View source for NUMAlink

{{Short description|Computer network technology}}
{{Infobox computer hardware bus
| name        = NUMAlink
| fullname    = NUMAlink
| logo = 
| image       = 
| caption     = 
| invent-date = 1990 (as '''Stanford DASH'''), 1996 (as NUMAlink)
| invent-name = [[Silicon Graphics]]
| replaces    = 
| super-date  = 
| super-name  = 
| width       = 
| pin count   = 
| numdev      =
| speed       = 
| style       = 
| hotplug     = Optional
}}
'''NUMAlink''' is a system interconnect developed by [[Silicon Graphics]] (SGI) for use in its [[distributed shared memory]] [[ccNUMA]] computer systems. NUMAlink was originally developed by SGI for their [[SGI Origin 2000|Origin 2000]] and [[SGI Onyx2|Onyx2]] systems. At the time of these systems' introduction, it was branded as "CrayLink" during SGI's brief ownership of [[Cray Research]].<ref name="press">{{Cite news |title= Silicon Graphics and Cray Research Unveil Modular Origin Server Family: High-Bandwidth Systems Revolutionize Computer Buying Economics With Seamless Scalability |date= October 7, 1996 |work= Press release |url= http://www.sgi.com/Headlines/1996/October/originserver_release.html |url-status= dead |archiveurl= https://web.archive.org/web/19970707174608/http://www.sgi.com/Headlines/1996/October/originserver_release.html |archivedate= July 7, 1997 |accessdate= September 21, 2013 }}</ref>

[[Hewlett Packard Enterprise]] entered an [[original equipment manufacturer]] (OEM) arrangement with [[Silicon Graphics International]] (SGI) to use Numalink as the foundation in some mission critical servers.<ref>{{cite web|url=http://blog.sgi.com/hpe/|title=SGI Enters OEM Agreement with Hewlett Packard Enterprise (HPE) to Deliver SGI UV Technology through HPE Mission Critical Solutions - SGI Blog|date=9 February 2016|publisher=|access-date=11 August 2016|archive-url=https://web.archive.org/web/20161104192455/http://blog.sgi.com/hpe/|archive-date=4 November 2016|url-status=dead}}</ref>
{| class="wikitable"
|+
!Generation
!Introduction
!Bandwidth{{Efn-lr|name="both-directions"|Bandwidth per port in each direction (each NUMAlink port is a dual simplex channel).}}
!Systems
|-
|Stanford&nbsp;DASH
|~1990
|0.12 GB/s{{Efn-lr|DASH used separate request and reply mesh networks, so this bandwidth number is not directly comparable to NUMAlink.}}
|
|-
|NUMAlink&nbsp;2
|1996
|0.8 GB/s
|[[SGI Origin 200|Origin 200]], [[SGI Origin 2000|Origin 2000]], [[SGI Onyx2|Onyx2]] 
|-
|NUMAlink&nbsp;3
|2000
|1.6 GB/s
|[[SGI Origin 3000|Origin 3000]], [[Altix|Altix 3000]]
|-
|NUMAlink&nbsp;4
|2004
|3.2 GB/s
|[[Altix|Altix 4000]]
|-
|NUMAlink&nbsp;5
|2009
|7.5 GB/s
|[[SGI Altix|Altix UV]]
|-
|NUMAlink&nbsp;6
|2012
|6.7 GB/s
|UV 2000, UV 3000, UV 30
|-
|NUMAlink&nbsp;7
|2014
|14.9 GB/s
| UV 30EX, UV 300 / HPE Integrity MC990 X
|-
|NUMAlink&nbsp;8
|2017
|13.3 GB/s
| (HPE) Superdome Flex, Compute Scale-up
|}
; Notes
{{Notelist-lr}}

== NUMAlink 2 ==
There was no NUMAlink 1, as SGI's engineers deemed the system interconnect used in the [[Stanford DASH]] to be the first generation NUMAlink interconnect. 
NUMAlink 2 (branded as CrayLink) was announced in October 1996 for the [[SGI Onyx2|Onyx2]] [[visualization (computer graphics)|visualization]] systems, the [[SGI Origin 200|Origin 200]] and the [[SGI Origin 2000|Origin 2000]] servers and supercomputers.<ref name="press" /> The NUMAlink 2 interface is the [[SGI Origin 2000#Hub ASIC|Hub ASIC]]. NUMAlink 2 is capable of 1.6 GB/s of peak bandwidth through two 800 MB/s, [[Positive emitter-coupled logic|PECL]] 400&nbsp;MHz 16-bit unidirectional links.<ref name="white">{{Cite web |title= SGI® NUMAlink™ Industry Leading Interconnect Technology |work= White paper |date= April 13, 2005 |url= http://www.sgi.com/pdfs/3771.pdf |url-status=dead |archiveurl= https://web.archive.org/web/20060328173509/http://www.sgi.com/pdfs/3771.pdf |archivedate= March 28, 2006 |accessdate= September 21, 2013 }}</ref>

== NUMAlink 3 ==

NUMAlink 3 is the third generation of the interconnect, introduced in 2000 and used in the [[SGI Origin 3000|Origin 3000]] and [[Altix|Altix 3000]]. NUMAlink 3 is capable of 3.2 GB/s of peak bandwidth through two 1.6 GB/s unidirectional links.<ref name="white" />
The name NUMAflex reflects the modular design approach around this time.<ref>{{Cite web |title= NUMAflex Modular Design Approach: A Revolution in Evolution |author= John Mashey |author-link= John Mashey |url= http://homepages.cwi.nl/~robertl/mash/numaflex |date= August 30, 2000 |accessdate= September 28, 2016 }}</ref>

== NUMAlink 4 ==

NUMAlink 4 is the fourth generation of the interconnect, introduced in 2004 and used in the [[Altix|Altix 4000]]. NUMAlink 4 is capable of 6.4 GB/s of peak bandwidth through two 3.2 GB/s unidirectional links.<ref name="white" />

== NUMAlink 5 ==

NUMAlink 5 is the fifth generation of the interconnect, introduced in 2009 and used in the [[SGI Altix|Altix UV]] series. NUMAlink 5 is capable of 15 GB/s of peak bandwidth through two 7.5 GB/s unidirectional links.<ref name="altix-uv-ds">{{cite web|url=http://www.sgi.com/pdfs/4185.pdf|title=SGI Altix UV|year=2009|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2009-11-18}}</ref>

== NUMAlink 6 ==

NUMAlink 6 is the sixth generation of the interconnect, introduced in 2012 and used in the [[SGI UV|SGI UV 2000]], [[SGI UV|SGI UV 3000]], [[SGI UV|SGI UV 30]]. NUMAlink 6 is capable of 6.7 GB/s of bidirectional peak bandwidth for up to 256 socket system and 64TB of coherent shared memory.<ref name="sgi-uv-2000-ds">{{cite web|url=https://www.sgi.com/pdfs/4552.pdf|title=SGI UV 2000 Datasheet|year=2015|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2016-02-10|archive-url=https://web.archive.org/web/20170517033759/http://www.sgi.com/pdfs/4552.pdf|archive-date=2017-05-17|url-status=dead}}</ref><ref name="sgi-uv-3000-ds">{{cite web|url=https://www.sgi.com/pdfs/4555.pdf|title=SGI UV 3000, UV 30 Datasheet|year=2015|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2016-02-10|archive-url=https://web.archive.org/web/20170611092401/http://www.sgi.com/pdfs/4555.pdf|archive-date=2017-06-11|url-status=dead}}</ref>

== NUMAlink 7 ==

NUMAlink 7 is the seventh generation of the interconnect, introduced in 2014 and used in the HPE Integrity MC990 X/[[SGI UV|SGI UV 300]], [[SGI UV|SGI UV 30EX]], [[SGI UV|SGI UV 300H]], [[SGI UV|SGI UV 300RL]]. NUMAlink 7 is capable of 14.94 GB/s of bidirectional peak bandwidth for up to 64 socket system and 64TB of coherent shared memory.<ref name="sgi-uv-300-ds">{{cite web|url=https://www.sgi.com/pdfs/4530.pdf|title=SGI UV 300, UV 30EX Datasheet|year=2015|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2016-02-10}}</ref><ref name="sgi-uv-300h-ds">{{cite web|url=https://www.sgi.com/pdfs/4559.pdf|title=SGI UV 300H Datasheet|year=2015|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2016-02-10}}</ref><ref name="sgi-uv-300rl-ds">{{cite web|url=https://www.sgi.com/pdfs/4562.pdf|title=SGI UV 300RL Datasheet|year=2015|work=www.sgi.com|publisher=[[Silicon Graphics International]]|accessdate=2016-02-10}}</ref>

== NUMAlink 8 (Flex ASIC) ==

NUMAlink 8 is the eighth generation of the interconnect, introduced in 2017 and used in the HPE Superdome Flex and Compute Scale-up. NUMAlink 8 provides 13.3 GB/s of bandwidth per port<ref>{{Cite web |url=https://www.nas.nasa.gov/hecc/support/kb/endeavour-configuration-details_662.html |title=Endeavour Configuration Details |date=2021-08-05 |access-date=2023-10-28 |website=NASA Advanced Supercomputing (NAS) Division |quote=A Superdome Flex ASIC (also known as NUMAlink 8) provides 16 flex grid ports, each capable of 13.3 GB/s data rates for maximum flex grid bandwidth. The total bi-sectional crossbar grid bandwidth for a 32-socket Superdome Flex server is more than 850 GB/s.}}</ref> and systems using it are capable of 853.33 GB/s of bisection peak bandwidth (64 links are cut) across a 32 socket system with up to 48 TB of coherent shared memory.<ref name="hpe-flex-tnp">{{cite web|url=https://www.nextplatform.com/2017/11/06/hpes-superdome-gets-sgi-numalink-makeover/|title=HPE’s Superdome Gets An SGI NUMAlink Makeover|year=2017|work=www.nextplatform.com|publisher=The Next Platform|accessdate=2017-11-14}}</ref><ref name="hpe-flex-docs">{{cite web|url=https://www.hpe.com/h20195/v2/default.aspx?cc=us&lc=en&oid=1010323140|title=HPE Superdome Flex Server - Product documentation|year=2017|work=hpe.com|publisher=HPE|accessdate=2017-11-14}}</ref><ref name="HPE-HPC-2018">{{cite web|url=http://193.62.125.70/CIUK2017/HP.PDF|title=HPE High Performance Computing & AI Solutions|year=2017|work=www.hpe.com|publisher=[[Hewlett Packard Enterprise]]|accessdate=2018-03-24}}</ref>

==See also==
* [[List of device bit rates]]
* [[Compute Express Link]]
* [[InfiniBand]]
* [[RapidIO]]
* [[Myrinet]]
* [[QsNet II]]
* [[QuickRing]]

== References ==
{{Reflist}}

* Joseph Heinrich, ''[http://www.futuretech.blinkenlights.nl/origin/007-3439-002.pdf Origin and Onyx2 Theory of Operations Manual]'', 007-3439-002, Silicon Graphics.

== External links ==
* [https://hpe.com/superdome HPE NUMAlink servers], Superdome Flex and Compute Scale-up.
* {{webarchive |url=https://web.archive.org/web/20161028040934/http://www.sgi.com |date=October 28, 2016 |title=Silicon Graphics International, official website (pre-acquisition) }}
* {{webarchive |url=https://web.archive.org/web/20090327194141/http://www.sgi.com |date=March 27, 2009 |title=Silicon Graphics, official website (pre-acquisition) }}

{{Silicon Graphics}}

[[Category:Silicon Graphics]]