Jump to content

Zero ASIC: Difference between revisions

From Wikipedia, the free encyclopedia
Content deleted Content added
Monkbot (talk | contribs)
m Task 16: replaced (3×) / removed (0×) deprecated |dead-url= and |deadurl= with |url-status=;
clean up, typo(s) fixed: mm² → mm<sup>2</sup>
 
(24 intermediate revisions by 20 users not shown)
Line 1: Line 1:
{{Use mdy dates|date=May 2015}}
{{Use mdy dates|date=May 2015}}
{{Infobox company
{{Infobox company
| name = Adapteva, Inc
| name = Zero ASIC Corporation
| former_name = Adapteva, Inc.
| logo =
| logo =
| type =
| type =
Line 32: Line 33:
| divisions =
| divisions =
| subsid =
| subsid =
| homepage = {{URL|adapteva.com}}
| homepage = {{URL|zeroasic.com}}
| footnotes =
| footnotes =
| intl =
| intl =
}}
}}
'''Adapteva''' is a [[Fabless manufacturing|fabless]] [[semiconductor]] [[company]] focusing on low power [[many core]] [[microprocessor]] design. The company was the second company to announce a design with 1,000 specialized processing cores on a single [[integrated circuit]].<ref>{{cite news
'''Zero ASIC Corporation''', formerly '''Adapteva, Inc.''', is a [[Fabless manufacturing|fabless]] [[semiconductor]] [[company]] focusing on low power [[many core]] [[microprocessor]] design. The company was the second company to announce a design with 1,000 specialized processing cores on a single [[integrated circuit]].<ref>{{cite news
| title = Startup Has Big Plans for Tiny Chip Technology
| title = Startup Has Big Plans for Tiny Chip Technology
| publisher = Wall Street Journal
| publisher = Wall Street Journal
Line 49: Line 50:
</ref>
</ref>


Adapteva was founded in 2008 with the goal of bringing a ten times advancement in [[floating-point]] [[performance per watt]] for the mobile device market. Products are based on its Epiphany multi-core [[MIMD|multiple instruction, multiple data]] (MIMD) architecture and its Parallella [[Kickstarter]] project promoting "a supercomputer for everyone" in September 2012.
Adapteva was founded in 2008 with the goal of bringing a ten times advancement in [[floating-point]] [[performance per watt]] for the mobile device market. Products are based on its Epiphany multi-core [[multiple instruction, multiple data]] (MIMD) architecture and its Parallella [[Kickstarter]] project promoting "a supercomputer for everyone" in September 2012.
The company name is a combination of "adapt" and the Hebrew word "Teva" meaning nature.
The company name is a combination of "adapt" and the Hebrew word "Teva" meaning nature.


Line 55: Line 56:
Adapteva was founded in March 2008, by Andreas Olofsson. The company was founded with the goal of bringing a 10× advancement in [[floating-point]] processing [[Efficient energy use|energy efficiency]] for the [[mobile device]] market. In May 2009, Olofsson had a prototype of a new type of [[massively parallel]] multi-core [[computer architecture]]. The initial prototype was implemented in 65&nbsp;nm and had 16 independent microprocessor cores. The initial prototypes enabled Adapteva to secure US$1.5 million in series-A funding from BittWare, a company from [[Concord, New Hampshire]], in October 2009.<ref>{{cite web
Adapteva was founded in March 2008, by Andreas Olofsson. The company was founded with the goal of bringing a 10× advancement in [[floating-point]] processing [[Efficient energy use|energy efficiency]] for the [[mobile device]] market. In May 2009, Olofsson had a prototype of a new type of [[massively parallel]] multi-core [[computer architecture]]. The initial prototype was implemented in 65&nbsp;nm and had 16 independent microprocessor cores. The initial prototypes enabled Adapteva to secure US$1.5 million in series-A funding from BittWare, a company from [[Concord, New Hampshire]], in October 2009.<ref>{{cite web
| title = From RTL to GDSII in Just Six Weeks
| title = From RTL to GDSII in Just Six Weeks
| publisher = EETimes (via Wayback Machine)
| work = From RTL to GDSII in Just Six Weeks
| publisher = EETimes
| year = 2010
| year = 2010
| url = http://www.eetimes.com/electronics-blogs/other/4211089/From-RTL-to-GDSII-in-Just-Six-Weeks-
| url = https://www.eetimes.com/electronics-blogs/other/4211089/From-RTL-to-GDSII-in-Just-Six-Weeks-
| archive-url = https://web.archive.org/web/20101209044124/https://www.eetimes.com/electronics-blogs/other/4211089/From-RTL-to-GDSII-in-Just-Six-Weeks-
| accessdate = October 26, 2010}}</ref>
| accessdate = October 26, 2010| archive-date = December 9, 2010
}}</ref>


Adapteva's first commercial chip product started sampling to customers in early May 2011 and they soon thereafter announced the capability to put up to 4,096 cores on a single chip.
Adapteva's first commercial chip product started sampling to customers in early May 2011 and they soon thereafter announced the capability to put up to 4,096 cores on a single chip.
Line 66: Line 68:


== Products ==
== Products ==
Adapteva's main product family is the Epiphany scalable multi-core [[MIMD]] architecture. The Epiphany architecture could accommodate chips with up to 4,096 [[Reduced instruction set computing|RISC]] [[Out-of-order execution|out-of-order]] [[microprocessor]]s, all sharing a single [[32-bit]] flat memory space. Each RISC processor in the Epiphany architecture is [[superscalar]] with 64× 32-bit [[Processor register|unified register file]] (integer or [[single-precision]]) microprocessor operating up to 1&nbsp;[[Hertz|GHz]] and capable of 2&nbsp;[[GFLOPS]] (single-precision). Epiphany's RISC processors use a custom [[instruction set architecture]] (ISA) optimised for [[single-precision floating-point format|single-precision floating-point]],<ref>{{cite web
Adapteva's main product family is the Epiphany scalable multi-core [[Multiple instruction, multiple data|MIMD]] architecture. The Epiphany architecture could accommodate chips with up to 4,096 [[Reduced instruction set computing|RISC]] [[Out-of-order execution|out-of-order]] [[microprocessor]]s, all sharing a single [[32-bit]] flat memory space. Each [[RISC]] processor in the Epiphany architecture is [[superscalar]] with 64× 32-bit [[Processor register|unified register file]] (integer or [[single-precision]]) microprocessor operating up to 1&nbsp;[[Hertz|GHz]] and capable of 2&nbsp;[[GFLOPS]] (single-precision). Epiphany's RISC processors use a custom [[instruction set architecture]] (ISA) optimised for [[single-precision floating-point format|single-precision floating-point]],<ref>{{cite web
|title=Epiphany Architecture Reference Manual
|title=Epiphany Architecture Reference Manual
|url=http://www.adapteva.com/support/docs/e3-reference-manual/
|url=http://www.adapteva.com/support/docs/e3-reference-manual/
Line 74: Line 76:
}}</ref> but are programmable in high level [[ANSI C]] using a standard [[GNU Compiler Collection|GNU-GCC]] tool chain. Each RISC processor (in current implementations; not fixed in the architecture) has 32&nbsp;[[Kibibytes|KB]] of local memory. Code (possibly duplicated in each core) and stack space should be in that [[Scratchpad memory|local memory]]; in addition (most) temporary data should fit there for full speed. Data can also be used from other processor cores local memory at a speed penalty, or off-chip RAM with much larger speed penalty.
}}</ref> but are programmable in high level [[ANSI C]] using a standard [[GNU Compiler Collection|GNU-GCC]] tool chain. Each RISC processor (in current implementations; not fixed in the architecture) has 32&nbsp;[[Kibibytes|KB]] of local memory. Code (possibly duplicated in each core) and stack space should be in that [[Scratchpad memory|local memory]]; in addition (most) temporary data should fit there for full speed. Data can also be used from other processor cores local memory at a speed penalty, or off-chip RAM with much larger speed penalty.


The memory architecture is does not employ explicit hierarchy of [[CPU cache|hardware caches]], similar to the Sony/Toshiba/IBM [[cell processor]], but with the additional benefit of off-chip and inter-core loads and stores being supported (which simplifies porting software to the architecture). It is a hardware implementation of [[partitioned global address space]].{{Citation needed|date=September 2016}}
The memory architecture does not employ explicit hierarchy of [[CPU cache|hardware caches]], similar to the Sony/Toshiba/IBM [[Cell processor]], but with the additional benefit of off-chip and inter-core loads and stores being supported (which simplifies porting software to the architecture). It is a hardware implementation of [[partitioned global address space]].{{Citation needed|date=September 2016}}


This eliminated the need for complex [[cache coherency]] hardware, which places a practical limit on the number of cores in a traditional [[multicore system]]. The design allows the programmer to leverage greater foreknowledge of independent data access patterns to avoid the runtime cost of figuring this out. All processor nodes are connected through a [[network on chip]], allowing efficient message passing.<ref>{{cite web
This eliminated the need for complex [[cache coherency]] hardware, which places a practical limit on the number of cores in a traditional [[multicore system]]. The design allows the programmer to leverage greater foreknowledge of independent data access patterns to avoid the runtime cost of figuring this out. All processor nodes are connected through a [[network on chip]], allowing efficient message passing.<ref>{{cite web
| title = Startup Launches Manycore Floating Point Acceleration Technology
| title = Startup Launches Manycore Floating Point Acceleration Technology
| work = Startup Launches Manycore Floating Point Acceleration Technology
| publisher = HPCWire
| publisher = HPCWire
| year = 2011
| year = 2011
Line 104: Line 105:
| Sum&nbsp;GFLOPS || 32 || 102
| Sum&nbsp;GFLOPS || 32 || 102
|-
|-
| mm² || 8.96 || 8.2
| mm<sup>2</sup> || 8.96 || 8.2
|-
|-
| nm || 65 || 28
| nm || 65 || 28
Line 122: Line 123:


=== Parallella project ===
=== Parallella project ===
[[File:Adapteva_Parallella_DK02_-_Top_(15478282925).png|thumbnail|Parallella single-board computer with 16-core Epiphany chip and Zynq-7010 FPGA]]
[[File:Adapteva Parallella DK02 - Top (15478282925).png|thumbnail|Parallella single-board computer with 16-core Epiphany chip and Zynq-7010 FPGA]]
In September 2012, Adapteva started project Parallella on [[Kickstarter]], which was marketed as "''A Supercomputer for everyone''." Architecture reference manuals for the platform were published as part of the campaign to attract attention to the project.<ref>Andreas Olofsson, [http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/323691 Epiphany Documentation Release]</ref> The US$750,000 funding goal was reached in a month, with a minimum contribution of US$99 entitling backers to obtain one device; although the initial deadline was set for May 2013, the first single-board computers with 16-core Epiphany chip were finally shipped in December 2013.<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/703717 Update #46: First Parallella User Created Video]</ref>
In September 2012, Adapteva started project Parallella on [[Kickstarter]], which was marketed as "''A Supercomputer for everyone''." Architecture reference manuals for the platform were published as part of the campaign to attract attention to the project.<ref>Andreas Olofsson, [http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/323691 Epiphany Documentation Release]</ref> The US$750,000 funding goal was reached in a month, with a minimum contribution of US$99 entitling backers to obtain one device; although the initial deadline was set for May 2013, the first single-board computers with 16-core Epiphany chip were finally shipped in December 2013.<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/703717 Update #46: First Parallella User Created Video]</ref>


Size of board is planned to be {{convert|3.4|x|2.1|in|mm|abbr=on|disp=flip}}.<ref name="eetimes-2012-09-27">Rick Merritt, [http://www.eetimes.com/electronics-news/4396876/Parallela-project-aims-to-spawn-open-source-parallel-effort Adapteva Kickstarts Hundred-Dollar Supercomputer] // EETimes, September 27, 2012</ref><ref name="Andreas Olofsson describes Parallella">{{cite AV media |url=https://www.youtube.com/watch?v=taAG9QNWLNo |title=Parallella - Supercomputing for Everyone |format=slidecast |work=Adapteva founder & CEO Andreas Olofsson |date=September 28, 2012}}</ref><ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone Parallella: A Supercomputer For Everyone by Adapteva], Project page at Kickstarter</ref>
Size of board is planned to be {{convert|3.4|x|2.1|in|mm|abbr=on|disp=flip}}.<ref name="eetimes-2012-09-27">Rick Merritt, [http://www.eetimes.com/electronics-news/4396876/Parallela-project-aims-to-spawn-open-source-parallel-effort Adapteva Kickstarts Hundred-Dollar Supercomputer] // EETimes, September 27, 2012</ref><ref name="Andreas Olofsson describes Parallella">{{cite AV media |url=https://www.youtube.com/watch?v=taAG9QNWLNo |title=Parallella - Supercomputing for Everyone |format=slidecast |work=Adapteva founder & CEO Andreas Olofsson |date=September 28, 2012}}</ref><ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone Parallella: A Supercomputer For Everyone by Adapteva], Project page at Kickstarter</ref>


The Kickstarter campaign raised US$898,921.<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone Parallella: A Supercomputer For Everyone] // Kickstarter project, by Adapteva</ref><ref>Hiawatha Bray, [http://www.boston.com/business/technology/2012/12/03/adapteva-creates-efficient-cheap-microchip-with-help-from-kickstarter/IGJX8lYhHfHHIiC1MTUVyI/story.html Adapteva creates efficient, cheap microchip with help from Kickstarter. ‘Crowdfunding’ puts a tiny, fast computer closer to production] // The Boston Globe, December 2, 2012</ref> Raising US$3 million goal was unsuccessful, so no 64-core version of Parallella will be mass-produced.<ref name="linux.com.2013-01">Andrew Back, [https://www.linux.com/news/enterprise/systems-management/692990-introducing-the-99-linux-supercomputer Introducing the $99 Linux Supercomputer], Linux.com, January 24, 2013: "pledges of $99 or more being rewarded with at least one board with a 16-core device. ... The 16-core Epiphany chip delivers 26&nbsp;GFLOPS of performance and with the entire Parallella computer consuming only 5 watts"</ref> Kickstarter users having donated more than US$750 will get "parallella-64" variant with 64-core coprocessor (made from initial [[Multi-project wafer service|prototype manufacturing]] with 50 chips yield per wafer).<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/336213 64-core version of the Parallella board now offered!] // Adapteva blog at Kickstarter, October 25, 2012: "The Epiphany-IV (64+2) core Parallella board will be offered for pledges above $750. ... the fact that we only get 50 dies per wafer for these initial prototype runs. We can't disclose wafer pricing and yields at 28nm,"</ref>
The Kickstarter campaign raised US$898,921.<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone Parallella: A Supercomputer For Everyone] // Kickstarter project, by Adapteva</ref><ref>Hiawatha Bray, [http://www.boston.com/business/technology/2012/12/03/adapteva-creates-efficient-cheap-microchip-with-help-from-kickstarter/IGJX8lYhHfHHIiC1MTUVyI/story.html Adapteva creates efficient, cheap microchip with help from Kickstarter. ‘Crowdfunding’ puts a tiny, fast computer closer to production] // The Boston Globe, December 2, 2012</ref> Raising US$3 million goal was unsuccessful, so no 64-core version of Parallella will be mass-produced.<ref name="linux.com.2013-01">Andrew Back, [https://www.linux.com/news/enterprise/systems-management/692990-introducing-the-99-linux-supercomputer Introducing the $99 Linux Supercomputer] {{Webarchive|url=https://web.archive.org/web/20151117014717/https://www.linux.com/news/enterprise/systems-management/692990-introducing-the-99-linux-supercomputer |date=November 17, 2015 }}, Linux.com, January 24, 2013: "pledges of $99 or more being rewarded with at least one board with a 16-core device. ... The 16-core Epiphany chip delivers 26&nbsp;GFLOPS of performance and with the entire Parallella computer consuming only 5 watts"</ref> Kickstarter users having donated more than US$750 will get "parallella-64" variant with 64-core coprocessor (made from initial [[Multi-project wafer service|prototype manufacturing]] with 50 chips yield per wafer).<ref>[http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/336213 64-core version of the Parallella board now offered!] // Adapteva blog at Kickstarter, October 25, 2012: "The Epiphany-IV (64+2) core Parallella board will be offered for pledges above $750. ... the fact that we only get 50 dies per wafer for these initial prototype runs. We can't disclose wafer pricing and yields at 28nm,"</ref>


{| class="wikitable"
{| class="wikitable"
Line 142: Line 143:
|-
|-
! Processor
! Processor
| colspan="2" | [[Multi-core processor|Dual-core]] 32-bit [[ARM architecture|ARM]] [[Cortex-A9]] with [[ARM architecture#Advanced SIMD .28NEON.29|NEON]] at 1&nbsp;GHz (part of [[Xilinx#Zynq|Zynq]] Z7010 chip by Xilinx)
| colspan="2" | [[Multi-core processor|Dual-core]] 32-bit [[ARM architecture|ARM]] [[Cortex-A9]] with [[ARM architecture#Advanced SIMD (NEON)|NEON]] at 1&nbsp;GHz (part of [[Xilinx#Zynq|Zynq]] XC7Z010 chip by Xilinx)
| [[Multi-core processor|Dual-core]] 32-bit [[ARM architecture|ARM]] [[Cortex-A9]] with [[ARM architecture#Advanced SIMD .28NEON.29|NEON]] at 1&nbsp;GHz (part of [[Xilinx#Zynq|Zynq]] Z7020 chip by Xilinx)
| [[Multi-core processor|Dual-core]] 32-bit [[ARM architecture|ARM]] [[Cortex-A9]] with [[ARM architecture#Advanced SIMD (NEON)|NEON]] at 1&nbsp;GHz (part of [[Xilinx#Zynq|Zynq]] XC7Z020 chip by Xilinx)
|-
|-
! Coprocessor
! Coprocessor
Line 180: Line 181:
|-
|-
! Size
! Size
| colspan="3" | {{convert|3.5|x|2.1|x|0.625|mm|in|abbr=on}}
| colspan="3" | {{convert|3.5|x|2.1|x|0.625|in|mm|abbr=on}}
|-
|-
! SKU
! SKU
Line 195: Line 196:


=== Epiphany V ===
=== Epiphany V ===
By 2016, the firm had [[taped out]] a 1024-core [[64-bit computing|64-bit]] variant of their Epiphany architecture that featured: larger local stores (64&nbsp;KB), 64-bit addressing, [[double-precision floating-point format|double-precision floating-point]] arithmetic or [[SIMD]] single-precision, and 64-bit [[integer instruction]]s, implemented in the [[16nm|16&nbsp;nm process node]].<ref>{{cite web|title=epiphany v announcement|url=https://www.parallella.org/2016/10/05/epiphany-v-a-1024-core-64-bit-risc-processor/}}</ref> This design included instruction set enhancements aimed at [[deep-learning]] and [[cryptography]] applications. In July 2017, Adapteva's founder became a [[DARPA]] [https://www.darpa.mil/about-us/offices/mto MTO] program manager<ref>{{Cite web|url=https://www.darpa.mil/staff/mr-andreas-olofsson|title=Mr. Andreas Olofsson|last=Olofsson|first=Andreas|date=March 11, 2017|website=DARPA|archive-url=https://www.darpa.mil/staff/mr-andreas-olofsson|archive-date=March 11, 2017|url-status=live|access-date=December 16, 2018}}</ref> and announced that the Epiphany V was "unlikely" to become available as a commercial product.<ref>{{Cite web|url=http://www.adapteva.com/andreas-blog/adapteva-status/|title=Adapteva Status Update|last=Olofsson|first=Andreas|date=July 9, 2017|website=Adapteva Blog|archive-url=https://web.archive.org/web/20180423045817/http://www.adapteva.com/andreas-blog/adapteva-status/|archive-date=April 23, 2018|url-status=live|access-date=December 16, 2018}}</ref>
By 2016, the firm had [[taped out]] a 1024-core [[64-bit computing|64-bit]] variant of their Epiphany architecture that featured: larger local stores (64&nbsp;KB), 64-bit addressing, [[double-precision floating-point format|double-precision floating-point]] arithmetic or [[Single instruction, multiple data|SIMD]] single-precision, and 64-bit integer instructions, implemented in the 16&nbsp;nm process node.<ref>{{cite web|title=epiphany v announcement|url=https://www.parallella.org/2016/10/05/epiphany-v-a-1024-core-64-bit-risc-processor/}}</ref> This design included instruction set enhancements aimed at [[deep-learning]] and [[cryptography]] applications. In July 2017, Adapteva's founder became a [[DARPA]] [https://www.darpa.mil/about-us/offices/mto MTO] program manager<ref>{{Cite web|url=https://www.darpa.mil/staff/mr-andreas-olofsson|title=Mr. Andreas Olofsson|last=Olofsson|first=Andreas|date=March 11, 2017|website=DARPA|url-status=live|access-date=December 16, 2018|archive-url=https://web.archive.org/web/20170311170937/http://www.darpa.mil:80/staff/mr-andreas-olofsson |archive-date=March 11, 2017 }}</ref> and announced that the Epiphany V was "unlikely" to become available as a commercial product.<ref>{{Cite web|url=http://www.adapteva.com/andreas-blog/adapteva-status/|title=Adapteva Status Update|last=Olofsson|first=Andreas|date=July 9, 2017|website=Adapteva Blog|archive-url=https://web.archive.org/web/20180423045817/http://www.adapteva.com/andreas-blog/adapteva-status/|archive-date=April 23, 2018|url-status=live|access-date=December 16, 2018}}</ref>


=== Performance ===
=== Performance ===
The 16-core Parallella achieves roughly 5.0&nbsp;GFLOPS/W, and the 64-core Epiphany-IV made with 28&nbsp;nm estimated as 50&nbsp;GFLOPS/W (single-precision),<ref>{{cite web|url=http://www.hpcwire.com/2012/08/22/adapteva_unveils_64-core_chip/|title=Adapteva Unveils 64-Core Chip|last= Feldman|first=Michael|date=August 22, 2012|publisher=HPCWire|accessdate=September 3, 2014}}</ref> and 32-board system based on them achieves 15 GFLOPS/W.<ref>{{cite web|url=http://www.hpcwire.com/off-the-wire/adapteva-unveils-worlds-smallest-supercomputing-platform-isc14/|title=Adapteva Reveals A-1 Supercomputing Platform at ISC14|date=June 23, 2014|publisher=HPCWire, press-release of Adapteva|accessdate=September 3, 2014}}</ref> For comparison, top GPUs from AMD and Nvidia reached 10 GFLOPS/W for single-precision in 2009–2011 timeframe.<ref>{{cite web|url=http://www.karlrupp.net/2013/06/cpu-gpu-and-mic-hardware-characteristics-over-time/|title=CPU, GPU and MIC Hardware Characteristics over Time. Raw Compute Performance - Comparison of GFLOP/sec per Watt for single precision arithmetics. Higher is better.|date=June 24, 2013|publisher=Karl Rupp|accessdate=September 3, 2014}}<!-- this is a blog, but he has 113 citations according to scholar.google.com/citations?user=glyauuMAAAAJ --></ref>
{{POV section|date=June 2016}}
Joel Hruska from [[ExtremeTech]] had the following opinion about the 64-core Parallella project, prior to the 1024-core design: "Adapteva is drastically overselling what the Epiphany IV can actually deliver. 16–64 tiny cores with small amounts of memory, no local caches, and a relatively low clock speed can still be useful in certain workloads, but contributors aren't buying a supercomputer&nbsp;— they're buying the real-world equivalent of a self-sealing stem bolt."<ref>{{cite news |author=Joel Hruska |url=http://www.extremetech.com/electronics/137085-adapteva-turns-to-kickstarter-to-fund-massively-parallel-processor |title=Adapteva turns to Kickstarter to fund massively parallel processor |publisher=Extremetech |date=September 28, 2012}}</ref>

The criticism that the Epiphany chips cannot provide anywhere near the performance of modern supercomputers is nevertheless correct: actually, Epiphany chips with 16-cores or 64-cores and {{circa|25}} or 100&nbsp;GFLOPs in single-precision, respectively, do not even match the floating-point performance of modern desktop PC processors (Core i7-4770K (Haswell), 4× cores @ 3.5&nbsp;GHz AVX2: 177 GFLOPS<!-- usually listed – double-precision. Multiply CPU performance by 2 to get single-precision; coefficient from www.karlrupp.net ref -->,<ref>{{cite web |author=Dr. Donald Kinghorn |url=http://www.pugetsystems.com/blog/2013/08/26/Haswell-Floating-Point-Performance-493/ |title=Haswell Floating Point Performance |work=Puget Systems Blog |date=August 26, 2013}}</ref> double-precision){{snd}}a fact that is acknowledged by Adapteva.{{cn|date=November 2017}}

However, the latest Parallella boards with E16 Epiphany chips<ref>{{cite web |author=Andreas Olofsson |url=http://www.parallella.org/2014/07/14/new-parallella-product-offerings/ |title=New Parallella Product Offerings |publisher=Parallella Blog |date=July 14, 2014 |accessdate=September 3, 2014}}</ref> can be compared to many historic supercomputers in terms of raw performance (just as an example, the Cray 1{{snd}}the first supercomputer per se{{snd}}had a peak performance of 80&nbsp;[[MFLOPS]] at 1976, and its successor the Cray 2 had a peak performance of 1.9&nbsp;GFLOPS at 1985), and can certainly be used for parallel code development. The architectural similarities to supercomputers (message passing and [[non-uniform memory access|NUMA]]) make the Parallella a potentially useful development system, compared to traditional SMP machines.{{cn|date=November 2017}}

The point being that for a power envelope of 5&nbsp;W and in terms of GFLOPS/mm<sup>2</sup> of chip die space, the current E16 Epiphany chips provide vastly more performance than anything else available to date{{when|date=November 2017}}, with an architecture designed to scale, and applicable to more than just [[embarrassingly parallel]] GPU tasks.{{citation needed|date=September 2014}} (e.g. it would be capable of running the [[actor model]] with many concurrent, fully independent states). It is also suitable for DSP-like tasks where data could be fed directly on chip (from an FPGA or other ASIC) without having to create buffers in temporary memory as for a GPU), making it ideal for robotics & other intelligent sensor applications. The architecture also allows parallella boards to be combined into a cluster with a fast inter-chip 'eMesh' interconnect, extending the logical grid of cores (creating almost unlimited scaling potential).{{cn|date=November 2017}}

The 16-core Parallella has roughly 5.0&nbsp;GFLOPs/W, and the 64-core Epiphany-IV made with 28&nbsp;nm estimated as 50&nbsp;GFLOPs/W (single-precision),<ref>{{cite web|url=http://www.hpcwire.com/2012/08/22/adapteva_unveils_64-core_chip/|title=Adapteva Unveils 64-Core Chip|last= Feldman|first=Michael|date=August 22, 2012|publisher=HPCWire|accessdate=September 3, 2014}}</ref> and 32-board system based on them has 15 GFLOPS/W.<ref>{{cite web|url=http://www.hpcwire.com/off-the-wire/adapteva-unveils-worlds-smallest-supercomputing-platform-isc14/|title=Adapteva Reveals A-1 Supercomputing Platform at ISC14|date=June 23, 2014|publisher=HPCWire, press-release of Adapteva|accessdate=September 3, 2014}}</ref> For comparison, top GPUs from AMD and Nvidia reached 10 GFLOPs/W for single-precision in 2009–2011 timeframe.<ref>{{cite web|url=http://www.karlrupp.net/2013/06/cpu-gpu-and-mic-hardware-characteristics-over-time/|title=CPU, GPU and MIC Hardware Characteristics over Time. Raw Compute Performance - Comparison of GFLOP/sec per Watt for single precision arithmetics. Higher is better.|date=June 24, 2013|publisher=Karl Rupp|accessdate=September 3, 2014}}<!-- this is a blog, but he has 113 citations according to scholar.google.com/citations?user=glyauuMAAAAJ --></ref>


== See also ==
== See also ==
Line 227: Line 219:


[[Category:Computer companies of the United States]]
[[Category:Computer companies of the United States]]
[[Category:Computer hardware companies]]
[[Category:Companies based in Lexington, Massachusetts]]
[[Category:Companies based in Lexington, Massachusetts]]
[[Category:American companies established in 2008]]
[[Category:American companies established in 2008]]
[[Category:Electronics companies of the United States]]
[[Category:Semiconductor companies of the United States]]
[[Category:Reconfigurable computing]]
[[Category:Reconfigurable computing]]
[[Category:Manycore processors]]
[[Category:Manycore processors]]

Latest revision as of 22:45, 21 November 2024

Zero ASIC Corporation
FormerlyAdapteva, Inc.
IndustrySemiconductor industry
FoundedMarch 2008
FounderAndreas Olofsson
Headquarters,
US
Key people
Andreas Olofsson, CEO
ProductsCentral processing units
OwnerPrivately funded
Websitezeroasic.com

Zero ASIC Corporation, formerly Adapteva, Inc., is a fabless semiconductor company focusing on low power many core microprocessor design. The company was the second company to announce a design with 1,000 specialized processing cores on a single integrated circuit.[1][2]

Adapteva was founded in 2008 with the goal of bringing a ten times advancement in floating-point performance per watt for the mobile device market. Products are based on its Epiphany multi-core multiple instruction, multiple data (MIMD) architecture and its Parallella Kickstarter project promoting "a supercomputer for everyone" in September 2012. The company name is a combination of "adapt" and the Hebrew word "Teva" meaning nature.

History

[edit]

Adapteva was founded in March 2008, by Andreas Olofsson. The company was founded with the goal of bringing a 10× advancement in floating-point processing energy efficiency for the mobile device market. In May 2009, Olofsson had a prototype of a new type of massively parallel multi-core computer architecture. The initial prototype was implemented in 65 nm and had 16 independent microprocessor cores. The initial prototypes enabled Adapteva to secure US$1.5 million in series-A funding from BittWare, a company from Concord, New Hampshire, in October 2009.[3]

Adapteva's first commercial chip product started sampling to customers in early May 2011 and they soon thereafter announced the capability to put up to 4,096 cores on a single chip.

The Epiphany III, was announced in October 2011 using 28 nm and 65 nm manufacturing processes.

Products

[edit]

Adapteva's main product family is the Epiphany scalable multi-core MIMD architecture. The Epiphany architecture could accommodate chips with up to 4,096 RISC out-of-order microprocessors, all sharing a single 32-bit flat memory space. Each RISC processor in the Epiphany architecture is superscalar with 64× 32-bit unified register file (integer or single-precision) microprocessor operating up to 1 GHz and capable of 2 GFLOPS (single-precision). Epiphany's RISC processors use a custom instruction set architecture (ISA) optimised for single-precision floating-point,[4] but are programmable in high level ANSI C using a standard GNU-GCC tool chain. Each RISC processor (in current implementations; not fixed in the architecture) has 32 KB of local memory. Code (possibly duplicated in each core) and stack space should be in that local memory; in addition (most) temporary data should fit there for full speed. Data can also be used from other processor cores local memory at a speed penalty, or off-chip RAM with much larger speed penalty.

The memory architecture does not employ explicit hierarchy of hardware caches, similar to the Sony/Toshiba/IBM Cell processor, but with the additional benefit of off-chip and inter-core loads and stores being supported (which simplifies porting software to the architecture). It is a hardware implementation of partitioned global address space.[citation needed]

This eliminated the need for complex cache coherency hardware, which places a practical limit on the number of cores in a traditional multicore system. The design allows the programmer to leverage greater foreknowledge of independent data access patterns to avoid the runtime cost of figuring this out. All processor nodes are connected through a network on chip, allowing efficient message passing.[5]

Scalability

[edit]

The architecture is designed to scale almost indefinitely, with 4 e-links allowing multiple chips to be combined in a grid topology, allowing for systems with thousands of cores.

Multi-core coprocessors

[edit]
16-core Adapteva Epiphany chip, E16G301, from Parallella single-board computer

On August 19, 2012, Adapteva posted some specifications and information about Epiphany multi-core coprocessors.[6]

Technical info for    E16G301[7]   E64G401[8]
Cores 16 64
Core MHz 1000 800
Core GFLOPS 2 1.6
"Sum GHz" 16 51.2
Sum GFLOPS 32 102
mm2 8.96 8.2
nm 65 28
W def. 0.9 1.4
W max. 2 2

In September 2012, a 16-core version, the Epiphany-III (E16G301), was produced using 65 nm[9] (11.5 mm2, 500 MHz chip[10]) and engineering samples of 64-core Epiphany-IV (E64G401) were produced using 28 nm GlobalFoundries process (800 MHz).[11]

The primary markets for the Epiphany multi-core architecture include:

Parallella project

[edit]
Parallella single-board computer with 16-core Epiphany chip and Zynq-7010 FPGA

In September 2012, Adapteva started project Parallella on Kickstarter, which was marketed as "A Supercomputer for everyone." Architecture reference manuals for the platform were published as part of the campaign to attract attention to the project.[12] The US$750,000 funding goal was reached in a month, with a minimum contribution of US$99 entitling backers to obtain one device; although the initial deadline was set for May 2013, the first single-board computers with 16-core Epiphany chip were finally shipped in December 2013.[13]

Size of board is planned to be 86 mm × 53 mm (3.4 in × 2.1 in).[14][15][16]

The Kickstarter campaign raised US$898,921.[17][18] Raising US$3 million goal was unsuccessful, so no 64-core version of Parallella will be mass-produced.[19] Kickstarter users having donated more than US$750 will get "parallella-64" variant with 64-core coprocessor (made from initial prototype manufacturing with 50 chips yield per wafer).[20]

Parallella-16 Micro Server Parallella-16 Desktop Computer Parallella-16 Embedded Platform
Usage Ethernet connected headless server A personal computer Leading edge embedded systems
Processor Dual-core 32-bit ARM Cortex-A9 with NEON at 1 GHz (part of Zynq XC7Z010 chip by Xilinx) Dual-core 32-bit ARM Cortex-A9 with NEON at 1 GHz (part of Zynq XC7Z020 chip by Xilinx)
Coprocessor 16-core Epiphany III multi-core accelerator (E16)
Memory 1 GB DDR3L RAM
Ethernet 10/100/1000
USB 2× USB 2.0 (USB 2.0 HS and USB OTG)
Display HDMI
Storage 16 GB microSD
Expansion 2 eLinks + 24 GPIO 2 eLinks + 24 GPIO
FPGA 28K programmable logic cells
80 programmable DSP slices
80K programmable logic cells
220 programmable DSP slices
Weight 36 g (1.3 oz) 38 g (1.3 oz)
Size 3.5 in × 2.1 in × 0.625 in (88.9 mm × 53.3 mm × 15.9 mm)
SKU P1600-DK-xx P1601-DK-xx P1602-DK-xx
HTS Code 8471.41.0150
Power USB-powered (2.5 W) or 5 V DC (≈5 W)

Epiphany V

[edit]

By 2016, the firm had taped out a 1024-core 64-bit variant of their Epiphany architecture that featured: larger local stores (64 KB), 64-bit addressing, double-precision floating-point arithmetic or SIMD single-precision, and 64-bit integer instructions, implemented in the 16 nm process node.[21] This design included instruction set enhancements aimed at deep-learning and cryptography applications. In July 2017, Adapteva's founder became a DARPA MTO program manager[22] and announced that the Epiphany V was "unlikely" to become available as a commercial product.[23]

Performance

[edit]

The 16-core Parallella achieves roughly 5.0 GFLOPS/W, and the 64-core Epiphany-IV made with 28 nm estimated as 50 GFLOPS/W (single-precision),[24] and 32-board system based on them achieves 15 GFLOPS/W.[25] For comparison, top GPUs from AMD and Nvidia reached 10 GFLOPS/W for single-precision in 2009–2011 timeframe.[26]

See also

[edit]

References

[edit]
  1. ^ Clark, Don (May 3, 2011). "Startup Has Big Plans for Tiny Chip Technology". Wall Street Journal. Retrieved May 3, 2011.
  2. ^ "IBM says Kilocore technology will outrun today's mobile processors". Tom's Hardware. 2006.
  3. ^ "From RTL to GDSII in Just Six Weeks". EETimes (via Wayback Machine). 2010. Archived from the original on December 9, 2010. Retrieved October 26, 2010.
  4. ^ "Epiphany Architecture Reference Manual". Archived from the original on October 9, 2012.
  5. ^ "Startup Launches Manycore Floating Point Acceleration Technology". HPCWire. 2011. Retrieved May 3, 2011.
  6. ^ "Epiphany Multicore IP. Example Configurations". August 19, 2012.
  7. ^ Epiphany-III 16-core 65nm Microprocessor (E16G301) // admin (August 19, 2012)
  8. ^ Epiphany-IV 64-core 28nm Microprocessor (E64G401) // admin (August 19, 2012)
  9. ^ Silicon devices // Adapteva
  10. ^ Linley Gwennap, Adapteva: More Flops, Less Watts. Epiphany Offers Floating-Point Accelerator for Mobile Processors. // Microprocessor Report, June 2011
  11. ^ Michael Feldman, Adapteva Unveils 64-Core Chip // HPCWire
  12. ^ Andreas Olofsson, Epiphany Documentation Release
  13. ^ Update #46: First Parallella User Created Video
  14. ^ Rick Merritt, Adapteva Kickstarts Hundred-Dollar Supercomputer // EETimes, September 27, 2012
  15. ^ Parallella - Supercomputing for Everyone (slidecast). Adapteva founder & CEO Andreas Olofsson. September 28, 2012.
  16. ^ Parallella: A Supercomputer For Everyone by Adapteva, Project page at Kickstarter
  17. ^ Parallella: A Supercomputer For Everyone // Kickstarter project, by Adapteva
  18. ^ Hiawatha Bray, Adapteva creates efficient, cheap microchip with help from Kickstarter. ‘Crowdfunding’ puts a tiny, fast computer closer to production // The Boston Globe, December 2, 2012
  19. ^ Andrew Back, Introducing the $99 Linux Supercomputer Archived November 17, 2015, at the Wayback Machine, Linux.com, January 24, 2013: "pledges of $99 or more being rewarded with at least one board with a 16-core device. ... The 16-core Epiphany chip delivers 26 GFLOPS of performance and with the entire Parallella computer consuming only 5 watts"
  20. ^ 64-core version of the Parallella board now offered! // Adapteva blog at Kickstarter, October 25, 2012: "The Epiphany-IV (64+2) core Parallella board will be offered for pledges above $750. ... the fact that we only get 50 dies per wafer for these initial prototype runs. We can't disclose wafer pricing and yields at 28nm,"
  21. ^ "epiphany v announcement".
  22. ^ Olofsson, Andreas (March 11, 2017). "Mr. Andreas Olofsson". DARPA. Archived from the original on March 11, 2017. Retrieved December 16, 2018.
  23. ^ Olofsson, Andreas (July 9, 2017). "Adapteva Status Update". Adapteva Blog. Archived from the original on April 23, 2018. Retrieved December 16, 2018.
  24. ^ Feldman, Michael (August 22, 2012). "Adapteva Unveils 64-Core Chip". HPCWire. Retrieved September 3, 2014.
  25. ^ "Adapteva Reveals A-1 Supercomputing Platform at ISC14". HPCWire, press-release of Adapteva. June 23, 2014. Retrieved September 3, 2014.
  26. ^ "CPU, GPU and MIC Hardware Characteristics over Time. Raw Compute Performance - Comparison of GFLOP/sec per Watt for single precision arithmetics. Higher is better". Karl Rupp. June 24, 2013. Retrieved September 3, 2014.

Further reading

[edit]
[edit]