[HN Gopher] A Comparison of ARM Cortex-A Series Processor Perfor...
       ___________________________________________________________________
        
       A Comparison of ARM Cortex-A Series Processor Performance
       Classifications
        
       Author : teleforce
       Score  : 25 points
       Date   : 2023-09-28 05:40 UTC (17 hours ago)
        
 (HTM) web link (www.forlinx.net)
 (TXT) w3m dump (www.forlinx.net)
        
       | dmitrygr wrote:
       | This is a museum piece, missing most of the last 7 years of ARM
       | cores
        
       | meepmorp wrote:
       | > This is ARM's latest A-Series processor released in 2016, the
       | Cortex-A73 supports full-size ARMv8-A architecture, and the
       | ARMv8-A is ARM's first processor architecture to support 64-bit
       | instruction sets, including ARM TrustZone, NEON, virtualization,
       | and encryption.
       | 
       | The article is 6-7 years old.
        
       | zauguin wrote:
       | (2017) (?) (Based on referring to the Cortex A73 as "ARM's latest
       | A-Series processor")
        
       | uxp8u61q wrote:
       | Blog posts without a date are a scourge.
        
       | schaefer wrote:
       | If this write-up has a target audience, I'm in it.
       | 
       | However... the graphics are pulled directly off the ARM
       | website[1] and IMO, having an automatic pop-up on the page that
       | says "talk to us about sales" subverts the good will that could
       | other wise have been generated.
       | 
       | Finally, anyone who tries to tell me that A53 is "high-end" in
       | 2023 will get some side-eye.
       | 
       | [1]: https://www.arm.com/products/silicon-ip-
       | cpu/cortex-a/cortex-...
        
         | dragontamer wrote:
         | > Finally, anyone who tries to tell me that A53 is "high-end"
         | in 2023 will get some side-eye.
         | 
         | You know that Microchip's SAM9x60 from 2020 is running
         | ARM926EJ-S (aka: ARMv5) with J for Jazelle instructions, right?
         | 
         | "Embedded" world moves slower than consumer world. ARM
         | Cortex-A53 is high-end for embedded. This is a world where 8051
         | and Z80 chips are still being made.
         | 
         | ----------
         | 
         | I feel like a lot of MPUs released in the last 2 years (i.MX
         | from NXP, STM32MP1 and soon-to-be-released MP2, etc. etc.) are
         | in the ARM-Cortex A7-ish tier.
        
           | RicoElectrico wrote:
           | Any advantage to use ARMv5 at this point? Knowing that ARM
           | designs a whole spectrum of processors over
           | performance/power/area triangle, has there really been no
           | alternative?
        
             | dragontamer wrote:
             | Its an eye-raising decision for sure, but this is literally
             | Microchip's most recent MPU design. Microchip has Cortex-A5
             | and Cortex-A7 chips available for those who want them. But
             | the most recently released chip is in fact SAM9X60.
             | 
             | My assumption is that Microchip has a large customer who is
             | demanding as perfect of a match to some 20+ year old
             | design, and Microchip is more than happy to produce this
             | ancient design for them.
             | 
             | ARM926EJ-S gives us a big hint: J for Jazelle. If some
             | customer has some old Embedded Jazelle/Java code that they
             | haven't bothered to update in the last 20 years, perhaps
             | its easier for Microchip to just keep making ARM926EJ-S
             | rather than forcing them to upgrade to Cortex-A5 or A7
             | (which although newer... these newer chips have all gotten
             | rid of Jazelle / hardware execution of Java code).
             | 
             | -------------------
             | 
             | Code-wise, there's probably a lot of downsides. But
             | electrically speaking, the SAM9X60 is an extremely well
             | designed BGA with well-matched VCC/GND pins that has a
             | 4-layer reference design. (Or at least: the SiP version
             | SAM9X60-D1G). In a world where 6, 8, or 10-layer designs
             | are commonly needed to route DDR-RAM, seeing a 4-layer
             | board reference design is pretty nuts (https://ww1.microchi
             | p.com/downloads/en/Appnotes/AN_3311_SAM9... as well as the
             | 4-layer DDR2 layout that was validated on: https://ww1.micr
             | ochip.com/downloads/en/Appnotes/AN_3310_Conn...).
             | 
             | Designing an 14mm x 14mm BGA223 that can be routed on a
             | 4-layer board is no small feat. And therefore, this is
             | probably one of the easiest BGAs to layout for a beginner,
             | and lowers your PCB production costs.
        
               | IvanMilatForPM wrote:
               | [dead]
        
       ___________________________________________________________________
       (page generated 2023-09-28 23:02 UTC)