Maximum PC fast forward RSS Feed

Fast Forward: Parallelism

Tom Halfhill — Mon, 23 Nov 2009 12:30:12 -0600

After decades of fitful progress, parallel processing is suddenly hot and will soon be commonplace on ordinary PCs. For applications rich in data-level parallelism, performance is soaring by leaps and bounds.

Multicore CPUs from Intel and AMD are all good, but the game-changers are the next-gen GPUs from Nvidia and AMD/ATI. These chips are evolving from highly specialized 3D-graphics processors for games into broader computing engines for nongame software. Nvidia is leading the charge with a new GPU architecture that, for the first time, supports general-purpose computing as strongly as it supports graphics.

Nvidia’s new Fermi GPUs will support error-correction codes (ECC), one terabyte of memory, concurrent kernels, and faster double-precision floating-point math. These features are largely unnecessary for 3D graphics but vital for high-performance general-purpose computing. (In fact, ECC slows down graphics processing, which is why it can be disabled in Fermi chips sold for the consumer market.)

With Nvidia’s CUDA development tools, programmers are accelerating some tedious media-processing tasks, such as video transcoding. CUDA uses the GPU’s programmable 3D-graphics shaders as massively parallel processor cores, delivering performance that today’s PC processors can’t match. In addition, GPUs are finding new applications in scientific computing, financial analysis, medical imaging, energy exploration, and engineering.

Other developments are equally exciting. Microsoft’s DirectCompute brings a parallel-processing API to millions of mainstream PCs running Vista and Windows 7. The new OpenCL standard makes parallel programming easier and less proprietary. Apple’s Snow Leopard (Mac OS X 10.6) supports OpenCL and Apple’s Grand Central Dispatch technology (now open source), allowing programmers to distribute workloads across multicore CPUs and GPUs.

Intel is busy, too. With its own new GPU (Larrabee) on the way, Intel has acquired two small companies specializing in software tools for parallel programming—RapidMind and Cilk Arts. RapidMind is especially cool, because its software bridges GPUs, multicore x86 processors, and even IBM’s Cell Broadband Engine.

Parallel processing is spreading to the masses, and parallel-programming tools are catching up with parallel-processing hardware. When these trend lines finally converge, we’ll wonder why it took so long.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: Picoprojectors

Tom Halfhill — Sun, 01 Nov 2009 14:00:12 -0600

In August, Nikon introduced the world’s first digicam with a built-in video projector. The Coolpix S1000pj has a tiny projector—called a picoprojector—that can display photos and videos at 640x480-pixel resolution. In a dark room, projected images are visible up to six feet away, up to 40 inches wide.

Although picoprojector technology has been appearing in small video projectors and a few other devices, the S1000pj moves this revolutionary technology into a mainstream consumer product. Soon, “embedded” picoprojectors will be everywhere.

An embedded picoprojector is one that’s built into a device other than a stand-alone video projector. Digital cameras, video camcorders, and camera-equipped cell phones are obvious candidates. Embedded picoprojectors will probably become as common as webcams in notebook computers. Hand-held videogames, media players, portable TVs, and ebook readers are additional possibilities. Picoprojectors will be used for advertising displays, vehicle entertainment systems, heads-up control panels, and other applications that can benefit from their space-saving properties.

Several companies have been working on picoprojector technology for years. Texas Instruments uses LEDs with millions of movable micromirrors to project the image by reflection. Microvision uses red, green, and blue lasers with an oscillating mirror to scan the image onto the screen. Nikon uses liquid crystal on silicon (LCoS) technology instead of micromirrors, and LEDs instead of lasers. The picoprojector in the Coolpix S1000pj has a brightness rating of 10 lumens.

By comparison, my Kodak Carousel slide projector from the 1970s is rated at 525 lumens. It easily projects a 35mm slide onto a five-foot-wide screen from 20 feet away. But the bulb dissipates a scorching 300 watts—enough to melt the little Nikon camera in minutes.

Because picoprojectors rely on the much smaller, cooler light sources of LEDs or low-power lasers, they can’t yet match the brightness of old-fashioned projectors. But they have an advantage that, in time, will make up the difference—they concentrate all their light into a highly directional beam. A slide-projector bulb radiates light in all directions, wasting most of it.

Embedded picoprojectors are a game-changing technology, like LCD screens. Someday we’ll wonder how we ever lived without them.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: Chrome Won't Shine

Tom Halfhill — Wed, 07 Oct 2009 14:00:25 -0500

If there were such a thing as post-traumatic stress disorder for weary veterans of OS wars, I’d have it. Frightening flashbacks of MS-DOS vs. CP/M... Windows 3.0 vs. Apple System 6... OS/2 vs. Windows NT... Windows vs. Mac again... then Linux vs. Windows vs. Mac. And that’s not counting the smaller conflicts that engaged OS-9, CP/M-86, AmigaDOS, and others too numerous to mention.

Now Google’s Chrome OS is challenging Windows? Please.

Look, I’ve railed at Microsoft as much as anyone, sometimes in these very pages. And my other computer is an iMac. But one thing I’ve learned is that a new OS needs a strategic advantage before it can defeat a deeply entrenched OS.

MS-DOS overwhelmed CP/M in the 1980s because Microsoft’s OS was basically free with the 16-bit IBM PC, which was more powerful than 8-bit CP/M machines and legitimized personal computers in corporate America. The Mac’s graphical OS debuted in 1984 with a huge advantage in usability over the primitive MS-DOS command line, but Apple’s reluctance to license Mac clones allowed Microsoft to retaliate with Windows and win 90 percent market share.

Chrome OS isn’t totally outgunned. With Google’s resources, Chrome OS could become, in effect, the leading Linux distro. A united front would help Linux, which suffers from the same dispersal of forces that weakened Unix. Also, the trend toward cloud computing favors a secure, nimble OS designed to run network-based software. And Microsoft’s adoption of an XML-based format for Office allows challengers to offer file compatibility with the world’s most popular productivity suite.

Unfortunately for Google, Windows 7 looks like a worthy atonement for Vista. The widespread rejection of Linux-based netbooks shows that people still expect a PC to run their familiar Windows software. Although the success of Apple’s App Store proves that a newcomer can build a software library faster than ever before, most iPhone apps are toys, not replacements for serious desktop programs.

To fight Windows, Chrome OS needs heavier artillery or a faster transition to the cloud-computing battlefield than I anticipate. A smarter strategy is a flank attack on Microsoft. Google’s Android OS can capture the new high ground of smartphones and other next-generation computing gadgets.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: The Next Next Thing

Tom Halfhill — Tue, 29 Sep 2009 16:15:07 -0500

It’s getting almost impossible to be a fully equipped techie. There’s always another new gadget threatening to leave you behind, even if you’ve already got a desktop PC, laptop, netbook, home WLAN, game console, e-book reader, smart phone, iPod, GPS, portable DVD, digicam, DSLR, HDTV, HD camcorder, Blu-ray, DVR, dish, and surround-sound home theater.

What’s next? Media phones.

Nope, they’re not smart phones. We’ve already got that. Media phones are next-gen landline phones tethered to broadband Internet service in a home or office. Typically, they have cordless handsets for voice calls and a fairly large (8-inch or so) touch screen. Built-in DSL or Wi-Fi provides fast, always-on Internet access. VoIP can provide cheap long-distance calling. Like conventional phones, media phones needn’t be booted or shut down.

The touch screen offers numerous applications: web browsing, email, network-based address books, phone-directory lookups, news updates, weather forecasts, YouTube videos, music, TV program guides, Twitter, recipes, and quick e-commerce (e.g., ordering pizza or movie tickets). When idle, a media phone can be a digital picture frame, MP3 player, or Internet radio.

Basically, media phones update the old-fashioned home or office landline phone with the same services and applications now appearing on mobile phones, except with bigger screens and no need to recharge batteries. Media phones can be mounted on walls (ideal for kitchens and workshops) or put anywhere conventional phones are found (desks, tables, night stands). Some plug into an ordinary phone jack, others connect wirelessly to a dedicated Wi-Fi router. They don’t need a PC.

Telephone companies are introducing media phones in the same way that DSL modems were rolled out. Your local telco will probably subsidize the phone’s cost in return for a service contract. You may also find media phones sold in stores, probably bundled with service. In offices, they will tie into Internet-based PBX systems.

An ulterior motive is at work here. More people are dropping landline telephones in favor of cell phones and cable services. The telcos hope media phones will reverse this trend. Nevertheless, I can see the usefulness of a media phone. At the right price, this techie would buy one.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: Why Apple Designs Chips

Tom Halfhill — Fri, 28 Aug 2009 17:45:48 -0500

Many people still think of Apple as a relatively small computer company, even though it’s a large consumer-electronics company. Those folks were surprised by recent reports that Apple is hiring more chip designers. They question the wisdom of plunging deeper into the risky and costly venture of designing custom chips.

But Apple’s moves are a logical response to current events. We are witnessing a fundamental shift in computing, as important as the debut of personal computers in the 1970s.

Desktop PCs—and to a lesser extent, notebook PCs—are the old wave. The new wave integrates mobile computing and communications with ubiquitous Internet access. Although notebook PCs can ride this wave, they are the largest species of new personal computers. Netbooks are better examples. Best of the new breed are the Apple iPhone, RIM Blackberry, and Palm Pre. More are coming.

It’s been two years since the wildly popular iPhone appeared, yet major cell-phone makers still haven’t introduced similar models. They were caught flat-footed. The iPhone redefined the smartphone with innovative hardware and software, just as the Macintosh redefined desktop computers in 1984.

For 25 years, Apple’s strategy has been consistent. Apple differentiates its products with custom hardware and software so it can charge higher prices and earn greater profits than commodity-minded competitors using off-the-shelf components.

Small systems like the iPhone need custom chips optimized for specific tasks. These chips burn less power (lengthening battery life), occupy less space (improving portability), and deliver greater performance (enhancing the user experience). It’s true that chip design is risky and expensive, but Apple has the experience, money, and talent required for these projects, and is accumulating more.

No one questions why consumer-electronics companies like Sony, Canon, Matsushita, Toshiba, and Samsung design custom chips. With products like the iPod and iPhone, Apple is joining that league. Apple has always been good at surfing waves—largely because it’s a tidal force helping to make them.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: Intel's Crystal Ball

Tom Halfhill — Thu, 13 Aug 2009 17:30:33 -0500

In a rare example of limb-crawling, Intel’s technical marketing manager recently made 10 predictions for the next 10 years. But he didn’t crawl very far. Most predictions were boring references to long-standing development projects at Intel and elsewhere.

“Realistic computer-generated images.” (Hey, Intel, we’ve already got that.) “New classes of portable devices with 10 times more battery life.” (Who else saw that coming?) “Personal Internet devices will be truly personal.” (Like I’ve been saying for years.) “Low-cost silicon photonics for faster, more reliable data transmission.” (Intel and many others have been working on that technology forever.)

Nevertheless, two predictions are interesting. The boldest was “Malware will become a thing of the past.” The idea is that microprocessors will incorporate security features to stop malicious software from attacking the operating system and application software. It’ll be like a Roach Motel for malware—bugs crawl in, but they won’t crawl out.

Bold as it may seem, that prediction is an extension of well-worn technology. For decades, many processors have been able to protect regions of memory from unauthorized accesses. Operating systems have long been able to run in a privileged mode that lets them perform operations forbidden to regular apps. These features have improved over the years, and they will keep improving. Will they ever make malware impossible? I doubt it. The weakest link is still the user. We need a Moore’s Law for brains.

Another Intel prediction: “Interactive computing devices make ‘composable computing’ a reality.”

The concept is that different wireless devices in close proximity could be linked to share data and computational resources. You could download a video to your iPod and beam it to an HDTV, wirelessly and effortlessly. Some digicams can already share photos wirelessly with PCs and printers, so this prediction isn’t exactly a stretch goal.

Even so, anything that improves interoperability is fine with me. I’ve been computing since the 1970s and have a junk box of obsolete cables you wouldn’t believe.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Foward: Intel vs. ARM

Tom Halfhill — Thu, 16 Jul 2009 11:17:54 -0500

Intel, the world’s largest semiconductor company, suffers from a Freudian case of appendage envy. The appendage is an ARM.

Simply put, smartphones (and other mobile consumer-electronics gizmos) are the next PCs, and Intel wants them to run on Intel x86 processors. Right now, your mobile phone, MP3 player, or digicam probably has a custom chip with a microprocessor core licensed from ARM. Although most people have never heard of ARM, it makes the most popular 32-bit microprocessor architecture in the world.

Yet ARM doesn’t make a single chip. It licenses its 23 different processor cores to other companies that design and make chips. These chips are very different from most of Intel’s. They are system-on-chip (SoC) devices—highly integrated chips that surround the processor core with built-in peripherals, memory, I/O interfaces, and application-specific logic.

Intel knows it needs SoCs to conquer mobile electronics. Using separate chips is too costly, burns too much power, and requires too much space. Although Intel makes a few SoCs, it’s impractical for one company—even one as large as Intel—to make a different SoC for every gadget.

ARM is much smaller than Intel. But ARM’s strength is its global army of licensees, who make an awesome variety of SoCs. Intel fears to license the x86 in the same way. The last time Intel licensed the x86 architecture was in the 1980s—to AMD.

Intel’s latest solution? Offer to design Atom-based SoCs for makers of smartphones and other small systems, then use Taiwan Semiconductor Manufacturing Company, the world’s largest independent chip manufacturer, to fabricate the chips. Intel will provide design specs and custom logic. TSMC can provide the integrated peripherals, memory, and interfaces. Presto! An x86 SoC.

This arrangement isn’t as flexible as ARM’s licensing, but it’s a big step for a company as paranoid as Intel. The big question is whether having an x86 instead of an ARM processor in a mobile device matters to anyone but Intel and ARM. In netbooks, it matters—because netbooks are PCs, and most people want Windows. In an iPhone, who cares? I think Intel will have trouble muscling in on ARM.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.

Fast Forward: Fretting over Netbooks

Tom Halfhill — Mon, 06 Jul 2009 08:30:00 -0500

The recession is getting so bad that stock market refugees are snapping up Treasury bills at 0.2 percent interest, and car dealers have tried everything but adding immortality to their option packages. So you would think that a hot-selling product would be universally welcomed.

Netbook computers are a rare bright spot in a dimming economy. They’re selling faster than copies of Foreclosure for Dummies. The Asus Eee PC opened the door. Now there are too many to count.

However, critics say netbooks might be a bad thing. Their reasoning is that most netbooks use Intel’s Atom processor, which costs less and has lower profit margins than Intel’s other mobile processors. Atom’s popularity, they say, might actually hurt Intel and drag down profits for system vendors and their suppliers.

Enough of that. Netbooks are a good thing. In the first place, market surveys indicate that netbooks aren’t displacing notebooks. Most buyers either have a notebook already and want something more portable, or they weren’t considering the purchase of a mobile computer at all until netbooks came along.

Of course, the surveys could be wrong or premature. I’m sure some people are bypassing traditional notebooks for smaller, lighter netbooks. But the choice isn’t easy, because most netbooks aren’t much cheaper than full-featured notebooks with superior screens. Intel, genetically paranoid, is carefully positioning netbooks as less-capable machines suitable for casual email, web browsing, and social networking. When Nvidia recently tried to expand the scope of netbooks into gaming by introducing a chipset with better graphics, Intel responded with aggressive countermarketing. Intel is also trying to limit the screen sizes of netbooks.

It’s inevitable that netbooks will cannibalize sales of larger computers to some degree. Computing is going mobile, a trend no one can stop and that wise companies will exploit. Intel is promoting a “new” class of mobile Internet devices (MIDs)—basically, PDAs reborn. Intel is also pushing Atom into smartphones.

The success of netbooks, Apple’s iPhone, and wireless networking show that people want the Internet wherever they go. If larger notebooks can’t make the grade, too bad. Resistance is futile.

Tom Halfhill was formerly a senior editor for Byte magazine and is now an analyst for Microprocessor Report.