From WikiChip
Difference between revisions of "arm holdings/ethos"
< arm holdings

(ethos)
 
(N-Series)
 
(7 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
{{armh title|Ethos}}
 
{{armh title|Ethos}}
 +
{{family
 +
|name=Ethos
 +
|image=arm ethos.svg
 +
|developer=Arm Holdings
 +
|manufacturer=TSMC
 +
|manufacturer 2=Samsung
 +
|manufacturer 3=UMC
 +
|first announced=October 23, 2019
 +
|first launched=October 23, 2019
 +
|microarch=MLP
 +
}}
 
'''Ethos''' is a family of [[synthesizable]] [[neural processor]] IPs designed by [[Arm]] for IoT and edge applications.
 
'''Ethos''' is a family of [[synthesizable]] [[neural processor]] IPs designed by [[Arm]] for IoT and edge applications.
  
 
== Overview ==
 
== Overview ==
{{empty section}}
+
First introduced in late 2019, Ethos is a family of synthesizable [[neural processor]] IPs designed by Arm for various markets. The Ethos family represents the series of NPUs as part of {{armh|Project Trillium}}. The underlying microarchitecture for all the Ethos NPUs is the {{armh|MLP|l=arch}} which is designed to be scalable through various configurations based on the SRAM sizes and the number of compute engines.
  
 
== Members ==
 
== Members ==
 
=== N-Series ===
 
=== N-Series ===
{{empty section}}
+
[[File:ethos-n37,57,77 ce config.png|thumb|right|Ethos N37,57,77 CE Configurations.]]
 +
The Ethos-N series was introduced in October 2019. These are fully independent mainstream mobile and IoT NPUs which can be integrated into any SoC just like the {{\\|Cortex}} family. These NPUs come in a preconfigured number of quads and {{armh|MLP#Compute Engine (CE)|compute engines|l=arch}} which can scale from 1-4 [[TOPS]] and from 250 mW up to 1.5 W. Additionally, depending on the exact SKU, the size of the [[SRAM]] bank might be configurable as well. Multiple instances of those IPs may be combined using Arm's {{armh|CCN-500}} or {{armh|CMN-600}} interconnects to scale to higher performance. For example, eight of the {{armh|Ethos-N77|l=core}} can be integrated together on the {{armh|CCN-500}} to scale to 32.8 TOPS (or more using the {{armh|CMN-600}}).
 +
 
 +
 
 +
{| class="wikitable" style="text-align: center;"
 +
|-
 +
! colspan="2" | General || colspan="2" | Configuration || colspan="2" | SRAM || colspan="4" | Compute
 +
|-
 +
! NPU !! Introduction !! Quads !! {{armh|MLP#Compute Engine (CE)|CEs|l=arch}} || Bank || Total || MACs || OPS/clk || [[Int8]] || [[Int16]]
 +
|-
 +
| {{armh|Ethos-N37|l=core}} || Oct 23, 2019 || 1 || 4 || 128 KiB || 512 KiB || 512 MACs || 1024 OPs/clk || 1.024 TOPS || 256 GOPS
 +
|-
 +
| {{armh|Ethos-N57|l=core}} || Oct 23, 2019 || 2 || 8 || 64 KiB || 512 KiB || 1024 MACs || 2048 OPs/clk || 2.048 TOPS || 512 GOPS
 +
|-
 +
| {{armh|Ethos-N77|l=core}} || Oct 23, 2019 || 4 || 16 || 64 KiB<br>256 KiB || 1 MiB<br>4 MiB || 2048 MACs || 4096 OPs/clk || 4.096 TOPS || 1.024 TOPS
 +
|}
 +
 
 +
=== U-Series ===
 +
The Ethos-U series was introduced in early 2020. This series targets deeply-embedded AI applications. The 'microNPUs' in this series are not complete NPUs like the Ethos-N series. Instead, they feature a much slimmer design. The Ethos-U series is designed to tightly work with a companion Cortex-M processor such as the {{armh|Cortex-M55|l=arch}} (The M55 is preferred due to its  {{arm|Helium}} extension support but but other cores such as {{armh|Cortex-M7|M7|l=arch}}, {{armh|Cortex-M4|M4|l=arch}}, and the {{armh|Cortex-M33|M33|l=arch}} should do fine). Conceptually, the U series can be thought of as a single-{{armh|MLP#Compute Engine (CE)|compute engine (CE)|l=arch}} design whereby the {{armh|MLP#Programmable Layer Engine (PLE)|PLE|l=arch}} is removed and instead, relies on using a companion [[Cortex-M]] core for the extra processing. Due to the power and area constraints, the dedicated [[SRAM]] banks are also removed and instead rely on using the shared SoC (or ideally, the Cortex-M cache) for the weights and activations.
 +
 
 +
 
 +
{| class="wikitable" style="text-align: center;"
 +
|-
 +
! colspan="2" | General || colspan="2" | Configuration || colspan="4" | Compute
 +
|-
 +
! NPU !! Introduction !! SRAM || MACs || OPS/clk || Performance (Int8)
 +
|-
 +
| {{armh|Ethos-U55|l=core}} || February 10 || Shared with Cortex || 32-256 || 64-512 OPs/clk || 25.6-204.8 [[GOPS]] (@ 100-400 MHz, typical on [[55nm|50]]/[[40nm|40]] nm)<br>64-512 [[GOPS]] (@ 1 GHz, typical on [[16nm|16]]/[[7nm|7]]/[[5nm|5]] nm)
 +
|}
  
 
== See also ==
 
== See also ==
 
* Intel Habana {{habana|HL|HL Series}}
 
* Intel Habana {{habana|HL|HL Series}}

Latest revision as of 23:45, 10 February 2020

Edit Values
Ethos
arm ethos.svg
General Info
DesignerArm Holdings
ManufacturerTSMC, Samsung, UMC
IntroductionOctober 23, 2019 (announced)
October 23, 2019 (launched)
Microarchitecture
MicroarchitectureMLP

Ethos is a family of synthesizable neural processor IPs designed by Arm for IoT and edge applications.

Overview[edit]

First introduced in late 2019, Ethos is a family of synthesizable neural processor IPs designed by Arm for various markets. The Ethos family represents the series of NPUs as part of Project Trillium. The underlying microarchitecture for all the Ethos NPUs is the MLP which is designed to be scalable through various configurations based on the SRAM sizes and the number of compute engines.

Members[edit]

N-Series[edit]

Ethos N37,57,77 CE Configurations.

The Ethos-N series was introduced in October 2019. These are fully independent mainstream mobile and IoT NPUs which can be integrated into any SoC just like the Cortex family. These NPUs come in a preconfigured number of quads and compute engines which can scale from 1-4 TOPS and from 250 mW up to 1.5 W. Additionally, depending on the exact SKU, the size of the SRAM bank might be configurable as well. Multiple instances of those IPs may be combined using Arm's CCN-500 or CMN-600 interconnects to scale to higher performance. For example, eight of the Ethos-N77 can be integrated together on the CCN-500 to scale to 32.8 TOPS (or more using the CMN-600).


General Configuration SRAM Compute
NPU Introduction Quads CEs Bank Total MACs OPS/clk Int8 Int16
Ethos-N37 Oct 23, 2019 1 4 128 KiB 512 KiB 512 MACs 1024 OPs/clk 1.024 TOPS 256 GOPS
Ethos-N57 Oct 23, 2019 2 8 64 KiB 512 KiB 1024 MACs 2048 OPs/clk 2.048 TOPS 512 GOPS
Ethos-N77 Oct 23, 2019 4 16 64 KiB
256 KiB
1 MiB
4 MiB
2048 MACs 4096 OPs/clk 4.096 TOPS 1.024 TOPS

U-Series[edit]

The Ethos-U series was introduced in early 2020. This series targets deeply-embedded AI applications. The 'microNPUs' in this series are not complete NPUs like the Ethos-N series. Instead, they feature a much slimmer design. The Ethos-U series is designed to tightly work with a companion Cortex-M processor such as the Cortex-M55 (The M55 is preferred due to its Helium extension support but but other cores such as M7, M4, and the M33 should do fine). Conceptually, the U series can be thought of as a single-compute engine (CE) design whereby the PLE is removed and instead, relies on using a companion Cortex-M core for the extra processing. Due to the power and area constraints, the dedicated SRAM banks are also removed and instead rely on using the shared SoC (or ideally, the Cortex-M cache) for the weights and activations.


General Configuration Compute
NPU Introduction SRAM MACs OPS/clk Performance (Int8)
Ethos-U55 February 10 Shared with Cortex 32-256 64-512 OPs/clk 25.6-204.8 GOPS (@ 100-400 MHz, typical on 50/40 nm)
64-512 GOPS (@ 1 GHz, typical on 16/7/5 nm)

See also[edit]

Facts about "Ethos - ARM"
designerArm Holdings +
first announcedOctober 23, 2019 +
first launchedOctober 23, 2019 +
main imageFile:arm ethos.svg +
manufacturerTSMC +, Samsung + and UMC +
microarchitectureMLP +
nameEthos +