From WikiChip
Difference between revisions of "bitmain/sophon/bm1680"
< bitmain‎ | sophon

(bm1680)
 
Line 1: Line 1:
 
{{bitmain title|Sophon BM1680}}
 
{{bitmain title|Sophon BM1680}}
{{mpu}}
+
{{mpu
 +
|name=Sophon BM1680
 +
|no image=No
 +
|designer=Bitmain
 +
|manufacturer=TSMC
 +
|model number=BM1680
 +
|market=Artificial Intelligence
 +
|first announced=October 25, 2017
 +
|family=Sophon
 +
|process=28 nm
 +
|technology=CMOS
 +
|tdp=41 W
 +
|temp min=0 °C
 +
|temp max=125  °C
 +
|package module 1={{packages/bitmain/fcbga-1599}}
 +
}}
 
'''Sophon BM1680''' is a [[neural processor]] designed by [[Bitmain]] capable of performing both network inference and network training.
 
'''Sophon BM1680''' is a [[neural processor]] designed by [[Bitmain]] capable of performing both network inference and network training.

Revision as of 00:41, 7 November 2017

Template:mpu Sophon BM1680 is a neural processor designed by Bitmain capable of performing both network inference and network training.

Has subobject
"Has subobject" is a predefined property representing a container construct and is provided by Semantic MediaWiki.
Sophon BM1680 - Bitmain#package +
core voltage0.9 V (9 dV, 90 cV, 900 mV) +
core voltage tolerance5% +
designerBitmain +
familySophon +
first announcedOctober 25, 2017 +
first launchedNovember 8, 2017 +
full page namebitmain/sophon/bm1680 +
has ecc memory supporttrue +
instance ofmicroprocessor +
io voltage1.8 V (18 dV, 180 cV, 1,800 mV) +
io voltage tolerance5% +
ldateNovember 8, 2017 +
main imageFile:bitmain sophon bm1680.png +
manufacturerTSMC +
market segmentArtificial Intelligence +
max memory16,384 MiB (16,777,216 KiB, 17,179,869,184 B, 16 GiB, 0.0156 TiB) +
max memory bandwidth79.47 GiB/s (81,377.28 MiB/s, 85.33 GB/s, 85,330.263 MB/s, 0.0776 TiB/s, 0.0853 TB/s) +
max memory channels4 +
max operating temperature125 °C +
min operating temperature0 °C +
model numberBM1680 +
nameSophon BM1680 +
packageFCBGA-1599 +
peak flops (single-precision)2,000,000,000,000 FLOPS (2,000,000,000 KFLOPS, 2,000,000 MFLOPS, 2,000 GFLOPS, 2 TFLOPS, 0.002 PFLOPS, 2.0e-6 EFLOPS, 2.0e-9 ZFLOPS) +
process28 nm (0.028 μm, 2.8e-5 mm) +
supported memory typeDDR4-2666 +
tdp41 W (41,000 mW, 0.055 hp, 0.041 kW) +
tdp (typical)25 W (25,000 mW, 0.0335 hp, 0.025 kW) +
technologyCMOS +