Skip to content
\n

And this is exaclty the same hardware. Fresh installed DietPi with no addons. On Raspi-Lite is running a docker with ~8 Containers.

\n
root@DietPi:~# date\nThu May 30 13:20:17 BST 2024\nroot@DietPi:~# sudo curl https://raw.githubusercontent.com/TheRemote/PiBenchmarks/master/Storage.sh | sudo bash\n  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n                                 Dload  Upload   Total   Spent    Left  Speed\n100 36197  100 36197    0     0   359k      0 --:--:-- --:--:-- --:--:--  360k\nTrimming and syncing drives ...\n/boot/firmware: 95 MiB (99588096 bytes) trimmed on /dev/sda1\n/: 865.2 MiB (907272192 bytes) trimmed on /dev/sda2\nBoard information: Manufacturer: Raspberry Pi Foundation - Model: Raspberry Pi 5 Model B Rev 1.0 - Architecture: aarch64 - OS: Debian GNU/Linux 12 (bookworm)\nFetching required components ...\nClock speeds: CPU: 2400 - Core: 910\nSystem rootfs drive (/) has been detected as /dev/sda2 (sda2)\nStarting INXI hardware identification...\nSystem:\n  Kernel: 6.6.28+rpt-rpi-2712 arch: aarch64 bits: 64 compiler: gcc v: 12.2.0 Console: pty pts/2\n    Distro: Debian GNU/Linux 12 (bookworm)\nMachine:\n  Type: ARM System: Raspberry Pi 5 Model B Rev 1.0 details: N/A rev: c04170 serial: <filter>\nCPU:\n  Info: quad core model: N/A variant: cortex-a76 bits: 64 type: MCP arch: ARMv8 rev: 1 cache:\n    L1: 512 KiB L2: 2 MiB L3: 2 MiB\n  Speed (MHz): avg: 2100 min/max: 1500/2400 cores: 1: 2100 2: 2100 3: 2100 4: 2100 bogomips: 432\n  Features: Use -f option to see features\nGraphics:\n  Device-1: bcm2708-fb driver: N/A bus-ID: N/A\n  Device-2: bcm2712-hdmi0 driver: N/A bus-ID: N/A\n  Device-3: bcm2712-hdmi1 driver: N/A bus-ID: N/A\n  Display: server: No display server data found. Headless machine? tty: 80x40\n  API: N/A Message: No API data available in console. Headless machine?\nNetwork:\n  Device-1: driver: rp1 v: N/A port: N/A bus-ID: 01:00.0\n  IF-ID-1: eth0 state: up speed: 1000 Mbps duplex: full mac: <filter>\nDrives:\n  Local Storage: total: 119.24 GiB used: 1011.6 MiB (0.8%)\n  ID-1: /dev/sda vendor: Crucial model: M4-CT128M4SSD2 size: 119.24 GiB type: USB\n  Message: No optical or floppy data found.\nPartition:\n  ID-1: / size: 117.23 GiB used: 979.6 MiB (0.8%) fs: ext4 dev: /dev/sda2\nInfo:\n  Processes: 125 Uptime: 3m Memory: total: 4 GiB available: 3.95 GiB used: 159.9 MiB (4.0%)\n  igpu: 4 MiB Init: systemd target: graphical (5) Compilers: N/A Packages: 314 Client: Sudo\n  v: 1.9.13p3 inxi: 3.3.31\nRunning additional hardware identification tests...\nAdditional hardware identification tests completed.\nStarting mass storage device identification...\nRunning HDParm tests ...\n/dev/sda2:\n Timing O_DIRECT cached reads:   638 MB in  2.00 seconds = 318.73 MB/sec\n Timing O_DIRECT disk reads: 968 MB in  3.00 seconds = 322.41 MB/sec\nHDParm: 322.41 MB/sec - HDParmCached: 318.73 MB/sec\nRunning dd tests ...\n133120+0 records in\n133120+0 records out\n545259520 bytes (545 MB, 520 MiB) copied, 18.9546 s, 28.8 MB/s\nDD Write Speed: 28.8 MB/s\nRunning fio write test ...\nRunning fio read test ...\nFIO results - 4k RandWrite: 6714 IOPS (26859 KB/s) - 4k RandRead: 29767 IOPS (119069 KB/s)\nRunning iozone test ...\n        Iozone: Performance Test of File I/O\n                Version $Revision: 3.489 $\n                Compiled for 64 bit mode.\n                Build: linux\n        Run began: Thu May 30 13:21:04 2024\n        Auto Mode\n        Include fsync in write timing\n        O_DIRECT feature enabled\n        File size set to 81920 kB\n        Record Size 4 kB\n        Command line used: iozone -a -e -I -i 0 -i 1 -i 2 -s 80M -r 4k\n        Output is in kBytes/sec\n        Time Resolution = 0.000001 seconds.\n        Processor cache size set to 1024 kBytes.\n        Processor cache line size set to 32 bytes.\n        File stride size set to 17 * record size.\n                                                              random    random     bkwd    record    stride\n              kB  reclen    write  rewrite    read    reread    read     write     read   rewrite      read   fwrite frewrite    fread  freread\n           81920       4    26679    26653    39973    41363    16407    26781\niozone test complete.\nRandRead: 16407 - RandWrite: 26781 - Read: 39973 - Write: 26679\nEnter a description of your storage and setup (Example: Kingston A400 SSD on Pi 4 using StarTech SATA to USB adapter)\nDescription: Raspi5_diepti_fresh\n(Optional) Enter alias to use on benchmark results.  Leave blank for completely anonymous.\nAlias (leave blank for Anonymous): Sunnie_Micro\nResult submitted successfully and will appear live on https://pibenchmarks.com within a couple of minutes.\n\n     Category                  Test                      Result\nHDParm                    Disk Read                 322.41 MB/sec\nHDParm                    Cached Disk Read          318.73 MB/sec\nDD                        Disk Write                28.8 MB/s\nFIO                       4k random read            29767 IOPS (119069 KB/s)\nFIO                       4k random write           6714 IOPS (26859 KB/s)\nIOZone                    4k read                   39973 KB/s\nIOZone                    4k write                  26679 KB/s\nIOZone                    4k random read            16407 KB/s\nIOZone                    4k random write           26781 KB/s\n\n                          Score: 6157\n\nCompare with previous benchmark results at:\nhttps://pibenchmarks.com/\nroot@DietPi:~#\nroot@DietPi:~#\nroot@DietPi:~#\nroot@DietPi:~#\nroot@DietPi:~#\n------------------\n Benchmarks completed:                                                                                                │\n                                                │  - CPU performance : Duration = 2.13 seconds (lower is faster)                                                       │\n                                                │  - CPU temps       : Idle = 42 °C | Full load = 52 °C                                                                │\n                                                │  - RootFS I/O      : Write = 27 MiB/s | Read = 356 MiB/s                                                             │\n                                                │  - RAM I/O         : Write = 1156 MiB/s | Read = 4823 MiB/s                                                          │\n                                                │                                                                                                                      │\n                                                │ Compare these results online with other users, using the link below:                                                 │\n                                                │  - https://dietpi.com/survey#benchmark                                               ```\n
","upvoteCount":1,"answerCount":2,"acceptedAnswer":{"@type":"Answer","text":"

Since firmware and kernel are the same, there is two possible reasons/differences:

\n
    \n
  1. APM (Advanced Power Management) value set via hdparm: Most SSDs do not support APM at all, but we had one case, where it was supported, and where the default of 127 lead to slower disk I/O. 127 is the highest possible value which still allows idle spin down of spinning disks, hence we set it. Does this not report an error, and have an effect on your benchmark?\n
    hdparm -B 254 /dev/sda\n# or even (disabling APM compeletely)\nhdparm -B 255 /dev/sda
    \n
    2. \n
  2. The CPU governor, where DietPi uses schedutil by default, while RPi OS, AFIAK, still ondemand. schedutil is the more modern governor, meant as successor, which is more efficient, making more use of intermediate p-states, and with less processing overhead. But it indeed has weaknesses regarding disk and also RAM I/O on some SBCs. I observed it especially on big.LITTLE SBCs, where during I/O benchmarks, only the small cores' frequencies were raised, while the large ones remained at lower frequencies, negatively impacting I/O speeds. ondemand worked better in this case, even without the io_is_busy flag, somehow. Hence, check back at RPi OS, which governor it has, and in case test whether ondemand on DietPi leads to better results:\n
    cat /sys/devices/system/cpu/cpufreq/policy0/scaling_governor\necho ondemand > /sys/devices/system/cpu/cpufreq/policy0/scaling_governor
    \n
    3. \n
","upvoteCount":1,"url":"https://github.com/MichaIng/DietPi/discussions/7092#discussioncomment-9610466"}}}
Discussion options

Just2Confuse
May 30, 2024

You must be logged in to vote
Answered by MichaIng May 30, 2024

Since firmware and kernel are the same, there is two possible reasons/differences:

  1. APM (Advanced Power Management) value set via hdparm: Most SSDs do not support APM at all, but we had one case, where it was supported, and where the default of 127 lead to slower disk I/O. 127 is the highest possible value which still allows idle spin down of spinning disks, hence we set it. Does this not report an error, and have an effect on your benchmark? 
    hdparm -B 254 /dev/sda
# or even (disabling APM compeletely)
hdparm -B 255 /dev/sda
  2. The CPU governor, where DietPi uses schedutil by default, while RPi OS, AFIAK, still ondemand. schedutil is the more modern governor, meant as successor, which is mo…
View full answer

Replies: 2 comments 3 replies

Comment options

Just2Confuse
May 30, 2024
Author

You must be logged in to vote
0 replies
Comment options

MichaIng
May 30, 2024
Maintainer

You must be logged in to vote
3 replies
@Just2Confuse
Comment options

Just2Confuse May 31, 2024
Author

@Just2Confuse
Comment options

Just2Confuse Jun 1, 2024
Author

@MichaIng
Comment options

MichaIng Jun 5, 2024
Maintainer

Answer selected by Just2Confuse
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Category
Q&A
Labels
Question ❔ Investigating 🤔
2 participants
@Just2Confuse @MichaIng