Neoverse N1

Accelerating the transformation to a scalable cloud-to-edge infrastructure.

Neoverse N1 block diagram

Getting Started

The Arm Neoverse N1 CPU combines server-class features and thread performance with cutting -edge low power design techniques delivering revolutionary performance per watt for building the next-generation cloud-to-edge infrastructure. By co-optimizing the N1 CPU with the CMN-600 coherent mesh interconnect that supports chip-to-chip connectivity over CCIX, the N1 platform allows extreme scalability from 8 to16 cores/chip for networking, storage, security and edge compute nodes, to 128+ cores/socket in hyperscale servers. Built on a common Armv8.2-A software foundation with the Neoverse E1 CPU, seamless heterogeneous systems can be realized encompassing accelerated functions for infrastructure deployments in 5G, data analytics, and smart networking. Integrated with Arm’s ML processor, the N1 CPU can deliver a self-hosted inference accelerator for AI/ML applications.

Key Features

  • 30% more performance/watt than Cortex-A72 on the same technology node
  • Large 1MB private L2 cache with up to 128MB system level cache in the mesh interconnect
  • Low-latency direct-connect interface between the CPU and mesh interconnect
  • I-cache coherency to enable a broader range of server workloads
  • Microarchitecture for large footprint, branch-heavy workloads
  • Double the vector and crypto compute bandwidth over previous generation
  • Manageable, full-frequency, sustainable performance without compromising compute efficiency
  • Server-class virtualization, RAS and code profiling

Specifications

Architecture Armv8-A (Harvard)
Extensions
  • Armv8.1 extensions
  • Armv8.2 extensions
  • Cryptography extensions
  • RAS extensions
  • SPE extensions
  • Armv8.3 LDAPR instructions
  • Armv8.4 Dot Product instructions
  • Armv8.5 (security vulnerabilities mitigations)
ISA support
  • A64
  • A32 and T32 (at the EL0 only)
Microarchitecture Pipeline Out-of-order
Superscalar  Yes
Neon/Floating Point Unit  Included 
Cryptography Unit  Optional 
Max number of CPUs in cluster  Four (4) or direct-connect
Physical addressing (PA)  48-bit 
Memory system and external interfaces L1 I-Cache/D-Cache 64kB
L2 Cache 1MB to 512kB, or 256kB
L3 Cache  Optional, 512kB TO 4MB 
ECC Support  Yes 
LPAE  Yes
Bus interfaces  AMBA CHI or ACE
ACP  Optional with cluster
Peripheral Port Optional with cluster
Other Security  TrustZone
Interrupts  GIC interface, GICv34 
Generic timer  Armv8-A 
PMU  PMUv3
Debug  Armv8-A (plus Armv8.2-A extensions) 
CoreSight  CoreSightv3 
Embedded Trace Macrocell ETMv4.2 (instruction trace)

Get Support

Arm Support

Arm training courses and on-site system-design advisory services enable licensees to efficiently integrate the Cortex-M35P processor into their design to realize maximum system performance with lowest risk and fastest time-to-market.

Arm training courses  Arm Design Reviews  Open a support case

Community Forums

Answered Where do I find presentations and photos from SC'18? 1 votes 885 views 0 replies Started 5 months ago by John Linford Answer this
Suggested answer Code is not run after loading into chip 0 votes 52 views 1 replies Latest 7 hours ago by Bojan Potocnik Answer this
Not answered 是的刚说的话个离开接电话说过考虑的就是 0 votes 18 views 0 replies Started 13 hours ago by asgadgadg Answer this
Not answered What is the "Integer divide unit with support for operand-dependent early termination"? 0 votes 28 views 0 replies Started 20 hours ago by jing Answer this
Answered Binary Semaphore upset by FIQ
  • Cortex-A
 0 votes 804 views 20 replies Latest 3 days ago by 42Bastian Schick Answer this
Suggested answer Vulkan shader problem on Mali Gpu: Accessing array elements inside an Uniform buffer returns all-zero values
  • Graphics Application
  • Mali Vulkan SDK
  • vulkan sdk
 0 votes 466 views 10 replies Latest 3 days ago by Hongkun Wang Answer this
Answered Where do I find presentations and photos from SC'18? Started 5 months ago by John Linford 0 replies 885 views
Suggested answer Code is not run after loading into chip Latest 7 hours ago by Bojan Potocnik 1 replies 52 views
Not answered 是的刚说的话个离开接电话说过考虑的就是 Started 13 hours ago by asgadgadg 0 replies 18 views
Not answered What is the "Integer divide unit with support for operand-dependent early termination"? Started 20 hours ago by jing 0 replies 28 views
Answered Binary Semaphore upset by FIQ Latest 3 days ago by 42Bastian Schick 20 replies 804 views
Suggested answer Vulkan shader problem on Mali Gpu: Accessing array elements inside an Uniform buffer returns all-zero values Latest 3 days ago by Hongkun Wang 10 replies 466 views