to span a data center.

A networking person part of the effort is begrudgingly willing to support a separate proposal for both 40 and 100 Gbit/second standards, but is concerned the 40G work will lead to competing 40 and 100 G networking products that confuse and fragment the market. He also wants to see separate IEEE work groups hammer out the 40 and 100 G specs, a procedural difference.

As usual in a big IEEE group, there seems to be a certain amount of politics, clashing agendas and paranoia. But it's the telecom people I most want to hear from now because I understand they and the networking companies are at odds. So if you are with AlcaLu, Nortel, Ericcson, Nokia or Comcast, leave a posting here or drop me a line at rbmerrit@cmp.com

The HSSG chairman is upbeat about a breakthrough at the July meeting, but so far I am not hearing anything about a consensus. Ethernet has become the main highway for all communications so there is a lot riding on this work, including that little application called the Internet.

Monday, July 09, 2007

One more radio in your cellphone

The folks behind near field communications are stepping up their efforts to put one more radio in your cellphone. The NFC Forum announced today that it has begun work to develop a specification for a Host Controller Interface that will link NFC contactless controllers and application processors. Planned for release in 2008, the HCI is targeted at cell phones, PDAs and PC peripherals.


One of the hot applications for NFC is turning the cellphone into a debit card that can make payments at a suitably equipped check out stand with a waive of the handset, a use case that has already made its debut in Japan. Also, Nokia debuted NFC-enabled phones for the US market at CES this year.

NFC operates in the 13.56 MHz frequency range, over a typical distance of a few centimeters. It can ride on top of a variety of physical interfaces including the Serial Peripheral Interface Bus, Inter-Integrated Circuit, Universal Asynchronous Receiver/Transmitter, and Universal Serial Bus.

I'd love to hear any experiences of engineers with NFC on their implementation challenges and market acceptance in these early days of the technology. Feel free to post a comment here or email me at rbmerrit@cmp.com.

My boss at EE Times concluded last year there is still a long road ahead for NFC in the cellphone. In any case, it's amazing to think some of today's cellphones already might pack six or more radios—cellular, Bluetooth, Wi-Fi, GPS, FM and NFC.

Friday, July 06, 2007

Infiniband surges in supers

The news from the International Supercomputing Conference in Dresden last week looked pretty rosy for Infiniband. In its core application area of high-end technical computing clusters, Infiniband is still strongly ascendant.

The June 2007 version of the Top 500 list reports Infiniband was used in 127 of the world's most powerful systems, up from 78 systems six months ago. Gbit Ethernet, the leader at 207 systems, down from 211 six months back, is slipping because it is perceived as running out of gas.

Sun provided some extra shine on the Infiniband rise, rolling out its Project Magnum, the world's first 3,456-port Infiniband switch. It creates a five-stage Infiniband fabric aimed at the world's biggest computing sites. Designed by Sun co-founder Andy Bechtolsheim, it uses one-sixth the space, weight and cabling of previous IB switches of its class.

This is all cool stuff, but of course the big issue with Infiniband is whether it can break out of the nerdy HPC ghetto and gain significant traction in corporate data centers where it battles tooth and nail with Ethernet and Fibre Channel.

Thursday, July 05, 2007

Driving Gen2 Express

Moving to the PCI Express 2.0 is no slam dunk. The new spec requires more attention to managing jitter, simulating channels and signal integrity among other things. Kimkinyona Fox, a PCI Express product marketing manager at Rambus which sells Express intellectual property and design support, walked me through some of the challenges.

Engineers working with Gen2 Express need to track multiple kinds of jitter—random and deterministic—to finer tolerances than with the 1.0 interconnect. The magic number here is getting to less than 50 picoseconds of jitter in 100,000 data samples.

Gen 2 supports common, data or separate transmit and receiver clocks. The latter approach can be useful when running Express over a noisy wire. However separate clocks are so hard to manage the official Gen2 spec doesn't even try to define how to go about it and says interoperability is not guaranteed. "I found that amusing," said Fox

With tight tolerances all around, engineers need to simulate channel effects much more carefully with Gen2 than the existing Express spec so they can set effective but not restrictive guard band limits. In addition, Gen2 supports signal de-emphasis at 6dB in addition to the existing 3.5 dB support.

Despite the problems, Intel already has demoed a working Gen2 workstation chip set and third party graphics chip. Those chips and other graphics and server chips using Gen2 are expected to ship this year. They will be followed by ASICs in the works for high-end data center storage and networking systems likely to ship in 2008-9. And someday, probably a few years out, high-end peripherals such as high-def video decoders will start to transition to Gen2 which promises to deliver about 80 percent of its 5 Gbit/s theoretical transfer rate.

Monday, July 02, 2007

Clusters on a chip

Organizers of the annual Hot Interconnects conference know that on-chip interconnects will be one of the keys to the multi-core microprocessor battles between AMD, IBM, Intel and Sun, a point I made in a story last year. The August conference, now open for registration, has several sessions on the topic.

"It’s a new area of research. You can see a lot of techniques from supercomputers being used in multicore processors," said Fabrizio Petrini, a researcher from the Pacific Northwest Laboratory helping organize the event.


Indeed, Cray will present a tutorial on the Gemini interconnect based on HyperTransport it will use in its next generation supercomputers. "Gemini could be a major competitor for Infiniband in coming years," said Petrini.

Researchers from the University of Maryland and Columbia University will separately deliver papers on how they are baking system cluster concepts into silicon. An Intel microprocessor expert will deliver a keynote about on-die interconnects, and the topic is expected to come up in a Wednesday evening panel on the "multicore meltdown."

The iPhone's curious LCD link

In this weekend's scramble to find out who had chips inside the iPhone, National Semiconductor was one of the winners. Portelligent reports National's Mobile Pixel Link (MPL) interface chips were used for the iPhone display.

If true, it would be one of several examples of the iPhone using decent but not necessarily leading-edge component technology. More than a year ago the ad hoc Mobile Industry Processor Interface Alliance that defines cellphone interface standards said it was rolling out its Display Serial Interface to support screens with up to XGA resolutions of 1.024x768 pixels.

Even a spokesman for National said at the time it saw the MIPI work as the long-term future, though it planned at the time to roll out 16- and 18-bit versions of the interface. A Web site national set up to rally support for MPL looks pretty much moribund.

 
interconnects