intc on SWiK

Can Nvidia Play with the Big Boys?

Wed, 13 Aug 2008 08:05:55 -0700

Despite reporting a second-quarter loss last night, due in part to costs associated with the faulty packaging on some of its chips placed in thousands of laptops, Nvidia still has a plan for semiconductor domination through the GPU. But if it wants to execute, it needs to accept the realities that come with stepping into a competitive market. The earnings call shows Nvidia still has a lot to learn.

In yesterday evening’s call, CEO Jen-Hsun Huang admitted to a $196 million charge because of problems with its GPUs in some laptops. He also talked about some pricing mishaps that occurred as AMD pushed out a highly competitive product with a lower price. Nvidia is learning, but there are two bright spots in the call, related to its Tegra chipset for mobile Internet devices and smartphones and bringing high-level parallel processing to the consumer.

Huang said Tegra wouldn’t be shipping in products until next year (something he told us earlier this year in an interview), but growth from CUDA on laptops and desktops should have an impact over the second half of this year (which will be the second half of fiscal 2009 for the firm). CUDA is a programming tool that allows software coded in C languages to run on the multiple cores in a GPU. It helped the company make inroads in the scientific computing community, and thanks to software from startups such as Elemental Technologies, the goal is to bring that level of parallel processing to consumers.

“We can’t just keep selling chips that make graphics run faster and cheaper. I mean, that’s all very nice and it’s all good but we need to advance the visual computing field in some remarkable and important way, and parallel computing is one of the most important investments that we are making,” Huang said on the call.

Such a shift, if well executed, will bring a level of power to computers that has been reserved for research institutions and mainframes. The next laptop you purchase could very well be able to analyze real-time trading data and spit out investment decisions. The key will be building software that’s designed to take advantage of it.

image courtesy of Nvidia

Intel’s Larrabee Aims to Take on Nvidia and AMD

Mon, 04 Aug 2008 11:12:39 -0700

Last week, Intel offered up a sneak peak of its Larrabee graphics processor, due out in 2009 or 2010 and guaranteed to raise the competitive pressure on graphics chip makers Nvidia and AMD. Unlike its existing integrated graphics chips, Larrabee will be a standalone processor, but don’t expect that it will be a success.

As computing has required faster chips, Intel and other chip makers have added more cores, a tactic that GPU makers have used for years in order to increase parallel processing. GPUs from Nvidia contain as many as 240 cores while those from AMD, which that company acquired when it purchased ATI Technologies in 2006, have hundreds. So they’re faster.

But they’re also harder to program, something Nvidia is trying to solve with more flexible chips and a new programming tool called CUDA. But most enterprise and consumer software runs on x86 chips and needs adaptations to take advantage of GPUs. Intel’s Larrabee chip has multiple cores, but is not a GPU. Intel claims this offers people the performance gains and ability to render graphics much like a GPU does, but that Larrabee’s x86 architecture allows for easier programming.

It’s nearly impossible to judge a chip until you’ve seen it in action and tracked whether OEMs want to put it in their devices, but my bet is that Intel won’t be able to compromise with a many-cored CPU and believe it will beat a GPU at its own game. Nvidia and AMD are hoping as much, especially Nvidia, which has the dominant GPU market share — right behind Intel’s integrated chips — and wages an almost constant battle again Intel’s PR on this front. As Nvidia’s small but fierce marketing team faces off against Intel’s Goliath, grab some popcorn, because it’ll be a graphics showdown worth watching.

Google Eyeing a VC Arm?

Thu, 31 Jul 2008 08:12:22 -0700

According to the Wall Street Journal, Google may be trying to build out a corporate venture capital arm similar to other strategic venture groups at companies ranging from Intel to Motorola. The Journal reports that Google SVP David Drummond will be in charge, and that the search giant has hired 33-year-old former entrepreneur and investor William Maris to help.

Since Google already invests in plenty of companies, from Current Communications to 23&Me, the genetic information company run by Sergey Brin’s wife, I don’t think this news it terribly new or exciting. It also has a program in place for clean and renewable investments as part of its philanthropic arm. And so far, its track record as an investor has been unproven or weak given issues faced by portfolio companies. Current faces lackluster BPL use, Meraki is trying to stay relevant and electric vehicles are still far out there.

When a company sets up a venture arm it needs to decide if the investments it makes are part of a money-making effort similar to the way Intel now invests its capital, or to push product lines or ideas. Google already invests in companies to further its technology goals, so it may be looking for a strategic arm to create financial returns.

The downside of setting up a formal investment arm is that investing in startups is a long-term effort and many public companies are driven by short-term results. In tough times, it can be hard to write down the value in the portfolio, as Dell and Boeing discovered after the crash. They have both discontinued their strategic investment groups.

Update: To create a successful strategic investment group a company really needs to focus, something Google has not historically done well. If the goal is to complement and push Google’s technology efforts, the comapny needs to figure out what those efforts should be, and ferret out the best startups in that sector. If the fund is motivated by returns, then Google will need to temper its desire to only get involved with the best, the brightest and the hottest items right now (You know like accepting a $17,000 a year daycare instead of $30,000), because valuations on those deals can be sky high. I’m not terribly optimistic that Google has the discipline to handle strategic investing, or even if it really should.

photo of David Drummond courtesy of Google

HP, Yahoo and Intel Create Compute Cloud

Tue, 29 Jul 2008 09:04:59 -0700

Updated at the bottom: At long last, Hewlett-Packard is stepping up with an answer to cloud computing by inking a partnership with two other big technology vendors and three universities to create a cloud computing testbed. Through its R&D unit, HP Labs, the computing giant had has teamed up with Intel, Yahoo, the Infocomm Development Authority of Singapore (IDA), the University of Illinois at Urbana-Champaign, the National Science Foundation (NSF) and the Karlsruhe Institute of Technology in Germany.

The cloud will comprise six physical locations where mostly HP servers containing between 1,000 and 4,000 mostly Intel cores will run Apache Hadoop. The goal of the project is to give cloud access to academics and research institutions trying to build out services and work within the clouds. HP also hopes to use the testbed project to develop tools and software to push its Everything-as-a-Service idea. Researchers will be able to access the cloud through a proposal process later this year.

Such efforts parallel Google’s attmepts to create clouds for university students seeking to understand how their software might perform in a cloud environment. As Chris Bisciglia, a senior software engineer with Google, pointed out to me in June, most universities don’t have the resources to help a young programmer see if their application can work across hundreds of servers and achieve “Internet scale,” but it’s an important criteria for today’s developers.

Today’s announcement brings HP one step closer to offering a cloud service of its own rather than merely selling boxes to cloud providers — something that’s becoming more apparent after its deal for EDS and the launch of it’s new Scalability Computing Initiative in May. We’ll update the story with more information about testbed’s limits and what it means for other industry players later this morning.

Update: After chatting with the powers that be at all three companies, it’s clear that this cloud has two goals. The first is to figure out what people need to learn about building applications for the cloud and to tie such clouds together, and the second is to crush Google when it comes to building a cloud for researchers to play around with.

Prith Banerjee, SVP of research at HP and director of HP Labs, points out that the Yahooptel (thanks, Alistair) cloud is open all the way down to the hardware, which will allow researchers to use any programming language, operating system or other software to build applications at various layers of the cloud computing stack.

“We want, unlike other partnerships including Google and IBM’s where the lower-level stacks are not provided in a open manner to the world, open access to all levels of the hardware,” Banerjee said. “The Google approach is a proprietary way of building the hardware, and essentially all you see is the software layer at the top.”

For all of the openness, not everything built on the cloud testbed will be open sourced. The intellectual property of some applications and research could be owned and kept by the academics using the system or the companies building it. Prabhakar Raghavan, head of Yahoo research, which has played an instrumental role in Hadoop, said the details of IP ownership are still being worked out.

Much like the question of IP ownership is still in flux, the way this cloud will function across the six locations is still up in the air. As Banerjee points out, there is still too much they don’t know about allocating resources efficiently, security and linking disparate databases together in a cloud. But if this cloud gets the open-source community some useful code, Intel some chip sales, Yahoo a way to scale applications easily across its infrastructure and HP some tools to make money off of cloud software and hardware, then that’s good, too.

Welcome to the PS3 Data Center

Tue, 29 Jul 2008 16:04:10 -0700

Computerworld has done a nice job of encapsulating a corporate IT trend we’ve been writing about for the last couple of months with our focus on accelerator chips — among them graphics processors from Nvidia or AMD and Cell (which was designed originally for the PlayStation 3) from IBM — moving into the enterprise. To sum it up, the x86 processor, the workhorse of corporate computing, can do a lot, but accelerators such as Cell or GPUs can do some things better and faster, such as Monte Carlo simulations on Wall Street or video encoding and decoding.

That’s leading some IT managers to look into hybrid machines like the newly launched Roadrunner supercomputer, which uses AMD’s x86 chips and Cell. Hybrid machines won’t take over the data center, but plenty of firms that build high-performance computing systems for enterprises are eying the trend with interest.

The Computerworld article quotes Dan Olds, an analyst at Gabriel Consulting Group, as saying that 40 percent of Fortune 1,000 companies will be using hybrid computers within five years. One challenge will be getting enterprise software ported onto the different chip architecture through efforts like Nvidia’s CUDA or IBM’s software development kits for Cell, but there are plenty of companies working on that problem.

photos courtesy of IBM

Will the FCC Play Lollapalooza?

Mon, 21 Jul 2008 14:01:42 -0700

As the debate rages over who can access the white spaces between licensed digital television spectrum, the Federal Communications Commission itself has emerged as a hot ticket. Everyone from the NFL to Lollapalooza is clamoring to have its events be used as a staging ground by the agency for the testing of devices aimed at utilizing portions of the DTV spectrum for wireless Internet access. The FCC has said it will test interference of the white space devices in 10 geographic locations or buildings in the DC area; it’s looking for other venues as well.

At issue is the ability of these proposed white space devices to operate in the spectrum, which will become available after the conversion to digital TV signals next year. Companies such as Google, Motorola, Microsoft and Intel all would like to see that spectrum used for wireless broadband access. However users of wireless microphones — everyone from recording stars to preachers at megachurches — are against that plan as they’re worried about interference on their wireless mics. The National Association of Broadcasters is opposing the efforts as well, arguing that such devices could interfere with the transmission of DTV channels.

The FCC is expected to make a decision about them later this year. Depending on the summer concert lineup, it may want to hold off doing its field tests until Madonna’s latest tour starts in October or until Led Zeppelin reunites — might as well enjoy the music along the way.

If this story interests you, check out our upcoming conference:
Mobilize — Mobile Web Today and Tomorrow

AMD Won’t Offer Netbook Chips

Tue, 22 Jul 2008 12:55:00 -0700

AMD isn’t going after the mobile Internet device market that Intel and other chip vendors are eying. AMD’s senior VP and chief marketing officer, Nigel Dessau, told eWeek, “What we are saying is that we are a smaller company and we have to focus on what we do well at this point. We are watching that segment rather than playing in it, but as it matures we’ll see where it goes.” Sounds like a good plan for AMD, which already has chips that might work for Netbooks and needs to focus on righting itself.

AMD Has Gone From Scrappy to Sad

Thu, 17 Jul 2008 20:52:41 -0700

Confession: Back when AMD was pitching its Opteron chipset, I convinced my husband to buy shares in the company on the belief that its plans to build a backwards compatible 64-bit processor was so obviously better than Intel’s efforts with Itanium that the market would eventually see it. The market did, and AMD shares went up a bit, but we soon sold them after my company changed its policy regarding stock ownership.

I say this so you guys know that I once believed in AMD. I live in Austin, where the company at one time employed more workers than in its Sunnyvale headquarters. Where Hector Ruiz, who stepped down today from the president and CEO position, lives. But I look at the sad wreck that was once a scrappy upstart irritating Intel and I don’t know what to say. I can start with the facts.

Ruiz will remain as executive chairman of the company and Dirk Meyer, the former COO and president, will become the CEO and president. Ruiz had already named Meyer as his successor, but Ruiz had also said he would stay through 2008. But AMD had seven quarters of losses and wrote down $878 million last week (for a total loss this quarter of $1.2 billion).

Meanwhile, Meyer will preside over the sale of some of AMD’s consumer assets, as announced in the company’s fourth-quarter conference call on Thursday. These assets should include some of the non-core assets related to mobile and digital television AMD purchased as part of its ATI acquisition in 2006. Those are the facts.

Looking at those facts, and the string of things that have gone wrong, from delays with its Barcelona chip to the loss of its CTO earlier this year, and you have to wonder if Meyer, or anyone inside the company should really be the one to take over. Ruiz and Meyer are both known more for their engineering talents than their business ones, which may be one of the reasons AMD held onto non-core divisions for so long. I suppose I should stop caring. After all, it’s been years since I held stock in AMD, and it gets old rooting for the underdog.

Does Intel Know What It Wants From Atom?

Wed, 16 Jul 2008 11:55:55 -0700

Yesterday afternoon, Intel’s CEO Paul Otellini seemed a little hazy on the future home for Intel’s Atom processor during the chip maker’s quarterly earnings call — a fact I don’t find all that surprising since the netbooks or mobile Internet devices the chips are designed for exist only in a marketer’s imagination and failed product implementations.

Otellini was excited about Atom, calling demand for the chip” robust,” but analysts pressed Otellini about Atom’s end market and whether the chip would cannibalize Intel’s low-end Celeron processor. The Celeron ranges from speeds of 2.13 GHz to 3.6 GHz, and is faster than Atom’s 1.8 GHz or 1.6 GHz. Otellini’s responses were less than a ringing endorsement of the chip. “[Atom] is less than a third of the performance of our Centrino (high-end mobile processor),” said Otellini. “You’re dealing with something that most of us wouldn’t use.”

Wait a second. Just weeks ago before the Computex trade show in June, Otellini told the Financial Times he anticipated a $40 billion market opportunity for Atom chips over the next few years. If most of us aren’t using these low-end chips, then who is? Otellini envisions the Atom chip for small computers in emerging markets that happen to have IP-based voice, but in late 2009 Intel will launch an Atom chip for smartphones. In emerging countries, a lot of computing is already carried out on cell phones, begging the question of where Intel’s demand for Atom is coming from. Will those products actually succeed?

As for cannibalization, Otellini said, “We do not see [Atom] replacing Celeron. If you look at the netbook products being built around Atom, they’re all lower-priced, lower features, smaller screen size notebooks aimed at first-time buyers or the second, third or fourth machine in a household. We don’t see any cannibalization.”

So Atom chips are designed for slow web access on cheap, portable machines that will act as the backup computer in my home. Wait, I have one of those already. It’s called a smartphone and plenty of companies already make processors for that market.

If this story interests you, check out our upcoming conference:
Mobilize — Mobile Web Today and Tomorrow

How DreamWorks Puts Multicore Chips to Work

Fri, 11 Jul 2008 02:56:57 -0700

You wouldn’t think that next year’s DreamWorks movie, “Aliens vs. Monsters” and the search for more crude would be connected, but they are — in that they both take advantage of parallel programming for multicore chips. And when it comes to multicore chips, big bucks are on the line as the chip firm or software company that figures out how to write code to take advantage of them stands to make boatloads of money.

DreamWorks signed a deal with Intel this week aimed at parallelizing some of its code running on multicore chips to enable 3-D imaging for the 2009 animated movie. It’s not the first company to work with Intel to get more out of the multiple cores now embedded in servers, but it’s a nice example of how Intel is pushing its multicore efforts beyond simply throwing a bunch of chips at a computing bottleneck.

Like other chip firms, Intel knows that to keep compute power on the rise (and customers happy) it has to not only make the hardware more powerful with multicore chips, but also teach programmers how to use them. Otherwise, multicore chips don’t reach their full potential. James Reinders, director of marketing for Intel’s developer products division, pointed out that much of the work Intel was doing with regard to multicore, including investing in software research, selling tools to make parallel programming less cumbersome and participating in standards bodies, was done to deliver more computing power — something that can no longer be done efficiently by increasing clock speeds or adding even more cores.

“Every generation of hardware offers new capabilities, and we have rewritten our software to take advantage of it over time,” Reinders said. “Multicore will inspire us to do the same thing, but it won’t be overnight.”

It’s possible that the chip companies will be the vanguards of a new style of programming, much like programmers had to learn how to program for the web, graphical user interfaces or even e-commerce applications. Paula Richards, director of IBM’s Cell systems business thinks so. The Cell processor, designed for the Playstation 3, contains nine cores and also performs better if you adapt the code to take advantage of it.

So far IBM has focused on selling the Cell processor into financial firms, hospitals, and oil companies like Spain’s Reposal Repsol, which it inked a deal with last week. Richards said IBM doesn’t just dump that hardware and run — it spends time working with clients in each vertical to build software development kits the customer can use to get the most out of the processor. Those kits work with Intel and AMD multicore chips as well, although Richards says a user won’t see the same level of improvement they would using Cell processors.

“We knew multicore was a major inflection point in the industry,” Richards told me. “Everybody realized this and the company that addresses the [ease of programming] for this technology will win.” In some ways it’s not only about making it easy, it’s about attracting the hearts and minds of developers to a certain way of coding. That’s why IBM is offering SDKs to students who want to write parallel code on their PlayStations and Intel is pushing an undergraduate curriculum for parallel programming. This is a hardware battle fought using software.

image courtesy of DreamWorks

The iPhone Makes Semiconductors Fun Again!

Thu, 03 Jul 2008 12:55:02 -0700

For a while there, covering the chip industry was like covering a race run by a rabbit and a cheetah. AMD was the rabbit, while Intel — with its much larger market cap and greater profits — was the cheetah. Evey now and then the rabbit would fool you into thinking he was going to pull ahead, but we all knew who was going to win. In the past few years, however, two things have brought more runners and more diversity to the course: a challenge to the x86 architecture, and the iPhone.

I could probably find a way to credit the iPhone for changing the furniture industry if I tried hard enough (it could be the new Six Degrees of Kevin Bacon game for tech journalists.) But in this case the iPhone pushed the real Internet — as opposed to a carrier-defined portal — out to mobile consumers and showed them how compelling such access could be. That made clear to carriers that data usage, which was already on the rise, could become a huge revenue booster if consumers were given the right type of devices. Which prompted chip makers to see gold in the form of the 33.2 million high-end handsets sold around the world.

That pushed the chip world into viewing these devices as mini computers requiring their very own processors. Obviously these processors need to be small, use very little energy and still cycle fast enough to load and display web pages, pictures and other mobile computing tasks. Chip firms had been thinking about those functions for years, but the success of the iPhone showed how important the mobile computing experience could be. So Intel begat Atom, a chip designed not for a mobile phone but for a smaller laptop that Intel calls a mobile Internet device.

Other chips firms aren’t standing still, either. Via Technologies, which for a long time had the handheld computer market to itself, is refreshing its line of chips. Qualcomm now has Snapdragon, and Texas Instruments is offering OMAP chips. The dark horse in all of this frenzy comes from Nvidia’s Tegra offering, which is really compelling in demos. But Nvidia has an uneven record of supporting its products, so it remains to be seen if the real-life experience can meet the high expectations set by the demos.

Nvidia is also making my chip coverage fun with its efforts to knock out the x86 architecture. Intel and AMD dual-, triple- and quad-core chips will never go away, but both Nvidia and IBM are pushing credible alternatives for high-end processing. Nvidia’s dressing up its graphics processing chips (GPUs) to run scientific queries, visually intensive tasks and repetitive problems than can be done in parallel, such as video decoding and encoding. The influx of digital media is creating a need for such capabilities in an increasing number of data centers.

IBM, meanwhile, is pushing its Cell processor — which was designed with Sony and Toshiba eight years ago for the PlayStation 3 — for enterprise servers and high-performance computing. In many ways it’s attacking the same problems Nvidia’s GPUs are, with encoding and Monte Carlo simulations showing off the Cell’s specially designed, nine-core architecture. IBM may have an advantage over Nvidia because of its enterprise focus. It offers an enterprise-ready Cell-based blade server, while Nvidia sells its chips to firms such as Atrato and Rackable for corporate consumption.

So the two-company race that was never all that competitive has turned into several races with multiple players. Ironically AMD doesn’t have a mobile processor yet, and isn’t really pushing its GPUs into jobs other than running graphics. Perhaps it believes that if it stays the PC course it can pass the cheetah while Intel focuses on Atom and smaller devices.

AMD Already Missed the MID Boat

Thu, 19 Jun 2008 17:20:30 -0700

OK, so AMD refuses to comment on rumors that it plans to introduce a low-power chip aimed at the mobile Internet device market, where it would compete with Intel’s Atom chipset and offerings from several other rivals. And it refuses to claim a block diagram floated by eeepcnews.de as its plans for such a chip.

I was kind of hoping AMD might stay out of this MID market opportunity and focus on its core CPU business and getting its promising graphics processor and CPU platform off the ground instead of chasing Intel, Nvidia, Via, Qualcomm and Texas Instruments and their hopes for a pocket PC market. Plus, AMD’s been here and done that — back in 2002, when it purchased Alchemy Semiconductor and its line of MIPS-based, low-power personal device chips. That deal was a response to Intel’s Xscale assault, and AMD turned around and sold the Alchemy line in 2005.

AMD did, however, keep the low-power x86 chips for the embedded and personal device market that it purchased from National Semiconductor in 2003. The x86 architecture was more familiar to AMD’s existing chips, and the Geode line is still used in low-power devices, but isn’t very fast and wouldn’t be competitive for the MID opportunity.

It’s surprising that AMD doesn’t have anything better on offer already. Especially given AMD CEO Hector Ruiz’s 50×15 project, which aims to get computers and broadband to half of the population by 2015. An AMD-designed, low-power, high-performance chip would have been perfect for the project and then later for the MID market. However, the One Laptop Per Child laptops AMD is using for the project use a Geode processor. If this diagram represents AMD’s answer to Intel and the gang, why the heck has it waited so long? They had a perfect market for an MID chip and they let it pass them by. If anything, AMD could have sacrificed short-term profits for large volumes if it had to. Its main rival isn’t shy about doing that.

Multicore’s Not-So-Secret Problem

Thu, 19 Jun 2008 00:21:42 -0700

Parallel processing isn’t just for supercomputers or GPUs anymore. Computer makers are throwing multiple cores at everything from servers to your printer. But the focus on horsepower misses a crucial problem associated with adding more processors. To really take advantage of them, you have to rewrite your code.As anyone who’s ever hosted a demolition party well knows, you can only throw so many workers at a problem before people start to linger at the edges, swill your alcohol and generally stop helping. You need not just manpower, but a good way to organize those workers so that someone, says, preps a drop cloth before your walls get taken out. And others prep for cleanup while the plaster is flying.

Silicon doesn’t tend toward drunken destruction, but if you’re putting the cores in place, it would be great to give them better instructions. Otherwise the promise of performance is just a promise, which is why Microsoft and Intel recently pledged $20 million to two universities trying to figure out an easy way to translate the billions of lines of code into an instruction set for multicore chips.

Others are pushing Erlang as a potential solution to parallel programming, while those in the supercomputing industry are warning of a performance drop caused by applications not keeping up with the cores. Software startup VirtualLogix is trying to use virtualization software to govern how multicore chips run applications by making the programs think they’re running on one processor.

Last week, during the launch of the iPhone, Steve Jobs told the New York Times that the next generation of the Apple OS will not focus on new features, but will instead solve the problem of writing software for multicore processors. Apple has code-named the technology Grand Central, and based it on a programming language called OpenCL. It will parallelize C programming languages for graphics processors.

Besides investing millions of research dollars into the search for a magic compiler or reviving an older language, chip vendors are coming up with stopgaps. Unfortunately these stopgaps are focused solely on their own silicon. Nvidia has released a tool called CUDA to help translate C languages into parallel instructions that can be used by Nvidia’s GPUs for scientific computing. (Apple’s OpenCL looks similar to CUDA.) And AMD also has its own effort, called Stream.

Freescale on Monday announced a set of multicore embedded processors that come with software support in the form of a simulator that ships before the chips do. As a result, users can start their development efforts and test their multicore code weeks ahead of time. “Customers are not looking for suppliers to offer them a chip and then leave them to program it themselves,” explained Steve Cole, a systems architect for Freescale. “There’s a certain amount of support and market knowledge that we need to have to help our customers.”

With all the work it takes to rewrite code, it’s no wonder everyone from startups to established companies are desperately searching for the programming equivalent of a Babel fish to solve the problem. The one that succeeds will be responsible for taking computing to its next jump in speed.

Solarflare Gets $26M for 10 GigE

Tue, 17 Jun 2008 06:20:31 -0700

Solarflare Communications, a chip startup in Irvine, Calif., has raised $26 million in a third round of funding. That brings the total the company’s raised to $126 million, which is a lot of money for a chip startup, even when you consider that the amount includes money raised by Level 5 Networks, which Solarflare acquired in April 2006. But the startup is hoping to use that money to attack a big problem in the data center at prices lower than the current technology offers. And if it succeeds, it’ll make computing faster and data center operations more flexible.

Like many other communications chip companies, Solarflare is working on a way to deliver 10 Gigabit Ethernet over copper, which is cheaper than delivering it via fiber. That enables the high-speed transport technology to move outside of the telecommunications networks, where companies such as Infinera are already pursuing 100 Gigabit Ethernet over fiber, and into mass adoption in the data center. Getting the technology into servers at a reasonable cost would create a market 10 times bigger than that of networking switches.

Others chasing mass adoption of 10 GigE on the server side are Intel and Broadcom, which like Solarflare, have controller chips. Broadcom and Solarflare also have PHY chips sampling with customers. Solarflare CEO Russell Stern plans to integrate the PHY with the controller chip in 2009, beating Broadcom to the market. He will use some of the funding for that purpose.

It’s likely Broadcom will end up attempting an integrated 10 GigE over copper chip as well. Broadcom doesn’t talk about its chips until they’re sampling, but the company did make a mint by cornering the market for integrated 1 Gigabit Ethernet chips for servers. However, success for Solarflare or Broadcom is probably three years out and depends on creating an energy-efficient chip at the 32 nanometer process node, according to Bob Wheeler, an analyst at The Linley Group.

Power consumption is a big challenge for these chips because unless it’s managed properly, they run too hot for servers and switches. And because technology doesn’t stand still in the data center, where virtualization and ever-increasing amounts of data are screaming for fatter pipes, hybrid forms of networking technologies that mix fiber or Fibre Channel with Ethernet are emerging to bridge the Gigabit gap between servers and networking equipment. Broadcom has several products that take advantage of such a hybrid networking environment. Startups such as Arastra and Woven Systems are also in that sector, and may see gains at the expense of a unified 10GigE world, which means Solarflare’s market opportunity could fragment if cheap, integrated 10 GigE takes too long.

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

Solarflare Gets $26M for 10 GigE

Tue, 17 Jun 2008 05:20:51 -0700

If this story interests you then you should definitely check out our upcoming conference, Structure 08.

AMD Faces Nvidia With Dual Chip Plan

Mon, 16 Jun 2008 08:26:41 -0700

Nvidia and AMD today each launched two graphics chips for the PC market — but the two companies are pursuing divergent strategies. Both share a recent focus on high-end graphics, which underlines how important visual computing has become; but the different approaches taken by each firm may cost Nvidia market share if its monolithic high-end chips can’t deliver the graphic punch to compete with a multi-GPU strategy embraced by AMD and Intel.

Nvidia launched its GTX 280 and GTX 260 chips, which are larger multi-core processors on a single chip. AMD on the other hand, has taken a bottoms-up approach with smaller, multi-core chips that can be harnessed to a second graphics processing chip on a board to deliver higher-level performance. Lower-end PCs can rely on one AMD processor and those needing more power can turn to two AMD chips or Nvidia’s single, high-power chip.

The real question is how the graphics will look on the screen. And, as in most chip releases, the proof will be a while in coming. Nvidia already has HP signed up to use its new chip in a new Voodoo desktop especially for gaming. That makes sense. Nividia’s chip will rock the high-end application, while AMD’s is designed to provide compelling imagery for cheaper, power-efficient PCs and laptops at a large scale. The real battle will be whether AMD’s dual-chip strategy takes business away from Nvidia for specialty graphics computers and high-performance technical computing. If that occurs, Nvidia will have to be on guard: Intel’s planning to follow the same dual-chip path with its Larrabee GPUs.

Ouch. Intel to Face Formal FTC Probe.

Fri, 06 Jun 2008 10:55:06 -0700

The Federal Trade Commission, after two years of looking into allegations that Intel has behaved anticompetitively in the microprocessor market, has decided to act, announcing a formal probe. At issue is whether Intel offered PC makers rebates to use its chips instead of AMD’s. Intel issued a statement in response.

The company believes its business practices are well within U.S. law. The evidence that this industry is fiercely competitive and working is compelling. For example, prices for microprocessors declined by 42.4 percent from 2000 to end of 2007. When competitors perform and execute the market rewards them. When they falter and under-perform the market responds accordingly.

In Austin, the Intel fund at Dell was an open secret, although Dell eventually opened the door to AMD. While AMD may be tempted to applaud this and the $25.4 million fine imposed on Intel by South Korea, the FTC probe won’t lead to action anytime soon. The government moves slowly and the coming change in administration won’t help speed it up.

AMD Pushes Puma to Maul Intel

Wed, 04 Jun 2008 06:51:41 -0700

AMD’s Sisyphean task of grabbing market share from Intel begins anew with the launch of its latest line of laptop chips laptop platform formerly code-named Puma. Today, AMD launched a refresh of its Turion mobile processor combined with an integrated ATI graphics processor, designed for mobile use form the ground up. AMD also announced it would provide a discrete graphics processor that could work in conjunction with the integrated graphics processor to boost performance.

Puma will both help AMD compete with Intel again in the still growing laptop market and justify the company’s $5.4 billion acquisition of ATI Technologies back in 2006. As graphics become more important to the PC user, both Intel and AMD are shoring up their expertise in that department. AMD bought ATI, while Intel is pushing its own platform strategy with in-house graphics processing.

The Puma platform will launch in laptops from Toshiba, NEC, HP, Asus and Acer. Lucky for AMD, Intel’s planned upgrade to its Santa Rosa laptop platform — the Monetevina platform — has been delayed until July, giving AMD a few-month head start on wowing consumers and the back-to-school buyers.

Ozmo Teams With Intel to Target Bluetooth

Mon, 02 Jun 2008 17:24:28 -0700

Wi-Fi is the coax of the wireless world in that it’s cheap, is in a lot of homes and is familiar to consumers. So today’s launch of Ozmo Devices, with backing from Intel and Belkin, should strike not a small amount of fear into the hearts of Bluetooth SIG members. Ozmo makes software that uses the existing Wi-Fi chips inside a computer or laptop and allows that laptop to communicate with battery-operated peripherals containing its chip.

From the user perspective, this will eliminate USB dongles for communicating with your wireless keyboard, mouse, etc. It also allows for applications that Bluetooth, with its limited bandwidth, can’t do well, such as sending uncompressed stereo to wireless speakers.

Ozmo doesn’t currently have peripherals on the market, but Belkin has said it plans to use its chips in products later this year. Intel is also pushing Ozmo as part of its Cliffside project, which aims to build a chipset that can distinguish between Wi-Fi signals for local area networks (LANs) and personal area networks (PANs). Cliffside won’t only pick a fight with Bluetooth, but will be a blow to the underdogs in the wireless USB space that are seeking to use ultra-wideband as a wireless standard for sending large files across relatively short distances. If Intel starts pushing Cliffside in a big way, expect to see some PANdemonium.

Nvidia Joins The Ultra Mobile Computing Party

Mon, 02 Jun 2008 07:49:15 -0700

As we said they would a few weeks ago, Nvidia today showed off its line of Tegra chips designed for mobile Internet devices, becoming yet another entrant into the unproven market.

The Tegra chipsets are based on the APX2500 processor built for personal media players and navigation devices, but the Tegra target will be portable computers with screen sizes ranging from 4 to 12 inches. Pay close attention to news coming out of the Computex trade show in Taiwan this week, where more details should emerge from vendors using the Tegra chipset. Products based on Tegra will be out in time for the holiday season at the end of the year and cost about $200 to $250.

Also in the run-up to Computex, Intel’s CEO Paul Otellini told to the Financial Times his firm’s Atom chips (also aiming at MIDs) will chase $40 billion in market opportunities; Taiwanese computer vendor Asustek said it expected to double sales of it’s tiny Eee PCs in 2009 over this year. Even Dell is getting into the fray with a small computer. As products emerge, I’m eager to see how the market for the devices breaks down. Right now, the market opportunity is large because it’s ill-defined, with each vendor suggesting its own specs as the defining standard.

Will MIDs be small computers with voice as Otellini seems to think; phones with faster processing and media capabilities like Qualcomm, Apple and TI seem to envision; or will they be lightweight computers like the MacBook Air, Eee PC or what I bet the Dell effort is?

TI Joins the Portable Internet Device Race

Tue, 13 May 2008 13:18:29 -0700

No one knows exactly how big the market for mobile Internet devices will be, but the major chip makers are betting it will be huge (it’s one of the reasons they’re making chips for mobile devices at 45 nanometers.) We’ve covered efforts by Intel, Qualcomm, and Via Technologies to get their chips into devices sized somewhere between a smartphone and a PC, but Texas Instruments wants to play, too.

TI formalized its MID effort, based on its own OMAP architecture, last month. It’s entering this market with its third generation of OMAP multimedia processors, which were designed four years ago specifically to fit into smartphones. The second-generation chips are currently in the Nokia 800 and 770; the third-generation chips that underlie the formal MID group will be in an undisclosed number of products by the end of the year.

TI’s chips will compete directly with Qualcomm’s Snapdragon chipset and Intel’s Atom chips. Comparatively speaking, TI’s chips show a greater flexibility for the end products. The power-sipping (at 500 mW-750 mW) 800 GHz MHz processor is slower than both Qualcomm’s and Intel’s efforts and requires less power than Intel’s Atom processors, which can require up to 2.4 watts. Ramesh Iyer, a MID product strategy manager with TI, says the lower clock speed is a conscious decision to reduce the power consumption; combining several types of cores with TI software allows for a higher utilization of existing megahertz, he notes.

As products containing chips from competing vendors hit the market, my hunch is that TI’s might be the best when it comes to general purpose use and battery power, followed by Qualcomm’s Snapdragon, which will also be battery-friendly and perhaps perform better than TI’s in general purpose use. Device specs for MIDs based on Intel’s Atom processor are larger, but the x86 architecture might win converts because it’s familiar and plenty of applications are designed for it. And that raises the very legit question of what role the operating system will play in how MIDs are used. I’ll get back to that in a few posts.

HP Turns to Lasers to Cut Copper From Chips

Mon, 12 May 2008 08:18:49 -0700

HP is trying to eliminate copper on semiconductors to make them run faster, and today the company is gathering about 150 researchers at its Palo Alto campus to push lasers as a means to do this. If it and chip manufacturers such as Intel, IBM and Luxtera succeed, the chip firms will follow in the telcos footsteps, turning to light to transmit information quickly.

Only, in this instance, the light would provide short-haul transport on a chipset measured in nanometers or millimeters rather than over distances of miles. Lasers could replace the copper connecting multiple processing engines inside a chip, but could also act as interconnects between multiple chips on a board. Light pulses provided by a laser could reduce the bill of materials (if adapted for silicon), power consumption and solve some of the problems associated with following Moore’s Law because it reduces some of the materials needed on a chip. Improved chips mean more computing power and a faster, more dynamic web.

Such efforts are in the early stages with real products likely 10 years out. However, it isn’t so far-fetched. Already Infinera, a Sunnyvale, Calif.-based company whose products are sold to telecommunications companies, makes an optical chip, but it builds its chips on a far more expensive substrate than a silicon wafer.

Interop Vegas, Land of the Inexpensive and Powerful

Wed, 30 Apr 2008 10:06:19 -0700

I’m here in Las Vegas for the annual Interop show for IT professionals, and I’m finding it to be far livelier — and better attended — than I had expected it would be. In fact, I may need to rethink my belief that Web 2.0 has killed the networking engineer, as hordes of my compatriots are here, engaging in lively hallway discussions and even attending an Unconference.

While the activity here at Interop may show that IT is still a thriving business, there’s one trend that everyone I’ve spoken with so far has been observing: Inexpensive and powerful hardware is transforming IT networking infrastructure in ways we never would have conceived, even a few short years ago.

Commodity computing is dominating the show floor, with Intel processors not just in the server and desktop, but outside of it, in places including appliances serving as routers, load balancers, storage area network controllers, firewalls, application delivery controllers and so forth.

This makes intuitive sense if you realize that as a vendor you can easily find inexpensive hardware that can route multiple full-duplex Gigabit Ethernets per second, have memory that can hold a routing table twice the size of the entire Internet, implement firewall rules for every host in a typical organization — and not tax the Intel CPU with more than 10 percent load. If networking appliances are using something other than Intel processors, it’s often merchant silicon, from companies such as Broadcom, Cavium, Marvell and Nvidia.

As far as I can tell, only the highest-end networking appliances that serve multiple, 10-Gigabit Ethernet speeds are using custom silicon from specialized vendors. The market size for these highest-end appliances in the enterprise IT environment appears to be fairly small, even if the most optimistic bandwidth predictions come true.

With inexpensive and powerful networking hardware dominating the IT landscape, we may be ready for a shift in networking infrastructure. As an industry, we’ve been taught for over a decade about the three-tiered network design — access, distribution, core — but it strikes me that, with the commodity compute resources available today, this may need to be reconsidered. Networking engineers are already struggling to conform some of today’s modern technologies, like server and application virtualization, onto existing infrastructure design. Virtualized routers and firewalls are either in your network today or are just around the corner. Given the hardware that’s currently available to the networking industry, do we need a new network design to handle these new technologies? Dare I suggest that, with the processing power available today, some older networking protocols (IEEE802.5 with priorities or ATM with any-to-any direct communication) may be more relevant today?

If you know of a company here at Interop that is leveraging commodity compute to help transform network infrastructure, please let me know — I would be very anxious to meet with them.

Interested in web infrastructure? Check out our upcoming conference, Structure 08.

Cray Gives Intel a Chance

Mon, 28 Apr 2008 14:13:54 -0700

Poor AMD. Cray has decided to build its next generation of supercomputers around Intel’s Xeon chips. Cray’s previous line was built around AMD’s Opteron processor, which was introduced in 2003, but AMD’s latest quad-core chip — the Barcelona — has been delayed. Into that delay has stepped Intel, with an ambitious effort to not only provide chips, but also to build an R&D partnership around the future of supercomputing centered on Intel’s multicore processors. With Cray bringing on such a potent partner (Intel has almost 71 percent of the market when it comes to providing processors for the Top 500 supercomputers) other chip vendors such as IBM should take note, too. Chart image from Top500.org.

Buffalo Can’t Roam But Still Charges Ahead

Sun, 27 Apr 2008 00:13:03 -0700

Patent lawsuits always seem to be one of two things: Little more than a slight annoyance, or a business-ending death blow. Rarely does the tech world see companies who resemble Timex watches in their ability to take a patent lickin’ and keep on tickin’. But the U.S. headquarters of Buffalo Inc. is one such entity.

In Japan, the company reports sales of about $1.3 billion a year. Yet it generates a mere $100 million of revenue out of the U.S., where it offers four types of products, two of which (Wi-Fi routers and flash memory devices) it currently can’t sell because of court injunctions. Another line — multimedia — is new, with the first product due to hit the shelves in June. Its best-performing line is storage, which is profitable despite the fact that the company buys the basic drives from its competitors. All in all, it reads like a prime candidate for business failure. But so far, Buffalo is making it work.

Let’s start with wireless, since the story there is pretty simple. Four years ago, Buffalo started selling 802.11n routers in the U.S., going up against Linksys, D-Link and Netgear. As the smallest player in the market it was first hit with a patent infringement lawsuit from Australia’s Commonwealth Scientific and Industrial Research Organisation. Last year, a court sided with the Australian firm and ordered an injunction against Buffalo’s routers, despite protest (and amicus briefs) from Netgear, 3Com, Atheros, Dell, Intel and others. The case is being appealed, but in the meantime Buffalo can’t sell Wi-Fi devices in the U.S. and the 802.11n IP is still up in the air.

Patent cases have stymied Buffalo in its flash memory business as well. Last month Buffalo stopped selling USB drives and memory cards (its first line of business in the U.S., which it acquired over a decade ago), due to a patent infringement lawsuit filed by SanDisk against it and several other industry players.

Legal fun aside, Buffalo has established itself as a well-regarded provider of storage for the small- to medium-sized business market and, to a lesser extent, consumers. Buffalo has a pretty loyal following for its network-attached storage products, but the company has to purchase the hard drives from rival firms Seagate and Western Digital since, like other smaller storage vendors, Buffalo doesn’t manufacture its own. Making hard drives is a competitive business where economies of scale are important.

Buffalo adds applications and other features to its storage products to make them more compelling at what is generally a higher price point than those offered by Seagate and Western Digital, but storage is a commodity product, one in which cost-per-gigabyte is a customer’s primary consideration. Regardless, Buffalo makes money on each of its storage devices, so while the fact that its success in storage puts money into the pockets of its competitors pocket is galling, it doesn’t signal the end of that business for the firm.

Storage and its single product for multimedia streaming (wired, because it can’t sell wireless in the U.S. right now) are the cards Buffalo currently has to play, and the Austin, Texas-based Buffalo USA intends to play them for all it’s worth. Patent fights or no.

With iPhone In Mind, Apple Buys Chip Maker

Tue, 22 Apr 2008 22:22:00 -0700

Apple has acquired PA Semi, microprocessor design firm for $278 million in cash, reports Forbes’ Erika Brown. PA Semi was started by Dan Dobberpuhl, a chip designer closely associated with Alpha and StrongARM chips developed by Digital Equipment.

The decision to center the iPhone design around a chip that Apple could own marks a significant strategic choice by Apple Chief Executive Steve Jobs, and is aimed at ensuring Apple can continue to differentiate its flagship phone as a raft of competitors flood the market. According to a source affiliated with the chip company, Jobs and Senior Vice President Tony Fadell led the tiny group of executives who spearheaded the acquisition, which included negotiations that took place in Jobs’ home.

Apple’s decision is going to post a problem for Intel Corp. and its newly announced Atom chip. It is unlikely that Intel’s chip was going to find room in the handsets made by some of the larger players.

Chip industry insiders believe that Intel was betting on an Apple win to gain scale for Atom which in turn would allow it to dominate the “portable internet device” market. iPhone and iPod Touch are the early leaders in the PID category, and are unlikely to cede that spot for near foreseeable future.

PA Semi had designed a 64-bit dual core chip that consumer between 5-to-13 watts running at 2 gigahertz, making it a good chip for the PID category. So far, PA Semi has found takers in telecom equipment makers.

Beyond3D puts context on the news announcement. While they are mostly right about everything, I don’t think you can rule out the iPhone argument.