Arm Announces Cortex-A78AE CPU, Mali-G78AE GPU & Mali-C71AE ISP for autonomous automotive & industrial applications

Arm has announced new CPU, GPU, and ISP specifically designed for autonomous automotive and industrial applications with respectively Cortex-A78AE CPU, Arm Mali-G78AE GPU, and Arm Mali-C71AE ISP.

Arm Cortex-A78AE CPU

Cortex-A78AE

Key features and specifications:

  • Architecture – Armv8.2-A (AArch32 at ELO only)
  • Extensions – Armv8.1, Armv8.2, and Armv8.3 extensions (LDAPR instructions only), RAS extensions, Armv8.4 Dot Product, Cryptography extensions, RAS extensions
  • Microarchitecture
    • Up to 4x CPU cores per cluster
    • Out of order pipeline
    • Neon / Floating Point Unit included with INT8 Dot Product and IEEE FP16
    • Optional Cryptography Unit
    • 48-bit Physical Addressing (PA)
  • Memory system and external interfaces
    • 32kB to 64kB L1 I-Cache / D-Cache
    • 256kB to 512kB L2 Cache
    • Optional 512kB to 4MB L3 Cache
    • ECC Support
    • LPAE
    • Bus interfaces – AMBA ACE or CHI
    • Optional ACP, peripheral port
    • Functional Safety Support – ASIL D Systematic1 and ASIL D Diagnostic2
    • Security – TrustZone
    • Interrupts – External GICv4
    • Generic timer – Armv8-A
    • Non-intrusive debug (trace)

Cortex-A78AE succeeds Cortex-A76AE processors and delivers 30% performance improvement at the same power efficiency, 25% increased performance at the same total power, or 60% reduction in power at the same performance level.

Cortex-A78AE vs Cortex-A76AE

Cortex-A78AE supports DynamIQ, and the high-performance cores would typically be paired with Cortex-A65AE low-power cores in an automotive SoC. You’ll find more details on the developer page, and Arm community’s blog post.

Arm Mali-G78AE GPU

Arm Mali-G78AE Automotive GPU

Mali-G78EU is the very first autonomous/automotive GPU with from the company, is designed for functional safety, and offers the following key features:

  • Configurable from 1 to 24 cores
  • Anti-aliasing – 4x/8x/16x MSAA (Multi-Sampling Anti-Aliasing)
  • API support
    • OpenGL ES 1.1, 2.0, 3.1, 3.2
    • Vulkan 1.1, 1.2
    • OpenCL 1.1, 1.2, 2.0 Full profile
    • OpenGL SC 1.0, 2.0
  • Bus interface – AMBA 5 ACE, ACE-LITE, and AXI
  • L2 cache – Configurable 256KB – 4MB
  • Adaptive Scalable Texture Compression (ASTC) – Low Dynamic Range (LDR) and High Dynamic Range (HDR); supports both 2D and 3D images.
  • Arm Frame Buffer Compression (AFBC) – Version 1.3

The GPU supports automotive and industrial safety standards to help manufacturers meet ASIL B / SIL 2 requirements, and adds features like Flexible Partitioning and virtualization to isolate various graphics & compute tasks and increase reliability.

Arm GPU Flexible Partitioning
12-core Mali-G78AE MC12 GPU with different workloads

Again you’ll find more details both on Arm’s developer page and corresponding blog post.

Arm Mali-C71AE ISP

Mali-C71AE ISP

Autonomous vehicles and robots requires penty of camera inputs, so it makes perfect sense for Arm to introduce the Mali-C71AE Image Signal Processor (ISP) with support for up to four real-time cameras or 16 buffered cameras with a total throughput of 1.2 GP/s (Giga pixel per second).

The ISP also supports advanced spatial (2D) noise reduction (“Sinter”), per-exposure noise profiling and chromatic aberration correction, integrates 400 fault-detection circuits to make sure data is reliable, and provides supports for independent dynamic range management both for optimum human vision (display) and for computer vision (CV) applications. More in-depth information about Mali-C71AE ISP can be found in the announcement’s post.

The new Arm Mali-78AE processor, Mali-G78AE GPU and Mali-C71AE ISP should soon be found in SoCs used in applications when functional safety is critcal from smart manufacturing to enhancing ADAS and digital cockpit applications in automotive.

Share this:
FacebookTwitterHacker NewsSlashdotRedditLinkedInPinterestFlipboardMeWeLineEmailShare

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK Pi 4C Plus

Leave a Reply

Your email address will not be published. Required fields are marked *

Khadas VIM4 SBC
Khadas VIM4 SBC