Naresh.Jois
Home of a professional developer & a aspiring photographer

Tilera debuts its sixty-four core processor

Silicon Valley startup Tilera today announced the Tile64, a processor with 64 programmable cores that, according to the company, houses ten times the performance and 30 times the power efficiency of Intel’s dual-core Xeon processors.

Intel may be getting tired of hearing about products performing better than its dual-core processors targeting server and embedded, as the company describes dual-core processors, at least when it comes to performance, as last year’s product. However, when there’s a company claiming that it can beat Intel’s last year’s product by a factor of 10x and 30x, depending on discipline, it’s certainly worth a look.

The Tile64 is a 90 nm RISC-based processor clocked between 600 MHz and 1 GHz aiming for integration in embedded applications such as routers, switches, appliances, video conferencing systems and set-top boxes. Its manufacturer claims that the CPU solves a critical problem in multi-core scaling and opens the door to hundreds or even thousands of cores using this new architecture.

Other than for example Intel’s Core architecture, which is expected to soon be running into a bottleneck caused by its centralized bus architecture (which acts as communication node between all cores), Tilera’s cores can exchange data with all other cores through a “mesh” architecture. Each of the 64 tiles consists of a CPU unit, a cache unit and a switch, which can send information into four directions to neighboring “tiles”. Each tile has a bandwidth of 500 Gb/s, with the Tile64′s aggregate bandwidth topping out at 32 Tb/s.

Besides the fact that Tilera claims that it has untangled a data traffic mess that otherwise would have surrounded a central bus, the company has come up with an interesting and flexible cache architecture for the tiles, each of which is able to act as a fully functioning system that can run an operating system. Each tile integrates two 8 KB L1 caches (8 KB iL1, 8 KB dL1) as well as a 64 KB L2 cache. There is no L3 cache per se, but if required by the application, a software developer can utilize all L1 and L2 caches as one 5 MB L3 cache.

Tile64 layout with 8×8 tiles and …

… the layout of one tile in detail.

The result is a claimed performance that is ten times what a dual-core Xeon offers, while performance per watt is exceeding the Xeon by 30x. The manufacturer states that each tile consumes a maximum of 300 mWatt, which translates into a maximum power consumption of 19.2 watts per Tile64 chip. Still, there is enough horsepower to encode eight parallel standard definition video streams at 2 Mb/s per stream, two high definition 720p streams at 7 Mb/s each or one 1080p video stream at 20 Mb/s.

According to Tilera, programmers can get their application up and running on Tile64 “very quickly”, while they mentioned that “fine-tuning” will optimize the software’s performance.

Tilera said that the processor is available now. For a new entry into the market, Tilera priced its product with confidence: 10K-tray pricing is set at $435 for each Tile64 – which appears cheap, if it can replace ten Xeon processors. But in a real world environment, the processor is priced against a quad-core Xeon 5345 (2.33 GHz, 8 MB L2 cache), which currently sells for a 1K tray price of $455.

Initial customers using the processor in upcoming products include 3Com, Top Layer, Codian and GoBackTV.

Of course, every time a new processor company comes around, there is the question if there is really enough room for another player – in this case, a market where heavyweights such as Intel and Texas Instruments battle for market shares. Other than PA Semi, a relatively new microprocessor company that does not build its processor, but licenses its technology, Tilera is actually manufacturing the Tile64, which is reminiscent of the rise and fall of Transmeta.

If you have been around in this industry for some time, then you may remember that Transmeta was in “stealth mode” from 1997-2000 and reason for media reports mainly because of one famous employee, Linux inventor Linus Torvalds. Transmeta launched with great fanfare in January 2000, but never got a foot on the ground and today is struggling to survive with revenues from licensing its LongRun2 technology to companies such as NEC.

Companies such as Tilera are exposed to the problems that broke Transmeta’s neck in CPU manufacturing: Microprocessor buyers today, for example, expect a track record of reliability that new companies cannot offer, buyers expect a support system, extensive manufacturing capabilities, a long-term roadmap that reveals what can be expected in terms of performance and feature set. Transmeta never made it much further than a presence in Asian markets, a few sub-notebooks on these shores as well as exotic computing solutions such as desktop clusters. Tilera, however, believes that it is competing in a different and “aggressively adopting” multi-core market in which it has the performance edge (which Transmeta never had).

During our briefing with Tilera, chief technology officer Anant Agarwal told us that he believes that the company has solved a “huge problem” of multi-core architectures and therefore has a substantial advantage: “We know how to get to hundreds of cores. This means that we are way ahead of the competition.”

Mouse brain simulated on computer

US researchers have simulated half a virtual mouse brain on a supercomputer. The scientists ran a “cortical simulator” that was as big and as complex as half of a mouse brain on the BlueGene L supercomputer. In other smaller simulations the researchers say they have seen characteristics of thought patterns observed in real mouse brains. Now the team is tuning the simulation to make it run faster and to make it more like a real mouse brain.

Life signs

Brain tissue presents a huge problem for simulation because of its complexity and the sheer number of potential interactions between the elements involved. The three researchers, James Frye, Rajagopal Ananthanarayanan, and Dharmendra S Modha, laid out how they went about it in a very short research note entitled “Towards Real-Time, Mouse-Scale Cortical Simulations”. Half a real mouse brain is thought to have about eight million neurons each one of which can have up to 8,000 synapses, or connections, with other nerve fibers. Modeling such a system, the trio wrote, puts “tremendous constraints on computation, communication and memory capacity of any computing platform”. The team, from the IBM Almaden Research Lab and the University of Nevada, ran the simulation on a BlueGene L supercomputer that had 4,096 processors, each one of which used 256MB of memory. Using this machine the researchers created half a virtual mouse brain that had 8,000,000 neurons that had up to 6,300 synapses. The vast complexity of the simulation meant that it was only run for 10 seconds at a speed ten times slower than real life – the equivalent of one second in a real mouse brain. On other smaller simulations the researchers said they had seen “biologically consistent dynamical properties” emerge as nerve impulses flowed through the virtual cortex. In these other tests the team saw the groups of neurons form spontaneously into groups. They also saw nerves in the simulated synapses firing in a ways similar to the staggered, co-ordinated patterns seen in nature. The researchers say that although the simulation shared some similarities with a mouse’s mental make-up in terms of nerves and connections it lacked the structures seen in real mice brains. Imposing such structures and getting the simulation to do useful work might be a much more difficult task than simply setting up the plumbing. For future tests the team aims to speed up the simulation, make it more neuro biologically faithful, add structures seen in real mouse brains and make the responses of neurons and synapses more detailed.

Internet2 operators set new internet speed record

We all know that data transmission records don’t hang around too long these days, but for operators of the Internet2 network, the final “official record” may have just been set. At the Internet2 consortium’s spring meeting, it was announced that officials “sent data at 7.67-gigabits per second using standard communications protocols,” but crushed even that milestone 24 hours later by achieving 9.08Gbps with IPv6 protocols. Notably, the data had to travel a whopping 20,000 or so miles roundtrip, and it’s being suggested that the newest record may be there awhile, considering the theoretical 10Gbps limit on Internet2 transmissions and the requirement for new records to sustain “a ten-percent improvement for recognition.” The best bit, however, is the mention of a “a new network with a capacity of 100Gbps,” which could see a full-blown DVD shot practically around the globe “within in a few seconds.”

A Single-Photon Server with Just One Atom

Every time you switch on a light bulb, 10 to the power of 15 (a million times a billion) visible photons, the elementary particles of light, are illuminating the room in every second. If that is too many for you, light a candle. If that is still too many, and say, you just want one and not more than one photon every time you press the button, you will have to work a little harder. A team of physicists in the group of Professor Gerhard Rempe at the Max Planck Institute of Quantum Optics in Garching near Munich, Germany, have now built a single-photon server based on a single trapped neutral atom. The high quality of the single photons and their ready availability are important for future quantum information processing experiments with single photons. In the relatively new field of quantum information processing the goal is to make use of quantum mechanics to compute certain tasks much more efficiently than with a classical computer. (Nature Physics online, March 11th, 2007)

A single atom, by its nature, can only emit one photon at a time. A single photon can be generated at will by applying a laser pulse to a trapped atom. By putting a single atom between two highly reflective mirrors, a so called cavity, all of these photons are sent in the same direction. Compared with other methods of single-photon generation the photons are of a very high quality, i.e. their energy varies very little, and the properties of the photons can be controlled. They can for instance be made indistinguishable, a property necessary for quantum computation. On the other hand, up to now, it was not possible to trap a neutral atom in a cavity and at the same time generate single photons for a sufficiently long time to make practical usage of the photons.

In 2005 the team around Prof. Rempe was able to increase the trapping times of single atoms in a cavity significantly by using three dimensional cavity cooling. In the present article they report on results where they have been able to combine this cavity cooling with the generation of single photons in a way that a single atom can generate up to 300,000 photons. In their current system the time the atom is available is much longer than the time needed to cool and trap the atom. Because the system can therefore run with a large duty cycle, distribution of the photons to a user has become possible: The system operates as a single-photon server.

The experiment uses a magneto-optical trap to prepare ultracold Rubidium atoms inside a vacuum chamber. These atoms are then trapped inside the cavity in the dipole potential of a focused laser beam. By applying a sequence of laser pulses from the side, a stream of single photons is emitted from the cavity. Between each emission of a single photon the atom is cooled, preventing it from leaving the trap. To show that not more than one photon was produced per pulse, the photon stream was directed onto a beam splitter, which directed 50% of the photons to a detector, and the other 50% to a second detector. A single photon will be detected either by detector 1 or by detector 2. If detections of both detectors coincide, more than one photon must have been present in the pulse. It is thus the absence of these coincidences that proves that one and not more than one photon is produced at the same time, which is demonstrated convincingly in the work presented.

With the progress reported now, quantum information processing with photons has come one step closer. With the single-photon server operating, Gerhard Rempe and his team are now ready to take on the next challenges such as deterministic atom-photon and atom-atom entanglement experiments.

Original work:

Markus Hijlkema, Bernhard Weber, Holger P. Specht, Simon C. Webster, Axel Kuhn, Gerhard Rempe
A Single-Photon Server with Just One Atom
Nature Physics online, March 11th, 2007

Danish scientists achieve advanced quantum teleportation

As you can imagine, I love it when science fiction becomes more science and less fiction. With that in mind, I am pleased to pass along the news that Danish scientists at Copenhagen University have made a breakthrough in the wacky world of quantum teleportation by transporting quantum information over a distance of half a meter (1.6 feet). In order to achieve this, Dr. Eugene Polzik and his team shined a strong laser beam into a cloud of room-temperature cesium atoms that shared the same directional spin. As Scientific American reports: “The laser became entangled with the collective spin of the cloud, meaning that the quantum states of laser and gas shared the same amplitude but had opposite phases. The goal was to transfer, or teleport, the quantum state of a second light beam onto the cloud.” (It should be noted that this process is more akin to duplication than actual teleportation, i.e. using this method on a human being would result in the formation of a doppelganger and not a magical Star Trek-like movement of matter). To achieve this goal, Polzik and other scientists added a second weaker laser pulse and split the two beams into separate branches in order to measure the difference between the quantum phases; through that measurement the scientists were then able to transfer the information of the spin state of the weak laser to the combination of the cesium atoms and the strong laser, without disturbing the quantum entanglement between the laser and the cesium. Umm, so the short of it is: one small step for a cesium atom, but one giant leap for quantum computing research and the advancement of teleportation theory.

Lockheed Martin eyes Quantum entanglement radar

The (in)famous defense contractor Lockheed Martin has apparently hit the loony sauce a bit too quickly on its latest patent application. In a proposed effort to concoct the ultimate omniscient radar, the firm is suggesting that it can break the boundaries of theoretical physics and create a “quantum entanglement” scanner that can “penetrate any type of defense to identify hidden weapons and roadside bombs from hundreds of miles away.” The theory — which hasn’t been realized in a product just yet — suggests that two particles can be joined so that whatever happens to one must also happen to its partner, however far apart they are, which could be used to detect contraband from faraway locales (or peek through suspicious garb). Interestingly, it doesn’t seem that we’re the only ones wondering just what type of Kool-Aid the outfit’s R&D department is sipping, as a physicist at Manchester University has reportedly insinuated that even in the far-reaching world of quantum physics, “the mechanics are just wrong.” Seriously, isn’t a Big Brother blimp enough for you guys?

Scientists Develop ‘Thinnest Material Ever,’ One Atom Thick

British-based scientists have created graphene, supposedly the thinnest material in the world: a one atom thick layer of carbon formed in hexagonal gropus. The substance, which is about 200,000 thinner than a human hair, was created two years ago but could be made only when stuck to another material. Researchers at the University of Manchester, England, have now managed to manufacture it as a film suspended between the nanoscale bars of scaffolding made from gold. Such a feat was held to be impossible by theorists, backed up by experimentation, because it is in effect a two-dimensional crystal that is supposed to be destroyed instantly by heat. It is so stable that it holds together in vacuums and at room temperature; it is thought to be able to exist because it undulates slightly. All other known materials oxidize, decompose and become unstable at sizes ten times the thickness. It was created by scientists at the University of Manchester, working with the Max Planck Institute in Germany. “This is a completely new type of technology — even nanotechnology is not the right word to describe these new membranes. We have made proof-of-concept devices and believe that the technology transfer to other areas should be straightforward. The real challenge is to make such membranes cheap and readily available for large-scale applications,” said Professor Andre Geim, of the University of Manchester.

Possible applications for the new advancement are expected in researching new drugs and to separate gases into their constituent parts. In medical research the membrane, which at single-atom thickness measures 0.35 nanometers, could be used as the support for molecules being analyzed by electron microscopes. The thinness of graphene membranes is such that the electrons would have much less irrelevant material to pass through and so be able to give a clearer picture of the structure of molecules, especially the proteins believed to hold the key to a generation of medicines. It can also be used to vastly increase computer speeds; it can potentially one day replace silicon because it can be used as a faster transistor that consumes less power. Leonid Ponomarenko, of the University of Manchester, is optimistic that it can be turned into a commercial success: “The technology has managed to progress steadily from millimetre-sized transistors to current microprocessors with individual elements down to ten nanometres in size. The next logical step is true nanometer-sized circuits.”

Building a Vista Premium PC for Rs.50k

Well the spanking new vista is here and I know most of us had refrained from upgrading waiting for Vista to arrive. As Vista is here I thought it might be a good idea for upgrading and get a new system, The rates mentioned here are steet prices, I think you have a hunt a little bit for products, but its certainly worth it. So here we go

Processor: AMD Athlon64 X2 4000+

I know all you Intel fanboys out there will say C2D is a better option, but you can never compare price to performance ratio to an AMD. AMD 4000+ is a cheap yet powerful option, Even in terms of gaming it gives a good performance.

RAM: Transcend 1GB DD2 800 MHz

I think this will be enough for most of our puposes.

Mother Board: Gigabyte GA-M55plus-S3G

The reason for choosing this Mother board is simple, This is a nForce 430 chipset board and has a onboard GeForce 6150 graphics controller, for those on a tight budget can skip buying a dedicated graphics card and still a good performance, Vista rates this 3.0 if 64MB is shared video memory, and it also has a PCI-Express slot for expansion for that cool graphics card. It has a 5.1 Channel HD Sound OnBoard. And is Vista Certified.

Hard Disk: Seagate 200 GB SATA HDD

I know this might not be enough for all the videos and music and stuff(you know what I mean), but It is still a good option, and you can record some shows too, but keep in mind if you want to do a lot of recording get a second HDD.

Graphics Card: XFX GeForce 7600 256MB

Should I say anything more??

Keyboard and Mouse: Logitech cordless Desktop

This is the perfect combination of usability and ergonomics, but if you want to use mouse a lot, I suggest you get a wired version cause you cant using those AA batteries which die out in a month.

TV Tuner : Haupauge WinTV PVR 150+

This TV tuner is certified for Vista, and if you really want get a second one by which you can record one show while watching other.

Monitor : Viewsonic VA1912W

This Monitor is simply great for viewing anything, it a 16:10 display ratio, with 1440×900 resolution with a response time of 5ms and a 500:1 Contrast Ratio. And is Vista Certified.

Optical Drive : Sony DVD Writer

Well this is the cheapest option, it doesn’t have Lightscribe but hey how many of us get Lightscribe compatible blank DVDs

Speakers: Creative Inspire 5.1

You cannot go wrong with these speakers, this the best possible speakers within 5k and it is the reason enough

CPU Casing : Really your choice

Nikon P5000

If you’re set on keeping it compact and fashionable, but still want to take some legitimate pictures, Nikon would like to tempt you with its new P5000 cam, which along with a 10 megapixel CCD, 3200 ISO and 3.5x zoom can accept add-on lenses or accessory shoe peripherals. I am pretty skeptical about that ISO 3200 mode, pictures at that ISO rarely look passable with a DSLR, much less a compact like this, and Nikon limits such pics to 5 megapixel shots, but we suppose having the option doesn’t hurt anything, and Nikon claims its new image processing engine with “enhanced noise reduction and an improved signal-to-noise ratio” should help. The camera includes all those Face-priority AF, In-Camera Red-Eye Fix and D-Lighting perks in the new S and L Series cameras, along with optical image stabilisation. The 2.5-inch LCD doesn’t sound stellar in an age of 3-inch viewfinders, but Nikon did leave room for an optical viewfinder, which is always nice. We’re also digging the i-TTL compatible accessory shoe, which supports external flashes, and the optional 24mm wide-angle lens and 378mm telephoto converter lenses. You can pick one of these up in March for $400 (Rs.20,000), but be sure to pick up a hefty SDHC card while you’re at it, those 10 megapixel pics are going to want some leg room.

Minox’s new 7 megapixel DC-7011

The inevitable bar-raising of low end tech now sees even crapcams rocking semi-advanced features spendier users have enjoyed for, well, nigh on years. Minox’s new DC-7011 has a 2.5-inch LCD, 3x optical zoom, and 7 megapixel sensor, USB 2.0, PictBridge support, even a top-mounted mode wheel — yet still somehow manages to look like like something you wouldn’t pay more than $75(Rs.3750) for. My advice to you: you probably shouldn’t.