Wednesday, November 19, 2008

Intel’ s Core i7 is FAST but getting there isn’t

I was invited to the Intel Core i7 victory lap party in way, way downtown south of the biotech area that’s exploding in San Francisco known as the dogpatch, is an area filled with old warehouses now used by film and art studios, and a few software development firms.

On my way to the Intel i7 rollout I was making a left to the designated parking area – parking being the most unavailable thing in this area of empty lots – go figure, and when I got to the right lane of on-coming traffic (the left two lanes were stopped waiting to make their left turn) a car hit me broadside in the passenger door (I was alone) and drove my heavy (4,220 pound) little car up the street 30 ft. He careened off. His car, a 1,095 pound Chevrolet Aveo was destroyed, his airbags popped and he got a ride to the hospital in a big red truck with flashing lights. He had, what looked to be minor injuries and a very bad day. Here’s the good part, or the bad part—time will tell—he’s a SF traffic safety officer, but it would seem someone should have explained speed limits to him.

That’s my car off to the right behind the fence - which ultimately stopped by forward motion. That’s a section of the explosive biotech development area in the background – Dogpatch is south to the left.

But now I had to wait for the twenty some cops who came rushing to scene and try to find the one who was in charge. Eventually he showed up on a motorcycle, and then the cops with cameras and tape measures went to work, and I kept trying to find someone to talk to. Early on there was a sergeant who seemed to know how things worked, but for some strange reason wasn’t in charge. Eventually I got my paperwork submitted and telling anyone who would listen I had to go to a meeting. They all told me forgetablutit.

I kept asking, and finally someone said, OK you can go. AND, I made it to the Intel event just a little late thanks to the sergeant at the scene who gave me a ride.

Got there, to Dogpatch Studios in time to hear that the Core i7 is the fastest thing ever – but I knew that because we’ve been testing one for a few weeks- and it IS fast!. And also go to hear that Intel has shipped over 100,000 processors to the mobo builders in Taiwan, and that unsurprising Alienware, Dell, Gateway and others will be building super-duper high-end hi-speed systems with the wonder chip.

When it comes to CPUs there are physical speed limits and Intel seems to be breaking them. We’re still checking the data but early indications are that using dual ATI Radeon HD4970x2s (that’s’s four GPUs chugging along with the four CPUs) the Core i7 based machine as compared to a Barcelona is over twice as fast!  We’re having trouble believing this and that’s why we’re still running the tests. The final results will be posted in Mt. Tiburon Testing Labs on our front page in a few days.

F A S T ! ! !

Saturday, May 24, 2008

How many FLOPS?

FLoating point Operations Per Second – FLOPS, one of the more obscure acronyms in our lives, and one of the oldest ones. It’s since been modified with a prefix of M (mega), G (giga) and most recently T (tera). A Terra is a million millions, one trillion (1012) a whole lot of anything, whether its cycles (Hertz), Bytes, dollars, or FLOPS. (And note - the ‘S’ in FLOPS is capitalized.)

So I was asked recently, how many TFLOPS in all the game consoles?

There are two answers to that question.

Do you mean in all the ones built, or just the FLOPS of the specific consoles?

So what is the TFLOPS rating of the game consoles? Well there are two answers to that question too.

If you count central processor FLOPS that’s one answer, if you count the FLOPS potential of the GPU and add it to the CPU’s FLOPS that’s another answer. And that answer provokes all kinds of debate. One side (the one I favor) is that the FLOPS of the GPU aren’t used in computations and therefore shouldn’t be counted because they can’t be measured and just represent a theoretical number. The other side argues that they are indeed being used in computation - the computation of shader operations. However, both sides agree that there isn’t yet a benchmark that can measure them. And therefore I conclude that we shouldn’t use them in evaluating the CPU FLOPS of game consoles.

The other main answer to the original question is do you mean in all the ones built, so I added in the installed base of all the consoles shipped.

The following table lists the FLOPS in consoles.

2007, October: Sony PS3 console, at US$400, that runs at a claimed 2 teraFLOPS; these figures represent the processing power of the GPU. The seven CPUs run collectively at a lower 218 GFLOPS.[14]

All of the GLOPS of all the consoles to date are only 402; where as a modern GPU is over two times that – Moore’s law in action.

You can read this several ways:

1. All of the CPU consoles added up don’t equal 1 TFLOPS.
2. All of the consoles added up don’t exceed 2 TFLOPS even if you count GPU and CPU.
3. All consoles shipped to date add up to 7.318 EFLOPS - Exa FLOPS - 1018 and that’s a number that that thrills the folks at .

Now why did I go through this laborious, pedantic, and sure to be arguable discussion? For several reasons:

1. It’s raining here on Mt Tiburon so I can’t go out and play.
2. I thought it was really interesting to look at how far consoles have progressed
3. I thought it was even more interesting to look at how far PC graphics have progressed – ATI and Nvidia are shipping TFLOPS AIBs, and the next gen coming out in June, with Dual GPUs will be approaching 5 TFLOPS per AIB.

Just think of what the game developers and the movie studios can do with kind of horsepower.

Think of the shader operations that will be offered soon. Those of you reading this who know me know Peddie’s first law – in computer graphics too much is not enough. And although I haven’t made it a law yet, if I did (it would be number three), I’m fond of saying – don’t just watch a movie – be in the movie.

That’s where we’re heading with all these tera and peta FLOPS and I can’t wait to get there because next stop after this one is the holodeck.

Monday, April 07, 2008

Educating the next crop of engineers

As I pondered this ponderous title and the challenge it represents for me to lead the round table at COFES it got me thinking about how we learn.

Studies have shown that children learn fast and do so until they become 19 to 20 years old, then their brains become less flexible and learning takes longer, and it’s more difficult. By the time one reaches full adulthood and middle age  you really have to work at it to learn new things; languages are particularly difficult because of the contextual and grammatical differences, they don’t easily fit our well honed models and views of the world.

Engineering is a language, as is medicine, music, and art.  And not everyone is able to learn those subjects as easily as others,  someone good in art may do poorly in medicine,  although there does seem to be a natural linkage between engineering and music.

As we grow we learn not just facts but also falsehoods,  philosophies, and fears. And those fears, philosophies, and falsehoods prevent us from learning other, usually important, and often enriching things – we are cast into a channel of self imposed ignorance.

It’s these falsehoods that are the greatest barrier to children entering the field of engineering. These localized common wisdoms found in school yards, barber shops, and local hangouts inhibit and stifle a child’s potential. They are taught that they can’t do certain things, or that such things are too hard, or not cool. The teachers are presenting their own failures as fact, and sadly embedding the crippling idea, don’t try you’ll only fail, into supple young minds.

Some however are so drawn to a vocation or profession that no amount of obstacle, peer pressure, or parental abandonment can dissuade them. They are the lucky ones, the ones who have found a passion early in life and managed to pursue it, and probably do well at it. And whereas we can’t ignore these people, they don’t need as much of our attention as the left behinds and passed over. The challenge is plucking the marginalized children out before they get too stuck in self inhibiting ways and can’t be inspired.

And so with regard to engineering, I think there could be simple tests, tests that are not formed like a test (and certainly not an element of the disastrous No Child Left Behind fiasco.) Rather testing would be coupled with sensitivity training for teachers, counselors, coaches, and other adults involved in the management, guidance, and education of children. The goal would be to identify those students who have a natural knack for problem solving, mechanics, and systems. And even though the theme of this rambling diatribe is about how to find and encourage the next crop of engineers, it is not constrained to just that narrow field of endeavor. The abilities we’re looking for apply equally to an orthopedic surgeon, and maybe even a composer.

In the past, in most countries including the United States and especially those countries with centrally controlled economies, aptitude testing was the norm. Children were evaluated at various grades, often as a way of directing them into studies where they would be best suited to make the greatest achievement (and subsequent contribution to society.) That concept has been abandoned in the US as being narrow minded and limiting. The idea being that testing and directing children towards particular areas of study limits a child’s self expression. Admittedly, using testing to sort children into rigid career paths or trade schools can be dangerous, cruelly limiting a child’s future. But to completely abandon the idea of testing in order to nurture and foster children’s aptitudes is also a mistake. I think it’s especially a mistake in underprivileged schools where children may not have a strong home life and learn about the world on the streets from people no better educated than they are. These are precisely the kids that need focused attention early in their careers.

One of the excuses for not providing such evaluations and guidance for children is the limitations in staff and of teachers. The teachers are over burdened with large classes, and administrative tasks such as checking home work, grading papers, submitting lesson plans, that eat into their time to offer any personal guidance.

So I have a proposal.

We hear the lament of industry that the US does not have sufficient technical people and therefore we should open up our immigration policy to allow more foreign workers in (H-1B visas.) However, if the industry would apply some of its resources to augmenting the schools with special information and teaching programs, evaluations of students, and guidance for students that show promise, I think US industry could find all the technical people it needs. The problem for US industry is that such a program would take at least 15 years from first contact in grammar school and US industry wants an instant solution.  Therefore, I propose the government grant the opening up of H-1B visas but with the provision that US industry pay a special tax to a fund for the evaluation, training, and most importantly, guidance of US students (regardless of where they were born.)

I further suggest that our professional societies assess their members an additional fee (some do this now as an educational fund) for student evaluation programs that are combined with student guidance programs.

I believe there are hundreds of thousands of potential engineers in grammar, middle and high school right now. In our ever more complex and challenging world—and especially so in the US—we need more engineers and technicians than ever before. We have these bright young flowers,  let’s find creative and imaginative ways of encouraging them before they get lost in the weeds.

For further reading
  http://www.nytimes.com/2005/01/16/education/edlife/EDSCIENCE.html
  http://members.shaw.ca/priscillatheroux/Glasser.htm
  http://www8.nationalacademies.org/onpinews/newsitem.aspx?RecordID=s12132002
  http://www.aaas.org/news/releases/2005/0518visa.pdf

Saturday, February 02, 2008

PMA in Las Vegas: new cameras cover old ground

The groundhog poked his nose out of his house and apparently he did not like what he saw — or maybe he did. I get confused. Whatever happened with the U.S. weather rodent, whether he went back in his hole or came out and danced, his priests declared six more weeks of winter. This came as little surprise to people in the U.S. who have been enduring generally awful weather across the country.

In Las Vegas the weather was bright and sunny if a tad on the cold side causing the Las Vegans to wear mittens and the rest of us to laugh at them. I was in Las Vegas for the PMA show &ndash the photography industry conference. This year the show took place at the same time as the Shot Show a gun show. Get it?

The gun lovers were in the North Halls and the camera lovers were in the South Halls. I would go into the differences between the two different groups but that way leads to trouble. I will put it this way, in most cases it was pretty clear what doors people would be using and with whom it might be wise to avoid spirited discussion. Cab drivers made a point to ascertain excactly what kinds of shots you were likely to take, snapshots or pot shots, before they let loose with any opinions.

I’m in the process of writing up the PMA show for TechWatch but I can tell you this right now.Hold off on buying a camera. There is some really good stuff coming down the pike, or will be as soon as the camera manufacturers can get their heads around the idea that consumers expect really new features and capabilities — not incremental upgrades before they’ll abandon their cameras that are doing the job for new cameras that will do the job better. Some of the trends you’ll be seeing include:

  * 14.2 MP cameras in all ranges because that’s the point at which digital meets the capabilities of 35mm film
  * Low end digital SLR cameras with through the LCD view finding. (Yes, that is kind of a contradiction in terms but camera users have come to like composing through the LCD)
  * Tiltable LCDs for DSLRs
  * HD video for DSCs

So, why wait? If you bought a camera three or four years ago, there might not be enough of an advantage for you move up. Do you really need a few more megapixels? And, yes, through the LCD view finding is cool, but haven’t you already adjusted back to through the lens? Personally, I’m holding out for a new camera with a full frame sensor that doesn’t require me to take out a loan to buy it or a bionic arm to hold it up for any length of time.

In the DSC range, I could go for a DSC with really good video. I don’t have to have HD — it takes up too much storage space — but I would like to have one small camera for snapshots and video. I keep getting lectures from well-intentioned manufacturers who tell me good video and good still photography can not co-exist in one sensor.  With the videos I shoot, however, I’m not sure quality is so important. It’s just important that I have the camera with me. For the most part products at PMA seemed to represent upgrades to the products that came out last year. Give it six months and see what the companies have to offer. Frankly, I’m going to wait until next year ... maybe.

Wednesday, January 23, 2008

Ink — I love ink

Electronic ink or electronic paper (e-paper)  was developed in the 1970s by Nick Sheridon at Xerox’s Palo Alto Research Center and it was called Gyricon. Its technical name is electrophoretic display meaning a display that forms visible images by rearranging charged pigment particles (i.e., powder like substances) using an applied electric field.

In the early 2000s the technology began to gain some traction and looked like it could be mass produced, albeit in small sizes. One of its first commercial applications was for the second display on mobile phones, and in 2001 E Ink and Philips Components announced plans to jointly develop high-resolution electronic ink displays for handheld devices such as PDAs and electronic books.

And then it all went quiet.

But at last year’s 3GSM conference (fro mobile phones, in Barcelona) I found the first early production rollup displays from the Philips spin-off, Polymer Vision. Needless to say, this long-anticipated and -wanted technology got lots attention from both the fans and the curious. The product itself is called Readius and is based on the organic/polymer TFT technology Philips invented.

But prior to that Sony released an e-book reader in 2004, when it introduced its first e-paper device, the Librié, in Japan. It didn’t hit the US or Europe until 2006, and it was re-christened Ebook. It’s been the darling of commuters and technophiles ever since. That is until December 2007.

Kindle
  In December, in one of the best executed PR programs since the iPhone, Amazon announced and then released the Kindle. It was the Ebook only better. A slightly larger display, and always on-line for almost instant down loading of books, and web surfing.

Kindle had all the things Sony’s Ebook didn’t; granted, Amazon had two plus years to figure it out. Nonetheless it was sold out the day it went on sale (300k units we were told) and was in such demand (mind you this was just before the holidays) that they were being bid up to $1,500 on eBay (The retail price from Amazon was $399.)

I’m reviewing the product for our Mount Tiburon Testing Labs and that will appear in a week or so on our front page.

But it’s the ink I want to talk about. I won’t go into a lengthy technobabble discussion about the polarization, molecules, powders or polymers, rather I want to speak about how damn great it looks.

The display uses reflected light – just like a book. That’s right, lights out – no more reading. But the display is so sharp and has such a high contrast ratio similar to that of a newspaper, and a very large viewing angle unlike many LCDs, that when people first encounter it they don’t believe it’s not back lit. It also has another interesting characteristic, albeit its use is not clear. The display leaves a ghost-like reanimate of the book on the screen when turned off. I kind of like that, it makes me feel like it’s ready to leap into action and bring the latest thriller to me.

And when its on, it looks like newspaper.
  The kindle is a white package with a liquid paper display
  There’s no doubt in anyone’s mind that this is the wave of the future. By 2014 e-paper made with electrophoretic materials reach almost $2.0 billion in sales, with liquid crystal based e-paper at $1.5 billion and electrochromic e-paper at just under a $1.0 billion.

The next challenge for the liquid ink folks is to produce color.  They’ve got red and some green working, but blue is a challenge still. And the color, as of now seems to have slowed the response time of the display down a little. I haven’t heard about the contrast ratios. There have also been some interesting experiments using two color powders and RGBW color filters combined with the powder.

For electronic books I don’t see the immediate need, although one can envision color illustrations and from there how about my entire Batman comic book library on my Kindle?