Posts Tagged ‘CPU’
Two Key Advances Bring Quantum Computers Closer to Reality Than Ever
In the first study, researchers at the University of California-Santa Barbara say they’ve built the first working quantum computer chip based on the von Neumann system. Named for the engineer who designed the concept, the von Neumann architecture combines processors and memory, and it’s the basis for every computer out there. (With one .)
This quantum CPU (quCPU?) is a big breakthrough, because quantum computers by definition are difficult to design. They’re based on the concept of superposition — that a quantum bit, or qubit, can exist in two different states at once. Put another way, it can be a 0 or a 1 at the same time, and it can therefore perform calculations more quickly than a system based on 0 or 1. But it’s hard to keep the qubits in a state in which this is possible, and interfering with them — i.e., reading their data — can destroy their superposition capabilities. So, a system that integrates random access memory into the qubits is a big step toward a working computer.
Researchers at UCSB super-chilled their quCPU to near absolute zero and performed a few calculations. Quantum information traveled back and forth among storage and processing elements, and the system performed pretty well — not perfectly, but it’s a start. They also found that the quantum memory can retain information for much longer periods than the qubits, which is also a good sign.
Next, the team is trying to increase the number of quantum devices integrated on a single chip, and they’re studying different metallic materials to make this easier, according to .
In another quantum paper, researchers in Austria report building the first working quantum simulator — kind of like a quantum computer, but different in scope. It can be used to model the behavior of quantum systems, which can potentially help improve quantum computers.
It would be useful for many reasons to model the behavior of quantum systems, but this is impossible with a traditional computer, as Richard Feynman figured out in 1982. It would take exponential time, with the system working more and more slowly as the calculations increased in number. For a general description of a quantum spin system with 300 particles, a computer would need more memory than exists in the world — even if all of the observable matter in the universe was processed into memory, as the Austrian researchers . But a quantum simulator, which can complete so many more calculations, would not experience this slowdown. To make one of these, you would have to very carefully control the setup of the simulator, and this is what the Austrians have done.
The team used six laser-cooled calcium atoms as qubits, and used laser pulses to initiate calculations. They found the system could simulate several types of interacting spin systems, according to Science magazine, which published both papers today. The simulator can be reprogrammed to simulate any type of quantum system, the researchers say.
Given breakthroughs like these, quantum computers may be closer than ever.
[, ]
Two Key Advances Bring Quantum Computers Closer to Reality Than Ever
In the first study, researchers at the University of California-Santa Barbara say they’ve built the first working quantum computer chip based on the von Neumann system. Named for the engineer who designed the concept, the von Neumann architecture combines processors and memory, and it’s the basis for every computer out there. (With one .)
This quantum CPU (quCPU?) is a big breakthrough, because quantum computers by definition are difficult to design. They’re based on the concept of superposition — that a quantum bit, or qubit, can exist in two different states at once. Put another way, it can be a 0 or a 1 at the same time, and it can therefore perform calculations more quickly than a system based on 0 or 1. But it’s hard to keep the qubits in a state in which this is possible, and interfering with them — i.e., reading their data — can destroy their superposition capabilities. So, a system that integrates random access memory into the qubits is a big step toward a working computer.
Researchers at UCSB super-chilled their quCPU to near absolute zero and performed a few calculations. Quantum information traveled back and forth among storage and processing elements, and the system performed pretty well — not perfectly, but it’s a start. They also found that the quantum memory can retain information for much longer periods than the qubits, which is also a good sign.
Next, the team is trying to increase the number of quantum devices integrated on a single chip, and they’re studying different metallic materials to make this easier, according to .
In another quantum paper, researchers in Austria report building the first working quantum simulator — kind of like a quantum computer, but different in scope. It can be used to model the behavior of quantum systems, which can potentially help improve quantum computers.
It would be useful for many reasons to model the behavior of quantum systems, but this is impossible with a traditional computer, as Richard Feynman figured out in 1982. It would take exponential time, with the system working more and more slowly as the calculations increased in number. For a general description of a quantum spin system with 300 particles, a computer would need more memory than exists in the world — even if all of the observable matter in the universe was processed into memory, as the Austrian researchers . But a quantum simulator, which can complete so many more calculations, would not experience this slowdown. To make one of these, you would have to very carefully control the setup of the simulator, and this is what the Austrians have done.
The team used six laser-cooled calcium atoms as qubits, and used laser pulses to initiate calculations. They found the system could simulate several types of interacting spin systems, according to Science magazine, which published both papers today. The simulator can be reprogrammed to simulate any type of quantum system, the researchers say.
Given breakthroughs like these, quantum computers may be closer than ever.
[, ]
Native Chinese Supercomputer Could Be World’s Most Efficient
The race for performance-per-watt is on
Officials at China’s Institute of Computing Technology recently announced a summer launch for the country’s first wholly indigenous supercomputer, the , which will achieve 300 teraflops using 3,000 1 GHz 8-core Godson (also known as Loongson) 3B chips. It’s very efficient at 3.2 gigaflops per watt — or 128 gigaflops using the power of a typical light bulb.
This is possible because of the Godson chip’s relatively low clock speed, at 1.0 GHz, and its use of the old-school 64-bit MIPS architecture, according to .
As computers grow ever more powerful, they also grow more power-hungry, requiring rooms full of cooling systems. So power-sipping supercomputers would be a major advancement. predicts performance per watt will become the dominant supercomputer awesomeness metric.
IBM’s new , slated to be delivered next year, will be one of the most energy-efficient in the world, thanks to improved chip designs and an efficient water-cooling system. Mira will be built on a Blue Gene/Q platform, currently tops in terms of performance per watt.
But HPC Wire says the Godson platform could give Blue Gene/Q a run for its money. The MIPS architecture is known for high function at low power, which is why MIPS cores can be found in technology ranging from portable video game consoles to wireless routers.
The People’s Republic is performing pretty well in the supercomputing race. Its supercomputer is the world’s fastest, achieving 2.507 petaflops via 7,168 NVIDIA GPUs and 14,336 CPU. But those components are made by American companies.
Ultimately, the Godson chips are unlikely to challenge companies like AMD and Intel in western computers. But HPC Wire notes they could cut into those companies’ imports into China.
[, ]
Could Intel’s new graphics/processor chip kill off Nvidia? (poll)
hopes to shake up the computer chip market as it unveils details on Monday about how it is putting a microprocessor and graphics processor onto a single silicon chip.
Code-named Sandy Bridge, the chip will figure prominently in the keynote speech of Intel chief executive Paul Otellini, who kicks off the event in San Francisco on Monday.
The processor-graphics combo chip is available in prototypes now and will ship to customers in early 2011. It will target low-end desktop computers and laptops, and could cause a big headache for , the maker of stand-alone graphics chips.
The Sandy Bridge chip is the latest attempt to put an end to the ongoing debate about whether graphics processing should be done with a much more powerful 3D graphics processing unit, or GPU, or if it can be handled better as part of a computer’s central processing unit, CPU, also known as the microprocessor.
Intel launched a combo chip last year that glued together two chips in a single package. But Nvidia laughed that chip off as brain-dead when it came to running the top 30 computer games. But this second-generation effort, Sandy Bridge, combines the two chips into one and poses a greater threat to Nvidia. , said he thinks entry-level GPUs will be unnecessary in the wake of Sandy Bridge’s introduction. (See Anandtech chart at right, which compares the performance of existing Intel-compatible integrated graphics chips, where a graphics chip is combined with a chip set).
“Is it enough to kill all discrete graphics? No,” Anandtech wrote. “But it’s good enough to really threaten the entry-level discrete market.”
One of the reasons Intel can do this is Moore’s Law, the notion formulated by former Intel chairman Gordon Moore that says chip performance doubles every couple of years. As it makes circuitry smaller, it can fit more electronics onto a chip, improving performance. By moving to a new generation of manufacturing technology, Intel can combine a capable microprocessor (with four cores, or processing brains) with a decent graphics component.
Intel’s Otellini said during the company’s recent earnings call that Intel was accelerating its investment in new 32-nanometer manufacturing technology because of strong demand from PC makers. Meanwhile, Advanced Micro Devices is preparing to launch its own hybrid chips, under the Fusion brand name, early next year.
Nvidia, meanwhile, continues to be a naysayer on the capabilities of hybrid chips and is skeptical of Intel’s (as yet unreleased) performance claims.
“They are launching a faster turbo-prop in a world that has moved on to jets,” said Ken Brown, an Nvidia spokesman.
You would expect that, since Nvidia doesn’t have its own microprocessor. But the company points out that Intel’s Sandy Bridge won’t be able to run DirectX 11, which is the graphics standard from Microsoft that governs whether developers can take advantage of certain cool graphics features. Nick Knupffer, spokesman for Intel, says that Intel is working on boosting its investment in driver technologies so that it will be able to run more games with better graphics.
“Integrated graphics and standalone graphics each have their place in the market in terms of performance and power consumption,” said Dean McCarron, analyst at Mercury Research. “Both graphics segments have coexisted for more than a decade, during which the capabilities of both segments have improved tremendously. It is highly unlikely either segment will displace the other, and increasingly we have seen active coexistence, particularly in the mobile market where computer makers are configuring notebooks to make use of both types of GPUs in the same platform.”
Sandy Bridge does have some features that aficionados will appreciate, such as the ability to display stereoscopic 3D Blu-ray videos and high-definition TV. It can support the Intel Wireless Display, which can display the image on your laptop on the big screen TV in your home. And it will have high-performance transcoding, or the ability to translate video from one format to another so that it can run on just about any device.
Intel recognizes that it isn’t going to kill off the market for $600 graphics cards that hardcore gamers want. But the market trends favor its direction, Knupffer said. By 2014, about 80 percent of the PCs shipped are expected to have hybrid chips, which Intel calls Processor Graphics, according to market researcher .
What do you think? Please take our poll.
Companies: , ,
People: , ,
Samsung’s new Orion processor brings dual-core power to phones
Samsung already has a plenty powerful mobile CPU with its Hummingbird processor — the brains behind its — but today it has even bigger plans for next year: A low-powered dual-core processor, dubbed Orion, with powerful 3D and video capabilities.
The processor combines two ARM Cortex A9 cores running at 1 gigahertz to create a computing powerhouse. Samsung says it provides 5 times the 3D performance of the Hummingbird processor, enough power to record and decode 1080 high-definition video, as well as the ability to power three displays. A manufacturer could use the chip to create a device with two screens, and it would still have enough power left over to send video to a computer monitor or HDTV via its built-in HDMI connection.
Orion is built with Samsung’s 45-nanometer low-power processor technology, meaning that it won’t drain battery life even though it’s packing a healthy amount of horsepower.
Samsung isn’t the only mobile chip-maker with dual-core plans. Intel is looking to finally get a piece of the smartphone pie with , and Qualcomm is also .
The move towards dual-core processors will make mobile devices significantly more powerful than they can be if restricted to a single CPU core. The challenge for chip manufacturers will be balancing the more powerful hardware with battery life concerns.
Samsung says its Orion processor will be available to a few customers later this year, and it will started building the chips en masse in the first half of 2011.
Companies: ,