From WikiChip
Difference between revisions of "bitmain/sophon"
< bitmain

(Models)
(Edge)
 
(4 intermediate revisions by the same user not shown)
Line 18: Line 18:
  
 
== Overview ==
 
== Overview ==
 +
[[File:sophon roadmap (2018).jpg|thumb|right|2018 Roadmap]]
 
Announced in late 2017 at the AIWORLD 2017 Artificial Intelligence Conference in Beijing, Sophon is a family of low-power [[neural processors]]. [[Bitmain]] started exploring the field of artificial intelligence and neural processors as early as [[2015]]. By April 2017, their first product, the BM1680, has taped-out. Their chips are designed for both inference and training of neural networks, suitable for working with the common ANNs such as CNN, RNN, and DNN.
 
Announced in late 2017 at the AIWORLD 2017 Artificial Intelligence Conference in Beijing, Sophon is a family of low-power [[neural processors]]. [[Bitmain]] started exploring the field of artificial intelligence and neural processors as early as [[2015]]. By April 2017, their first product, the BM1680, has taped-out. Their chips are designed for both inference and training of neural networks, suitable for working with the common ANNs such as CNN, RNN, and DNN.
  
 
== Models ==
 
== Models ==
 +
=== High-performance ===
 
<!-- NOTE:  
 
<!-- NOTE:  
 
           This table is generated automatically from the data in the actual articles.
 
           This table is generated automatically from the data in the actual articles.
Line 32: Line 34:
 
{{comp table header|main|5:List of Sophon Neural Processors}}
 
{{comp table header|main|5:List of Sophon Neural Processors}}
 
{{comp table header|cols|Launched|Process|TDP|Peak Perf (SP)}}
 
{{comp table header|cols|Launched|Process|TDP|Peak Perf (SP)}}
{{#ask: [[Category:microprocessor models by bitmain]] [[family::Sophon]]
+
{{#ask: [[Category:microprocessor models by bitmain]] [[family::Sophon]] [[market segment::!Edge Compute]]
 
  |?full page name
 
  |?full page name
 
  |?model number
 
  |?model number
Line 44: Line 46:
 
  |mainlabel=-
 
  |mainlabel=-
 
}}
 
}}
{{comp table count|ask=[[Category:microprocessor models by bitmain]] [[family::Sophon]]}}
+
{{comp table count|ask=[[Category:microprocessor models by bitmain]] [[family::Sophon]] [[market segment::!Edge Compute]]}}
 +
</table>
 +
{{comp table end}}
 +
 
 +
=== Edge ===
 +
In late 2018, Bitmain introduced a new series targetting edge computing. Those chips have much lower power consumption.
 +
<!-- NOTE:
 +
          This table is generated automatically from the data in the actual articles.
 +
          If a microprocessor is missing from the list, an appropriate article for it needs to be
 +
          created and tagged accordingly.
 +
 
 +
          Missing a chip? please dump its name here: https://en.wikichip.org/wiki/WikiChip:wanted_chips
 +
-->
 +
{{comp table start}}
 +
<table class="comptable sortable tc4">
 +
{{comp table header|main|5:List of Sophon Neural Processors}}
 +
{{comp table header|cols|Launched|Process|TDP (Typical)|Peak Perf (INT8)}}
 +
{{#ask: [[Category:microprocessor models by bitmain]] [[family::Sophon]] [[market segment::Edge Compute]]
 +
|?full page name
 +
|?model number
 +
|?first launched
 +
|?process
 +
|?tdp (typical)
 +
|?peak integer ops (8-bit)#TOPS
 +
|format=template
 +
|template=proc table 3
 +
|userparam=6
 +
|mainlabel=-
 +
}}
 +
{{comp table count|ask=[[Category:microprocessor models by bitmain]] [[family::Sophon]] [[market segment::Edge Compute]]}}
 
</table>
 
</table>
 
{{comp table end}}
 
{{comp table end}}

Latest revision as of 12:02, 25 December 2018

Sophon
sophon logo.png
Sophon Logo
Developer Bitmain
Manufacturer TSMC
Type Neural Processors
Introduction October 25, 2017 (announced)
October 25, 2017 (launch)
Process 28 nm
0.028 μm
2.8e-5 mm
, 12 nm
0.012 μm
1.2e-5 mm
Technology CMOS

Sophon is a family of low-power neural processors designed by Bitmain.

Overview[edit]

2018 Roadmap

Announced in late 2017 at the AIWORLD 2017 Artificial Intelligence Conference in Beijing, Sophon is a family of low-power neural processors. Bitmain started exploring the field of artificial intelligence and neural processors as early as 2015. By April 2017, their first product, the BM1680, has taped-out. Their chips are designed for both inference and training of neural networks, suitable for working with the common ANNs such as CNN, RNN, and DNN.

Models[edit]

High-performance[edit]

 List of Sophon Neural Processors
ModelLaunchedProcessTDPPeak Perf (SP)
BM16808 November 201728 nm
0.028 μm
2.8e-5 mm
41 W
41,000 mW
0.055 hp
0.041 kW
2 TFLOPS
2,000,000,000,000 FLOPS
2,000,000,000 KFLOPS
2,000,000 MFLOPS
2,000 GFLOPS
0.002 PFLOPS
BM1682March 201828 nm
0.028 μm
2.8e-5 mm
3 TFLOPS
3,000,000,000,000 FLOPS
3,000,000,000 KFLOPS
3,000,000 MFLOPS
3,000 GFLOPS
0.003 PFLOPS
BM1684September 201812 nm
0.012 μm
1.2e-5 mm
6 TFLOPS
6,000,000,000,000 FLOPS
6,000,000,000 KFLOPS
6,000,000 MFLOPS
6,000 GFLOPS
0.006 PFLOPS
BM1686June 201912 nm
0.012 μm
1.2e-5 mm
9 TFLOPS
9,000,000,000,000 FLOPS
9,000,000,000 KFLOPS
9,000,000 MFLOPS
9,000 GFLOPS
0.009 PFLOPS
Count: 4

Edge[edit]

In late 2018, Bitmain introduced a new series targetting edge computing. Those chips have much lower power consumption.

 List of Sophon Neural Processors
ModelLaunchedProcessTDP (Typical)Peak Perf (INT8)
BM188017 October 20182.5 W
2,500 mW
0.00335 hp
0.0025 kW
2 TOPS
2,000,000,000,000 OPS
2,000,000,000 KOPS
2,000,000 MOPS
2,000 GOPS
0.002 POPS
Count: 1

See also[edit]

Facts about "Sophon - Bitmain"
designerBitmain +
first announcedOctober 25, 2017 +
first launchedOctober 25, 2017 +
full page namebitmain/sophon +
instance ofintegrated circuit family +
main designerBitmain +
manufacturerTSMC +
nameSophon +
process28 nm (0.028 μm, 2.8e-5 mm) + and 12 nm (0.012 μm, 1.2e-5 mm) +
technologyCMOS +