A team of boffins has emerged from its smoke filled labs with a microchip with 1,000 independent programmable processors.
The team, from the University of California, Davis, Department of Electrical and Computer Engineering, developed the energy-efficient 621 million transistor “KiloCore” chip so that it could manage 1.78 trillion instructions per second.
Team leader Bevan Baas, professor of electrical and computer engineering said that it could be the world’s first 1,000-processor chip and it is the highest clock-rate processor ever designed in a university.
While other multiple-processor chips have been created, none exceed about 300 processors. Most of those were created for research purposes and few are sold commercially. IBM, using its 32 nm CMOS technology, fabricated the KiloCore chip.
Because each processor is independently clocked, it can shut itself down to further save energy when not needed, said graduate student Brent Bohnenstiehl, who developed the principal architecture. Cores operate at an average maximum clock frequency of 1.78 GHz, and they transfer data directly to each other rather than using a pooled memory area that can become a bottleneck for data.
The 1,000 processors can execute 115 billion instructions per second while dissipating only 0.7 Watts which mean it can be powered by a single AA battery. The KiloCore chip executes instructions more than 100 times more efficiently than a modern laptop processor.
The processor is already adapted for wireless coding/decoding, video processing, encryption, and others involving large amounts of parallel data such as scientific data applications and datacentre work.
Troubled chipmaker AMD has been sued over the number of cores in its Bulldozer chip.
A class action lawsuit, led by a bloke called Tony Dickey, claims AMD tricked punters into buying its Bulldozer processors by overstating the number of cores contained. Bulldozer was advertised as having eight cores when functionally it actually only had four.
Part of the problem is that AMD’s multi-core Bulldozer chips combine the functions of what would normally be two discrete cores into a single package, which the company calls a module. Each module is identified as two separate cores in Windows, but the cores share a single floating point unit and instruction and execution resources. This is different from Intel’s cores, which feature independent FPUs.
But the suit said that Bulldozer cores cannot work independently, and as a result, cannot perform eight instructions simultaneously and independently. The plaintiffs claim that this results in performance degradation, and average consumers in the market for a CPU lack the technical expertise to understand the design of AMD’s processors and trust the company to give accurate specifications regarding its CPUs.
The suit argues that tens of thousands of consumers were misled into buying a Bulldozer CPU that cannot perform in the same way as a true 8-core CPU. If this is true AMD violated the Consumer Legal Remedies Act, California’s Unfair Competition Law, and was guilty of false advertising, fraud, breach of express warranty, negligent misrepresentation, and unjust enrichment.
AMD is being sued for damages, including statutory and punitive damages, litigation expenses, pre- and post-judgment interest, as well as other injunctive and declaratory relief as is deemed reasonable.
Chipzilla has released most of its 6th generation Core processors which should be finding their way into PCs by next month.
Intel spilled the beans on the Skylake-K variants aimed at gamers, last month. Now, the firm has added to the list Skylake-U aimed at thin-and light laptops; Skylake-H for high performance systems; and Skylake-S which brings “desktop performance” to value-level systems, all-in-ones and mini PCs.
Last but not least there is the Skylake-Y, for the Core m7, m5 and m3 mobile chips for tablets, 2-in-1 systems and Intel’s Compute Stick.
For those who came in late, Skylake uses the same 14nm process as the previous Broadwell generation, but introduces a new and improved microarchitecture.
It is supposed to have boosts in performance and battery life along with support for new capabilities, such as Windows Hello and Cortana in Windows 10.
Intel said that Skylake will offer up to 2.5 times the performance and up to three times the battery life of older generations of PC hardware. These comparisons are typically based on three-year-old systems.
Also in the mix is DDR4 memory, along with enhanced Intel HD graphics offering up to a 30 times improvement in performance, support for high-speed PCI Express storage via Intel’s Rapid Storage Technology, and capabilities to support new features in Windows 10.
Among these is an integrated Image Signal Processor that accelerates processing of data from cameras, and helps to enable the Windows Hello password-free sign-on in Windows 10.
It is starting to look like things are hotting up in the data centre with a war brewing between ARM vendor Cavium and Intel.
Cavium has come up with a 48-core chip which it hopes to blow Intel’s efforts out of the water. Dubbed ThunderX, it will be built on the 28 nm process, with between 24 and 48 custom 64 bit ARMv8 cores at 2.5 GHz, and will run at between 20 and 95 W. Intel’s Xeon processors run 100 W just for the processor.
ThunderX will be sampling in Q4 2014 and there is little out there on performance. Cavium claims its SoCs should match the power dissipation of Intel-based systems but then it would say that.
Other things leaked about ThunderX range include statemetns like “hundreds of gigabits/second” and ” I/O, cache coherency across dual sockets, four DDR 3/4 72 bit memory controllers that will cope with up to 1 TB in a dual socket configuration, integrated hardware accelerators for applications like storage, networking, virtualisation and security”.
There will be five flavours of ThunderX. There will be the ThunderX_CP for content serving applications; the ThunderX_ST for Hadoop; the Thunder_SC for Web front-end, security and cloud; and the Thunder_NT for embedded applications and NFV workloads. The ThunderX CN87xx 8-16 core range, with two DDR3/4 controllers, multiple 10GbE, SATAv3 and PCIe Gen3 interfaces will look after the lower end.
Hopes that ARM might be able to push itself into the gaming market have been dashed as the Ouya console failed to live up to its expectations.
Benchmark tests are coming out which reveal that Ouya is not at all up to scratch and there are over 70 other mobile devices that have performed better.
Ouya is shipping only to developers who had signed up and customers who had pre-ordered the gaming console. It is not expected to be flogged to the great unwashed until the summer.
However, the benchmarks suggest that even for the comparatively cheap sub $100 price, it is going to be a turkey.
The android based gaming console packs a Nvidia Tegra 3 System-on-Chip (SoC) and the quad-core ARM Cortex-A9 processor is clocked at 1.7GHz. The graphics leans on the Nvidia A GeForce ULP.
But the Tegra 3 is a little slower than Nvidia’s latest efforts, the Tegra 4 and is pretty much a chocolate teapot when it comes to games.
Futuremark’s benchmarking on mobile devices has ranked Ouya at 73rd. This means that smartphones like the Samsung Galaxy S II plus and Galaxy Grand performed better. In fact Ouya would have been better off powered by Qualcomm processors with Adreno 320 graphics.
This is all a bit of a pity. Ouya had a few good things going for it. It was the the first totally open video game console and could have allowed a lot of tinkering from its users. Tinkering is widely seen as the way to get the next generation of developers roped into the IT industry and with a $99 price tag it was the sort of gizmo kids could buy.
It would have also moved the console market away from the x86 spec and allowed a greater cross-fertilisation between the ARM based mobile market and other hardware.
Anyone who thought that Intel might take years to catch up to ARM on the mobile front might want to think again.
In an early benchmark test an Intel based Lenovo IdeaPhone K900 gave the Samsung Galaxy S4 a good kicking.
The IdeaPhone has a dual-core Intel Atom Clover Trail+ processor. Unfortunately the benchmark did not indicate if the S4 used Samsung’s Exynos 5 Octa chipset or a variant bearing Qualcomm’s Snapdragon 600 quad-core CPU.
GSM Arena, which carried out the test, said that the Lenovo K900 reveals that the smartphone is a touch ahead of Samsung Galaxy S4 and the K900 is one of the fastest smartphone in the current market.
The Galaxy S4 was the fastest Android phone yet, so Intel knocking it off the top spot would be significant.
Intel as it is using a dual-core CPU design to combat ARM’s quad-core designs.
The dual-core Intel Atom CPU inside the IdeaPhone K900 is clocked at 2 GHz.
Chipzilla needs to stop mucking about and get itself an ARM licence so that it can push into the mobile market, claims, er, a co-founder of ARM.
Robin Saxby, a co-founder of ARM, at the GSA Entrepreneurship Conference at the British Museum, channelled the late Kenneth Williams and told Intel to stop messing about.
Years ago, Intel offered StrongARM and Xscale microprocessors based on ARMv4 and ARMv5 instruction-sets, but flogged off that division to Marvell Semiconductor in mid-2006.
At the time Chipzilla thought that ARM architecture was not scalable enough in terms of performance. While it turned out ARM was scalable, its power consumption increased along with performance. Intel did not think that the ARMv8 64-bit architecture will only be insignificantly more power-efficient than comparable x86 offerings based on AMD Jaguar or Intel Silvermont.
Saxby said that AMD recently came to the ARM party and decided to develop server-class system-on-chips based on ARMv8, just like loads of other companies, including Applied Micro, Calxeda, Nvidia, Samsung, Qualcomm and many others. But Intel just does not want to be friends with ARM, he moaned.
Talking to Electronics Weekly, he said that ARM had to turn its enemies into friends, the only one it did not manage to turn was Intel.
“I recommend Intel to take an ARM licence and stop messing about,” said Saxby. With a nice ARM licence, Intel could start building mobile chips around ARM cores, he added. A few years ago, Tudor Brown, a senior ARM executive, told us that Intel has several licences, something that Intel confirmed.
Of course this implies that Intel’s own mobile chips are going nowhere.
Intel has been given the thumbs up to radically expand its Kildare headquarters in a move which could see the creation of thousands of Irish jobs.
The Irish national planning authority, An Bord Pleanala, gave the green light today for a massive 40-hectare expansion at the plant in Leixlip. Kildare County Council had approved the plans last August but they were appealed to the national planning authority.
According to the Independent, Intel is supposed to be working on its next generation of computer chips in Leixlip, however, the Irish press has noted that there is a marked reluctance from Chipzilla to press ahead. Intel has yet to decide when or if it will go ahead because the Kildare site is competing with other Intel locations including Arizona and Israel for the work.
That has not stopped the Irish government, which has been bending over backwards to attract Intel with all sorts of tax breaks and other sweeteners, declaring a victory. Government chief whip Emmet Stagg claims that at least 4,300 jobs could be created.
Most of these are 3,500 construction jobs will be created as part of the expansion, but there will be 800 permanent manufacturing jobs when it’s completed. Over 4,500 workers are already based at the plant.
Stagg said that the expansion is “proof” that companies like Intel have confidence in Ireland.
While it is nice that someone is optimistic about Intel’s plans, Intel does not appear to be in any hurry. It has ten years to exercise its planning permission permit. If it does go ahead then it will involve the construction of manufacturing buildings with total floor area of 162,536 square metres, as well as a multi-storey car park accommodating 2,200 cars. A chemical store and other supporting work buildings are also part of the proposals.
Intel will pay Kildare County Council more than €27 million to improve the road connections to the site.
Leaked slides from Intel have been posted on the 3DCentre forum and show what Intel wants to be its ARM challenger, Bay Trail, in all its glory.
Bay Trail will be hitting the shops in 2014 and is expected to cause ARM a headache in what is left of the netbook and nettop space. This is the sort of headache which ARM will worry about. The netbook and nettop space will be non-existent by then, but the question is whether or not the spec will make it into new form factors or cause a resurgence in the netbook market.
Intel is trying to run the segment with Cedar Trail, which is a 32nm dual-core platform
Bay Trail means a shift to a quad-core, 22nm, out-of-order design. It speeds up the CPU with burst modes of up to 2.7GHz. It will also feature Intel’s own graphics processor instead of a licensed core from Imagination Technologies.
The slide suggests that TDPs will be 4.6-5W range at the low end, with 10W parts at the upper. It will also have full HD encode/decode, DDR3-1333L (low-power) support, and external resolutions of up to 2560×1600.
Bay Trail will give AMD’s Kabini a run for its money. Kabini is a 28nm, quad-core, expected to appear in the middle of this year.
Overall performance could jump by 50-70 percent which might mean that moves by Asus and Acer to kill off netbooks were premature.
The upgrades, however late they are, could breathe life into the dying market. If AMD tries something similar with Kabini then there would be pressure for the hardware makers to start to push something like a netbook onto the world.
Chipzilla plans to show off limited edition 3rd Generation Cores at CES after its plans for Internet telly were squashed by Big Content.
Everyone had been expecting to see Intel’s latest plan to kill off cable at CES, but the plan was abandoned because Big Content had been refusing to sign agreements with Chipzilla.
Instead Intel is going to make the more surprising move of showing off some chips at the show and a few ultrabooks and tablets from some vendors.
The limited run versions will be based on Ivy Bridge but will have a power rating below 10 watts. With power management controls the processor to throttle down to a sub-10-watt power envelope.
According to an Intel spokesperson who was leaking to CNET, the chips will enable new ultrabook designs. They are expected to appear soon after the show.
Intel’s mainstream Core series processors have ratings no lower than 17 watts — those versions are used in ultrabooks and Apple’s MacBook Air laptops today. Making them work below 10 watts is the realm of the Atom chip.
It will mean that ultrabooks will have a better battery life and be even thinner.
At the moment Intel is not saying how far below 10 watts these special “Y” series Ivy Bridge processors will go. Apparently they will be described in detail by Intel vice president Kirk Skaugen at CES.
The souped up Ivy Bridge chips were mentioned at Intel’s annual developer conference last year. At the time no one was certain if Intel and its partners would go ahead with the project.
Also on the agenda is the future Haswell processor — due later in the year. Y versions of that chip are expected to offer even better power efficiency than the low-power Ivy Bridge chips.
It is for this reason that the limited edition chips will be thin on the ground. Intel does not want to spoil the launch of Haswell later in the year.