From WikiChip
Difference between revisions of "bitmain/sophon"
(sophon) |
(→Edge) |
||
(6 intermediate revisions by the same user not shown) | |||
Line 16: | Line 16: | ||
}} | }} | ||
'''Sophon''' is a family of low-power [[neural processors]] designed by [[Bitmain]]. | '''Sophon''' is a family of low-power [[neural processors]] designed by [[Bitmain]]. | ||
+ | |||
+ | == 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. | ||
+ | |||
+ | == Models == | ||
+ | === High-performance === | ||
+ | <!-- 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|Peak Perf (SP)}} | ||
+ | {{#ask: [[Category:microprocessor models by bitmain]] [[family::Sophon]] [[market segment::!Edge Compute]] | ||
+ | |?full page name | ||
+ | |?model number | ||
+ | |?first launched | ||
+ | |?process | ||
+ | |?tdp | ||
+ | |?peak flops (single-precision)#TFLOPS | ||
+ | |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> | ||
+ | {{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> | ||
+ | {{comp table end}} | ||
+ | |||
+ | == See also == | ||
+ | * Intel Movidius {{movidius|Myriad}} | ||
+ | * Cambricon {{cambricon|MLU}} |
Latest revision as of 11:02, 25 December 2018
Sophon is a family of low-power neural processors designed by Bitmain.
Overview[edit]
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 | |||||
---|---|---|---|---|---|
Model | Launched | Process | TDP | Peak Perf (SP) | |
BM1680 | 8 November 2017 | 28 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 | |
BM1682 | March 2018 | 28 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 | ||
BM1684 | September 2018 | 12 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 | ||
BM1686 | June 2019 | 12 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 | |||||
---|---|---|---|---|---|
Model | Launched | Process | TDP (Typical) | Peak Perf (INT8) | |
BM1880 | 17 October 2018 | 2.5 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"
designer | Bitmain + |
first announced | October 25, 2017 + |
first launched | October 25, 2017 + |
full page name | bitmain/sophon + |
instance of | integrated circuit family + |
main designer | Bitmain + |
manufacturer | TSMC + |
name | Sophon + |
process | 28 nm (0.028 μm, 2.8e-5 mm) + and 12 nm (0.012 μm, 1.2e-5 mm) + |
technology | CMOS + |