AmorphousMemoryMark for macOS - measures memory read/write performance

Home | Store | Forum | About Us | Help

Amorphous
MemoryMark 3.0

Download

Date: January 5, 2022
OS: macOS 10.9.5 or later
License: Free

AmorphousMemoryMark measures memory read/write performance in GB/s.

Sequential 1 MiB block read/write tests with Threads=1, 2, 4, 8, 16, 32, or 64
Sequential 1 MiB block read/write tests
Random sequence 4 KiB block read/write tests with Threads=1, 2, 4, 8, 16, 32, or 64
Random sequence 4 KiB block read/write tests
Read/write GB/s scores
Test data: random or zero. This only applies to the read measurement. The write measurement always uses zeros.
Test size: 128 MiB, 256 MiB, 512 MiB, 1 GiB, 2 GiB, 4 GiB, 8 GiB, 16 GiB, 32 GiB, or 64 GiB
Test interval: 0 seconds, ..., 10 seconds

GB/s: gigabytes per second: 1,000,000,000 bytes per second

KiB: Kibibytes: 1,024 bytes

MiB: Mebibytes: 1,024 x 1,024 bytes: 1,048,576 bytes

GiB: Gibibytes: 1,024 x 1,024 x 1,024 bytes: 1,073,741,824 bytes

Visit AmorphousMemoryMark Forum.

Instructions:

Click the “All” button.
When the test is finished, press command - S to save a screenshot of the AmorphousMemoryMark window.
Write an Amazon review of your memory module, and attach the screenshot if you like.

Technical Details:

memmove()/bzero(): Executes the macOS provided memmove() function which reads from memory and write to memory for the read measurement, and the macOS provided bzero() function which writes zeros to memory for the write measurement. Both functions are optimized for each CPU architecture.
rep movsb/stosb (Intel only): Executes the "rep movsb" instruction which repeats a read-from-memory-write-to-memory operation, and the "rep stosb" instruction which repeats a write-to-memory operation. Both instructions are optimized on some CPU architectures with the Enhanced REP MOVSB and STOSB (ERMS) feature.
temporal load/store: Executes the temporal load instruction (Intel: movaps, Apple Silicon: ldp) which reads from memory through the memory cache, and the temporal store instruction (Intel: movaps, Apple Silicon: stp) which writes to memory through the memory cache.
non-temporal load/store: Executes the non-temporal load instruction (Intel: movntdqa, Apple Silicon: ldnp) which reads from memory without going through the memory cache, and the non-temporal store instruction (Intel: movntdq, Apple Silicon: stnp) which writes to memory without going through the memory cache.

Measurement result examples are available in the forum.

A permission was given by the author of CrystalDiskMark to use the identical or similar user interface in this app.

AmorphousMemoryMark screenshot

AmorphousMemoryMark saved score image

AmorphousMemoryMark memory module text field tooltip

AmorphousMemoryMark read/write result tooltip

-----------------------------------------------------------------------
AmorphousMemoryMark 1.0 (C) 2019-2020 Katsura Shareware
                       Katsura Shareware : http://katsurashareware.com/
-----------------------------------------------------------------------
* GB/s = 1,000,000,000 bytes per second
* MB = 1,000,000 bytes, MiB = 1,024 x 1,024 bytes
* KB = 1,000 bytes, KiB = 1,024 bytes

  Sequential Read 128KiB (T= 8) :   18.95 GB/s (memmove)
 Sequential Write 128KiB (T= 8) :   23.79 GB/s (bzero)
        Random Read 4KiB (T= 8) :   20.04 GB/s (memmove)
       Random Write 4KiB (T= 8) :   11.09 GB/s (bzero)
     Sequential Read 1MiB (T=1) :   10.06 GB/s (memmove)
    Sequential Write 1MiB (T=1) :   16.37 GB/s (bzero)
         Random Read 4KiB (T=1) :   10.33 GB/s (memmove)
        Random Write 4KiB (T=1) :    7.85 GB/s (bzero)

  Sequential Read 128KiB (T= 8) :   19.09 GB/s (rep movsb)
 Sequential Write 128KiB (T= 8) :   23.88 GB/s (rep stosb)
        Random Read 4KiB (T= 8) :   19.99 GB/s (rep movsb)
       Random Write 4KiB (T= 8) :   23.08 GB/s (rep stosb)
     Sequential Read 1MiB (T=1) :   10.34 GB/s (rep movsb)
    Sequential Write 1MiB (T=1) :   16.58 GB/s (rep stosb)
         Random Read 4KiB (T=1) :   10.06 GB/s (rep movsb)
        Random Write 4KiB (T=1) :   11.25 GB/s (rep stosb)

  Sequential Read 128KiB (T= 8) :   19.99 GB/s (temporal)
 Sequential Write 128KiB (T= 8) :   10.97 GB/s (temporal)
        Random Read 4KiB (T= 8) :   19.81 GB/s (temporal)
       Random Write 4KiB (T= 8) :   11.22 GB/s (temporal)
     Sequential Read 1MiB (T=1) :   14.28 GB/s (temporal)
    Sequential Write 1MiB (T=1) :    8.71 GB/s (temporal)
         Random Read 4KiB (T=1) :   10.43 GB/s (temporal)
        Random Write 4KiB (T=1) :    7.90 GB/s (temporal)

  Sequential Read 128KiB (T= 8) :   19.96 GB/s (non-temporal)
 Sequential Write 128KiB (T= 8) :   22.12 GB/s (non-temporal)
        Random Read 4KiB (T= 8) :   20.21 GB/s (non-temporal)
       Random Write 4KiB (T= 8) :   22.30 GB/s (non-temporal)
     Sequential Read 1MiB (T=1) :   14.07 GB/s (non-temporal)
    Sequential Write 1MiB (T=1) :   19.27 GB/s (non-temporal)
         Random Read 4KiB (T=1) :   10.30 GB/s (non-temporal)
        Random Write 4KiB (T=1) :   14.90 GB/s (non-temporal)

    Test : 512 MiB  (x5)  [Interval=0 sec]
  Device : DDR3-1600 SK Hynix HMT41GS6AFR8A-PB
    Date : 2020-01-22T17:09:37Z
      OS : macOS 10.14.6 18G1012

You can also copy/paste the measurements in text.

1x4GB (one-channel) vs. 2x4GB (two-channel) on Macmini5,1

AmorphousMemoryMark 3.0

Amorphous
MemoryMark 3.0