Tag: finfet

Polaris finally shines today

polarisAMD finally starts flogging its new RX 480 GPU today.

For those who came in late, the RX 480 uses the company’s latest Polaris architecture which is built lt on 14nm FinFET process technology.

The starting price is $199 for the 4GB model and $239 for the 8GB and has some interesting performance characteristics. Compared to the GeForce GTX 970 which sells for $280, the RX 480 performs is about five to ten percent better. But when it comes to DX12 games like Gears of War, Hitman and Rise of the Tomb Raider it is about 40 per cent faster.

Compared to previous AMD products, the RX 480 is as fast as a Radeon R9 390 but uses just 150 watts compared to 275 watts for the previous generation.

Rivals Nvidia are expected to have a competing product based on Pascal available sometime in July, so AMD’s advantage may be short-lived; but in the meantime, the Radeon RX 480 is clearly the best GPU for $200.

Of course the world is also waiting to see AMD’s entry into the CPU league tables with the much touted never seen Zen chip, which should be in the shops in December.

Nvidia creates “miracle” deep learning chip

5136037690_97d228fa58Nvidia chief executive Jen-Hsun Huang announced that the company has created a new chip which can do five miracles – the Tesla P100 for deep-learning computing.

With 15 billion transistors, it’s the biggest FinFET chip Nvidia ever made. Huang told the throngs at the GPUTech conference in San Jose, California. He unveiled the chip after he said that deep-learning artificial intelligence chips have already become the company’s fastest-growing business.

Huang is claiming a lot for the chip saying it could do “five miracles.”  Not quite Jesus’s 37 but clearly Nvidia is catching up – although Huang’s definition of a miracle might be a little different from Christian myth.

“Three years ago, when we went all in, it was a leap of faith,” Huang said. “If we build it, they will come. But if we don’t build it, they won’t come.”

The chip has 15 billion transistors, or three times as much as many processors or graphics chips on the market. It takes up 600 square millimeters. The chip can run at 21.2 teraflops. Huang said that several thousand engineers laboured  on it for years.

“We decided to go all-in on A.I.,” Huang said. “This is the largest FinFET chip that has ever been done.”

Nvidia says it is shipping P100 to IBM, HPE, Dell, Cray, AI and cognitive cloud players, and key research institutions.

Huang showed a demo from Facebook that used deep learning to train a neural network how to recognize a landscape painting. They then used the network to create its own landscape painting.

He said that deep learning has become a new computing platform, and the company is dealing with hundreds of startups in the space that plan to take advantage of the platform.

“Our strategy is to accelerate deep learning everywhere,” Huang said.

Nvidia has also built a 170-teraflop DGX-1 supercomputer using the Tesla P100 chip.

“This is a beast of a machine, the densest computer ever made,” he said.

 

Samsung puts 10nm FinFET on roadmap

SamsungSamsung has put its next-generation 10nm FinFET manufacturing plans onto its roadmap.

Two months ago the outfit said that the processing node will be functional by the end of next year and Samsung may alter its schedule in order to prevent clients going to TSMC.

Kelvin Low from Samsung Foundry confirmed in a video posted on YouTube that Samsung has formally added 10nm FinFET into the process roadmap, for chip designers working in mobile, consumer or networking market segment the new chips will provide significant performance and power consumption improvements.

Samsung has already shown off its first 10nm wafers which was more of a symbolic move to send a message to major clients that Samsung is more than capable of getting its 10nm production lines up and running with no major difficulties.

 

TSMC reiterates 450mm push

Foundry-for-hire TSMC has reiterated plans to build its first 450mm wafer plant by 2017.

The new plant will offer FinFET transistor technology at 10nm and 7mm nodes and it should be completed in 2016 or 2017. The company also expects to adopt extreme ultraviolet lithography for the 10nm node by 2017. 

The company said that it will invest a record $1.5 billion on research and development this year, marking a new record. TSMC’s R&D spending totalled just over $1 billion in 2010, it hit $1.36 billion last year and it will continue to go up over the next few years. The company is speeding up 16nm and 10nm development and it hopes to start building 16nm FinFET chips next year.

Although some optimists were hoping to see 20nm parts from TSMC by the end of this year, CEO Morris Chang said the company will start volume production of 20nm products in 2014, reports Digitimes.

However, TSMC will not ditch 28nm anytime soon. In fact, it is still expanding 28nm capacity and it hopes to increase capacity to 100,000 300mm wafers by the end of the year. Chang said TSMC will expand its 28nm capacity threefold in 2013 over 2012.

TSMC and ARM tape out first ARM Cortex A57 chip

TSMC and ARM have taped out the first Cortex A57 processors built using FinFET process technology.

The A57 is not just another ARM SoC for mobiles. It is the fastest ARM chip to date and it can handle compute intensive applications beyond phones and tablets. 

In other words we should see A57 derivatives in servers and other gear. The processor is based on 64-bit ARMv8 processor cores and it is built in 16nm FinFET technology. 

“This first ARM Cortex-A57 processor implementation paves the way for our mutual customers to leverage the performance and power efficiency of 16nm FinFET technology,” said Tom Cronk , executive vice president and general manager, Processor Division, ARM. “The joint effort of ARM, TSMC, and TSMC’s OIP design ecosystem partners demonstrates the strong commitment to provide industry-leading technology for customer designs to benefit from our latest 64-bit ARMv8 architecture, big.LITTLE processing and ARM POP IP across a wide variety of market segments.”

TSMC believes the FinFET process is ready for the new chip, although it might take quite a bit of time before we see any kit based on ARM’s new chip. It is also unclear whether we will see consumer oriented FinFET SoC designs any time soon.

GloFo speeds up move to 14nm process

GlobalFoundries will be one of the first foundries to offer 14nm processors with customer tape outs in 2013.

GloFo has announced that it will start producing processors using 3D FinFET transistors with its 14nm-XM production, which it claims is optimised for mobile system-on-chip designs, with test production already underway.

Chips produced at this node will offer a 40-60 percent improvement in battery life compared to the two dimensional planar transistor being produced at 20nm by GloFo.  According to a statement, the chips will also incorporate elements of the 20nm LPM process.

GlobalFoundries’ relatively early date for 14nm production is likely to be good news for ARM, with the chip design firm recently signing a multiyear deal for FinFET development.

Dipesh Patel, deputy general manager of the Physical IP Division at ARM, commented that FinFET would be integral to the next generation of mobile devices, while its partnership with GloFo would create a  “platform which is well-suited for SoCs based on the next generation of ARM processors and GPUs for the mobile market”.

Intel is expected to be the first to reach volume production at 14nm. It has been investing in 14 nm fabs at Oregon, Arizona and in Leixlip, Ireland, and is expected to reach volume production early 2014.

ARM and GlobalFoundries announce 20nm FinFET partnership

ARM is further pressing ahead in FinFET technology – going into partnership with GlobalFoundries shortly after it announced a collaboration with TSMC to work on 64-bit FinFET servers.

The two have agreed on a multi year agreement for system on chip ARM processors using GloFo’s 20nm and FinFET process technologies. They have been working together on the Cortex platform for some time, including 20nm testing at GloFo’s Malta fab. This announcement essentially extends their work on driving IP to consumer designs on 20nm, as well  as promoting 3D FinFET transistor technology. “This joint development will enable a faster time to delivering SoC for customers using next-generation ARM CPUs and GPUs in mobil devices,” the companies said in a joint statement.

ARM believes that its own POP technology, along with 20nm-LPM and FinFET, will deliver more advanced, energy efficient, and faster semiconductor devices. POP has artisan IP standard cell libraries and memory cache instances, specifically designed for the ARM processor and the foundry technology. It also includes benchmarking reports to note the conditions and results gained through processor implementation across a range of configurations and design targets, as well as design utilities, floor plans, scripts, and a POP implementation guide, which lets customers evaluate their designs quickly and at a low risk, ARM claims.

When asked why GlobalFoundries was the company of choice for FinFET, a spokesperson told us that there are several key points.

Speaking with TechEye, Jason Gorss of GlobalFoundries, said that the company has over 10 years of R&D to build FinFETs on, as well as the common platform partnership with IBM and Samsung, meaning the company owns over three quarters of the industry patents on the technology. “We are confident that this heritage of deep R&D will allow us to lead the foundry volume ramp of FinFETs as we did with High K Metal Gate,” Gorss said.

According to Gorss, observers will find a “high level of synergy” between GloFo’s 20nm planar and FinFET, “which means we will be able to offer our customers the fastest path to FinFET with the least amount of risk”.

“We have developed a new approach to technology definition that will allow us to offer a cost-effective and power optimised FinFET technology that is ideally suited for the mobile market,” Gorss said.

ARM and TSMC plan 64-bit FinFET for servers

British chip IP company ARM has announced a collaboration with TSMC – which contributes to roughly two percent of the world’s total GDP  – on 64-bit FinFET processors.

The multi-year agreement will see the two continue their partnership beyond the 20nm process to build ARM on FinFET transistors, which the British company is confident will build its leadership in application processors, leading the charge in the fabless industry.

ARM and TSMC will work on optimising 64-bit processors based on the ARMv8 architecture, along with TSMC’s FinFET process, and ARM’s Artisan physical IP. The chips will be aimed squarely at mobile and enterprise markets, which both hold high performance and energy efficiency as a high priority. The fabless industry will be able to make the most of improved silicon process, physical IP and processor technology.

This announcement may well shake Intel’s timbers. ARM has been looming on the periphery of the enterprise and server market for some time and it is expected its market share will creep up. The fact is, many companies prefer the flexible, more open model ARM offers – as fabless companies build on their SoCs as they see fit.

ARM claims that its v8 architecture is specifically designed with energy efficient 64 bit execution to offer high end performance for mobile, enterprise, and server applications. Of course, high energy efficiency has been ARM’s go-to selling point for some time.

According to the companies, TSMC’s FinFET process will offer serious gains in speed and power as well as leakage reduction. They claim that the advantages offered by TSMC and ARM’s technology will help advance further scaling of SoC technology.

With TSMC deemed “too big to fail” by one top chip analyst and ARM’s expertise in IP design it is tough to see how the collaboration would not be threatening for their rivals.

TSMC told TechEye that Intel is its customer and that relationship remains unchanged, however, it offers no further comment on its customer’s businesses.