News
 

Bookmark and Share

(8) 

Adapteva, a privately-held semiconductor technology start up, has announced that it is sampling the fourth generation of its Epiphany multi-core chip available for co-processing solutions to be implemented by industries requiring the next leap forward in parallel computing. The new chip boasts with 64 RISC cores and is made using 28nm super low-power (SLP) process technology at Globalfoundries.

“This is a watershed moment for low-power parallel computing, pushing many-core computing far ahead of what the market thought possible within this decade. With our latest OpenCL programmable 28nm chips we have created an easy-to-use multi-core processor solution that boosts energy efficiency for critical applications by an order of magnitude in a broad range of markets,” said Andreas Olofsson, chief executive officer of Adapteva.

The Epiphany IV chip includes 64 independent high-performance RISC cores on a single chip that operates at up to 800MHz, effectively offering over 50GHz of achievable programmable performance while consuming less than 2W in the "worst case" scenario. Peak performance of the Epiphany IV is around 100GFLOPS, whereas the peak core energy efficiency is 72GFLOPS/W at 500MHz.

Adapteva’s new OpenCL SDK provides a portable API for accessing the compute capabilities of a platform, accelerating performance in a wide spectrum of applications in market categories from gaming and entertainment to scientific and medical software.

Globalfoundries produces the chip using its 28nm-SLP (super low-power process) technology, which is believed to be one of the industry's most energy and cost efficient manufacturing process available today thanks to usage of gate-first high-k metal gate technology.

A key reason for the unmatched efficiency of the Epiphany-IV chips comes from early access to Globalfoundries leading-edge 28nm SLP technology. The industry to date has been challenged by the 28nm process node, but the Epiphany chip was able to exceed expectations in terms of performance, the two companies said.

Globalfoundries engineers started engaging Adapteva’s design teams back in 2010, which is two years prior to product launch. During the initial engagement phase, both teams rigorously discussed requirements and exchanged ideas in order to meet the stringent product leakage requirements while scaling the high performance needs from the previous generation. Time-to-market was obviously a priority, and it was essential for Adapteva to work on a 28nm technology that allows fast porting of their key IP blocks from the prior node. Glabalfoundries 28nm-SLP technology was the ideal fit for Adapteva, because it offers a combination of GHz class performance and low leakage to cover the  mobile segment as well as.

“The Epiphany efficiency breakthrough will enable significant savings in total cost of ownership for high-performance computing and will enable server level performance in mobile devices such as smart-phones and tablets,” added Mr. Olofsson.

Tags: Adapteva, Globalfoundries, 28nm, Semiconductor, Ephiphany

Discussion

Comments currently: 8
Discussion started: 09/04/12 01:49:49 PM
Latest comment: 09/06/12 01:38:31 PM
Expand all threads | Collapse all threads

[1-4]

1. 
What instruction set? If NEON is supported it could be the next killer GPU platform. 800 GFLOPS is Radeon 7750 territory which eats about 50W.
1 0 [Posted by: user99  | Date: 09/04/12 01:49:49 PM]
Reply
- collapse thread

 
I misread it 800 MHz not 800 GFLOPS. So not more than 3x power efficiency gain.
0 0 [Posted by: user99  | Date: 09/05/12 10:14:39 AM]
Reply

2. 
It's not ARM, it uses their own instruction set.

And it's not 800 GFLOPS, it's 100 GFLOPS (peak/theoretical)

And no, it's not good as GPU. GPU needs lots of dedicated HW this chip does not have.
2 0 [Posted by: hkultala  | Date: 09/04/12 02:59:27 PM]
Reply

3. 
Mix in MIPS Proaptive design with icubecorp opu design in this and you have super architecture that kills every thing else by far.
Hsa cud by dead before it even start!?
The only problem is software. It's final time to begin porting Dragonfly BSD to MIPS ISA.
1 0 [Posted by: Zola  | Date: 09/05/12 01:04:35 AM]
Reply

4. 
if that is 64 cores compatible with Linux i would say INTEL and AMD would be in their chairs shaking the fleas out in seconds. if it is not Linux compatible i do not see a need for it. unless INTEL and AMD CREATE A PORT allowing a co-processor that tasks can be push over to it and just keep waiting for the result of the task.
0 0 [Posted by: idonotknow  | Date: 09/06/12 12:11:07 PM]
Reply
- collapse thread

 
Linux is compiled for every architecture! Bat real problem is that there is no useful end user software to employ 64 trades. I don't se use of running let say 64 instances of angry birds, bat for cloud micro servers this is the real deal!
For high parallel multiprocessing we need new solutions and my best bet is DragonFly BSD.
1 0 [Posted by: Zola  | Date: 09/06/12 01:38:31 PM]
Reply

[1-4]

Add your Comment




Related news

Latest News

Monday, July 28, 2014

6:02 pm | Microsoft’s Mobile Strategy Seem to Fail: Sales of Lumia and Surface Remain Low. Microsoft Still Cannot Make Windows a Popular Mobile Platform

12:11 pm | Intel Core i7-5960X “Haswell-E” De-Lidded: Twelve Cores and Alloy-Based Thermal Interface. Intel Core i7-5960X Uses “Haswell-EP” Die, Promises Good Overclocking Potential

Tuesday, July 22, 2014

10:40 pm | ARM Preps Second-Generation “Artemis” and “Maya” 64-Bit ARMv8-A Offerings. ARM Readies 64-Bit Cores for Non-Traditional Applications

7:38 pm | AMD Vows to Introduce 20nm Products Next Year. AMD’s 20nm APUs, GPUs and Embedded Chips to Arrive in 2015

4:08 am | Microsoft to Unify All Windows Operating Systems for Client PCs. One Windows OS will Power PCs, Tablets and Smartphones