Wednesday, August 20, 2014

Robert Cringely blasts Memristor 'Vaporware'

Robert Cringely (pseudonym) is well known in the Valley, and in general very well respected for calling it 'like it is'     He historically was a huge HP fan, a fervor dissipated over the parade of CEOs in recent years.  I missed this article in June, but one of our readers pointed it out the other day.  It is angry, dismissive, and perhaps right--time will tell.  But at the least, it is worth reading:

By  | InfoWorldHewlett-Packard has gone and hit it -- the culmination of computing, the pinnacle of processing, the apex of annoyingness. It's solved all our computing problems with the most insightful and provocative move it's made since its VP of Vision passed on the Mac in the mid-'70s. I'm glad I lived to see it.
In fact, HP's news is so groundbreaking, it takes me straight back to the mid-'90s, when I sat through a similar announcement by NetWare executives somewhere in the badlands -- maybe Provo? -- to announce a supposedly soon-to-shake-the-earth, splendiferous project called SuperNOS. Though mostly wondering where they'd hid the open bar, I was mildly interested in what the heck SuperNOS might be. After all, that was my job, and I was nothing if not professional. Finally, a canyon-dwelling lab geek got up to explain, even as he was trembling like a banjo.
[ For a humorous take on the tech industry's shenanigans, subscribe to Robert X. Cringely's Notes from the Underground newsletter and follow Cringely on Twitter. ]
Back then NetWare still differentiated between a PC OS and a network OS. But every competitor, including HP, IBM, and especially that annoying Windows NT, was hot to move in on the company's 80-plus percent market share. Even the techno hermits in Utah knew they had to do something.
Remember SuperNOS? No one else does either
SuperNOS was the answer, combining the fabulousness of NetWare with the unshakable Unixness of UnixWare. The list of future benefits was long and distinguished, but apparently harder to build than NetWare thought. It gave up and sold the whole shamble to SCO in 1995, and we all know how that ended. It didn't matter anyway. Soon after, Linus Torvalds glanced up from his CRT long enough to make "open source" a household phrase and kicked SuperNOS, NetWare, and Darl McBride to the margins of computing history.
The lesson hasn't daunted the marketing brains at HP. If no one's paying attention to you, make some wild press announcements you can almost-maybe back up and enjoy the tweet spikes. Thus, HP has unveiled "The Machine," which boils down to a vision currently in little lab pieces glued together with promises reminiscent of that wonderful SuperNOS announcement so many years ago. According to HP, The Machine is a revolution in computing that will solve all our problems while finally satiating Meg Whitman's need for 80 percent margins.
The contraption will supposedly use fiber optics, memristors, and earnest prayer to increase bandwidth and push processing speed forward in big ol' leaps by essentially doing away with slow disks and moving to a memory-driven I/O system at the OS layer -- I think. I'd partaken of too much "milk" by the time they got to that. The best part? The operating layer is -- wait for it -- an open source-based OS that has yet to be developed.
HP: Breaking new ground in meaningless promises
That's all you got? It's a hardware platform stuffed with cool components you haven't actually integrated yet and an open source operating system tagged with a "coming soon" sign? How desperate for PR are you? Apparently quite a bit, because HP's using these lab avowals to make even bigger promises about how The Machine will eventually be capable of analyzing reams of big data in seconds or handling simultaneous chat sessions between multinational sports teams rejecting Steve Ballmer's overtures. Color me skeptical.
Sure, starting all over again from silicon on up would be a great opportunity to do away with all the PC OS Band-Aids we've had to apply since Wozniak first emerged from his garage, but that'll take more customers than HP has seen since this side of the Compaq years. That won't happen anyway if you're going to base the thing on pre-existing open source code snippets -- pretty much the definition of "Band-Aid." Also, Big Blue has been shipping a similar memory-based OS since the '80s, and I don't see that changing our lives.
This is, after all, coming from the same people who invented the revolutionary RISC chip and have remained on the bleeding edge of OS design with HP-UX and VMS. Ten gets you 20 that Linuxians will eat this, too, and have a Machine-capable distro out before HP's fantasy OS even hits beta.
This article, "Hewlett-Packard's 'Machine': Vaporware, meet empty suit," was originally published at Follow the crazy twists and turns of the tech industry withRobert X. Cringely's Notes from the Field blog, follow Cringely on Twitter, and subscribe to Cringely's Notes from the Underground newsletter.

3Q results for HP

Steve Johnson at the San Jose Mercury-News posted this article a few hours ago (highlights mine):

PALO ALTO -- Buoyed by a big increase in its personal computer business, tech giant Hewlett-Packard on Wednesday reported third-quarter sales that exceeded what Wall Street had forecast but its profit slumped from the same period last year.
The Palo Alto company said its sales totaled $27.6 billion for the three-month period that ended July 31, up about 1 percent from the same period a year ago. But its profit was just under $1 billion, compared with about $1.4 billion for the third quarter in 2013. That worked out to fully reported earnings of 52 cents per share.

FILE - In this Aug. 21, 2012, file photo, the Hewlett-Packard Co. logo is seen outside the company’s headquarters in Palo Alto, Calif.
FILE - In this Aug. 21, 2012, file photo, the Hewlett-Packard Co. logo is seen outside the company's headquarters in Palo Alto, Calif. (Paul Sakuma/AP Photo)
Analysts polled by Thomson Reuters generally had expected fully reported earnings of 68 cents a share on sales of $27 billion. HP's earnings also were far shy of its own prediction of 59 cents to 63 cents a share. On a measure that excludes certain expenses, however, the company's earnings met analysts' expectations.

"I continue to be very encouraged by the progress we're making," CEO Meg Whitman said during a conference call with analysts. But she noted that some aspects of HP's business -- notably printing and software, which saw a decline in sales -- "still face some challenges."

HP reported its earnings after the market's official close, when its shares fell 36 cents -- or about 1 percent -- to $35.12. Its stock price continued to slip in after-hours trading.

HP offers a wide variety of products, including computer networking switches, routers, servers, data storage devices, cybersecurity services and data-analysis software. But its biggest single source of money comes from sales of its personal computers, which grew 12 percent for the quarter compared with a year ago.

"Surprisingly, consumer PCs improved, which hadn't happened for many, many quarters," noted tech analyst Patrick Moorhead.

Nonetheless, Whitman noted that the overall market for PC's "continues to contract," due to dwindling interest from smartphone-obsessed consumers. That's been a big concern to HP's investors.

In addition, the corporation has been heavily criticized for some corporate acquisitions in recent years, including its $11 billion deal to buy British software company Autonomy in 2011. Shortly after completing that deal, HP wrote off $8.8 billion of Autonomy's value, saying it had been misled about Autonomy's worth. It also recently replaced two board members who'd been blamed for the purchase.
Autonomy's former CEO, Michael Lynch, has denied HP's claims that Autonomy's value had been artificially inflated. But those accusations are under investigation by U.S. and British authorities. In addition, the Autonomy purchase has sparked lawsuits by HP shareholders, alleging that HP didn't protect their interests when it pursued Autonomy.
Since becoming CEO in September 2011, Whitman has shuffled her executive ranks, attempted to focus the company on more profitable products and cut expenses, including laying off a good chunk of her workforce. After having nearly completed jettisoning 34,000 workers it began laying off in 2012, HP -- which has about 317,500 workers worldwide -- announced in May that it planned to cut an additional 11,000 to 16,000 employees.
The company's shares have been steadily rising for nearly a year. Nonetheless, Wall Street experts have a mixed view of its near-term prospects.
One positive is that many companies recently have been replacing their computers, in part because of Microsoft's decision to stop providing updates for its widely used Windows XP operating system. In addition, HP's market share in PCs and some computer-server niches "has also strengthened over the last three quarters," according to a recent report by Bernstein Research analysts. However, they added, "we worry that many of HP's end-markets are unlikely to see material revenue growth going forward."
In a separate note to their clients, Raymond James analysts concluded that "PC growth has peaked, as evidenced by commentary from several large resellers and distributors," and they called HP's ability to grow "questionable."

Sunday, August 10, 2014

bigger than memristors

Bob Burmeister, one of HP's long-term IC leaders, has been skeptical of the HP claims re the level of contribution from the memristor technology.  He has several times sent me notes that I have not included herein, but this one seems truly novel and significant.

This chip, at 46 million 'synapses' per second per watt, is much more like Carver Mead's strong work on neuromorphic computing from three decades ago.  Some of this early work, which resulted in the first significant cochlear implants and retinal simulators, is described in my new book, The Gentle Philosopher: Reminiscing--Carver Mead and me, available at:

Back to the IBM announcement, via the Kurzweil report--I mean, like four orders of magnitude less power per switching event, is BIG

Who said the breakthroughs are over with?