Re: NUMA Testing
Hi Dom,
I 'm using a 2GB Pi 4 doing a 7 minute compile, using all 4 cores & am not seeing any improvement, is this to be expected with only 2GB to play with? As far as I can tell NUMA is active, and numa=fake=1 was automatically added to my /proc/cmdline file. I have also tried changing it to numa=fake=2 but got exactly the same compile/build time.
Any thoughts?
Regards Kevin
I 'm using a 2GB Pi 4 doing a 7 minute compile, using all 4 cores & am not seeing any improvement, is this to be expected with only 2GB to play with? As far as I can tell NUMA is active, and numa=fake=1 was automatically added to my /proc/cmdline file. I have also tried changing it to numa=fake=2 but got exactly the same compile/build time.
Any thoughts?
Regards Kevin
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing
1GB and 2GB Pi4 are single rank, and we don't recommend changing BANKLOW on Pi4 as without iommus on the codec path, it hurts video decode performance (h264/hevc) and possibly camera.bleep42 wrote: ↑Fri Nov 01, 2024 3:13 pmI 'm using a 2GB Pi 4 doing a 7 minute compile, using all 4 cores & am not seeing any improvement, is this to be expected with only 2GB to play with? As far as I can tell NUMA is active, and numa=fake=1 was automatically added to my /proc/cmdline file. I have also tried changing it to numa=fake=2 but got exactly the same compile/build time.
Hence the default 1 numa region as there is no benefit available. A 4GB or 8GB Pi4 is dual rank and would benefit 2 numa regions.
If you don't use video hardware, you could try running with SDRAM_BANKLOW=1 which will default to 4 numa regions and may show a benefit.
But I'd describe this configuration as experimental, and not generally recommended.
Re: NUMA Testing
Hi Dom,
Thanks for your reply, for the sake of experimentation, I tried SDRAM_BANKLOW=1, which did indeed give me numa=fake=4, however, looking at the output from numactl --hardware and lscpu | grep -i numa it looked to me as though it had fowled up, and my compile time actually got worse, so I tried numa=fake=2 and that showed a speed up from 6:55 to 6:32, if you include the other memory tuning you said had been done, my compile time was 7:01, now with the settings indicated 6:32, so a roughly 7% speed up for some under the hood optimisations, tried a couple of videos using VLC and they seemed to play back ok, so it's looking good to me. :-) Maybe now I'll have to get a new Pi5.
Regards, Kevin.
Thanks for your reply, for the sake of experimentation, I tried SDRAM_BANKLOW=1, which did indeed give me numa=fake=4, however, looking at the output from numactl --hardware and lscpu | grep -i numa it looked to me as though it had fowled up, and my compile time actually got worse, so I tried numa=fake=2 and that showed a speed up from 6:55 to 6:32, if you include the other memory tuning you said had been done, my compile time was 7:01, now with the settings indicated 6:32, so a roughly 7% speed up for some under the hood optimisations, tried a couple of videos using VLC and they seemed to play back ok, so it's looking good to me. :-) Maybe now I'll have to get a new Pi5.
Regards, Kevin.
Re: NUMA Testing
Just to say it hasn't broken my Jellyfin .org setup on a Pi5 8GB.
No idea if it is faster or slower. Just that it still works.
No idea if it is faster or slower. Just that it still works.
Re: NUMA Testing
Has the NUMA-enabled kernel 6.6.62 just been released as a package update to all users? So one no more needs to do rpi-update, just the SDRAM_BANKLOW=, right?
Re: NUMA Testing
From what I understand, the 6.12.x kernel uses a completely different NUMA patch set.
viewtopic.php?p=2270416#p2270416
This would appear to imply that testing needs to start all over again to verify video and everything else works properly.
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing
Can you check "vcgencmd bootloader_version"? (and also "vcgencmd version" if on a Pi4). Newer than 23 Oct is wanted.
Re: NUMA Testing
Seems NUMA is active now for me with a regular apt update/full-upgrade:
https://browser.geekbench.com/v6/cpu/9140585
https://browser.geekbench.com/v6/cpu/9140585
Re: NUMA Testing. Is this way safe enough?
Hello everyone:
I enable NUMA follow the instruction of this topic,and got a high score with geekebnch.
I'm using a 8G pi5.Here is my configuration:
===========================================================
pi@Pi-5:~ $ vcgencmd version
2024年11月12日 16:10:44
Copyright (c) 2012 Broadcom
version 4b019946 (release) (embedded)
pi@Pi-5:~ $ vcgencmd bootloader_version
2024年11月12日 16:10:44
version 4b019946a06ea87c36d14fd6d702cc65fb458b9e (release)
timestamp 1731427844
update-time 1732995212
capabilities 0x0000007f
pi@Pi-5:~ $ vcgencmd bootloader_config
[all]
BOOT_UART=1
BOOT_ORDER=0xf461
NET_INSTALL_AT_POWER_ON=1
SDRAM_BANKLOW=1
pi@Pi-5:~ $ cat /proc/cmdline
reboot=w coherent_pool=1M 8250.nr_uarts=1 pci=pcie_bus_safe cgroup_disable=memory numa_policy=interleave numa=fake=8 system_heap.max_order=0 smsc95xx.macaddr=2C:CF:67:3C:6C:FD vc_mem.mem_base=0x3fc00000 vc_mem.mem_size=0x40000000 console=ttyAMA10,115200 console=tty1 root=PARTUUID=f580ed4f-02 rootfstype=ext4 fsck.repair=yes rootwait video=HDMI-A-1:1440x900M@60D plymouth.ignore-serial-consoles cfg80211.ieee80211_regdom=CN
pi@Pi-5:~ $
pi@Pi-5:/sys/class/gpio $ uname -a
Linux Pi-5 6.6.62+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.6.62-1+rpt1 (2024年11月25日) aarch64 GNU/Linux
===========================================================
To test memory stability ,I run 4 processes of memtester simultaneously . One of process report FAILURE at first run.
==================================================
pi@Pi-5:~ $ sudo memtester 1600M 2
memtester version 4.6.0 (64-bit)
Copyright (C) 2001-2020 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).
pagesize is 16384
pagesizemask is 0xffffffffffffc000
want 1600MB (1677721600 bytes)
got 1600MB (1677721600 bytes), trying mlock ...locked.
Loop 1/2:
Stuck Address : ok
Random Value : ok
Compare XOR : ok
Compare SUB : ok
Compare MUL : ok
Compare DIV : ok
Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : ok
Block Sequential : ok
Checkerboard : ok
Bit Spread : ok
Bit Flip : ok
Walking Ones : ok
Walking Zeroes : ok
8-bit Writes : |FAILURE: 0x57bd735cdc7eda8f != 0x57bd735ddc7eda8f at offset 0x000000002a229478.
16-bit Writes : ok
Loop 2/2:
Stuck Address : ok
Random Value : ok
Compare XOR : ok
Compare SUB : ok
Compare MUL : ok
Compare DIV : ok
Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : ok
Block Sequential : ok
Checkerboard : ok
Bit Spread : ok
Bit Flip : ok
Walking Ones : ok
Walking Zeroes : ok
8-bit Writes : ok
16-bit Writes : ok
Done.
==================================================
When 4 processes running simultaneously ,Soc temp is high to 67 degree.
==================================================
pi@Pi-5:/sys/class/gpio $ sensors
pwmfan-isa-0000
Adapter: ISA adapter
fan1: 4697 RPM
cpu_thermal-virtual-0
Adapter: Virtual device
temp1: +65.0°C
rpi_volt-isa-0000
Adapter: ISA adapter
in0: N/A
rp1_adc-isa-0000
Adapter: ISA adapter
in1: 1.49 V
in2: 2.52 V
in3: 1.40 V
in4: 1.42 V
temp1: +51.4°C
==================================================
I have overclock pi5 to 2.9G,without over volt.Maybe it's the reason of FAILURE?But my pi5 runs well for sevral week.
I may test more(eg. No overcloking,disable NUMA) to find the reason.
I enable NUMA follow the instruction of this topic,and got a high score with geekebnch.
I'm using a 8G pi5.Here is my configuration:
===========================================================
pi@Pi-5:~ $ vcgencmd version
2024年11月12日 16:10:44
Copyright (c) 2012 Broadcom
version 4b019946 (release) (embedded)
pi@Pi-5:~ $ vcgencmd bootloader_version
2024年11月12日 16:10:44
version 4b019946a06ea87c36d14fd6d702cc65fb458b9e (release)
timestamp 1731427844
update-time 1732995212
capabilities 0x0000007f
pi@Pi-5:~ $ vcgencmd bootloader_config
[all]
BOOT_UART=1
BOOT_ORDER=0xf461
NET_INSTALL_AT_POWER_ON=1
SDRAM_BANKLOW=1
pi@Pi-5:~ $ cat /proc/cmdline
reboot=w coherent_pool=1M 8250.nr_uarts=1 pci=pcie_bus_safe cgroup_disable=memory numa_policy=interleave numa=fake=8 system_heap.max_order=0 smsc95xx.macaddr=2C:CF:67:3C:6C:FD vc_mem.mem_base=0x3fc00000 vc_mem.mem_size=0x40000000 console=ttyAMA10,115200 console=tty1 root=PARTUUID=f580ed4f-02 rootfstype=ext4 fsck.repair=yes rootwait video=HDMI-A-1:1440x900M@60D plymouth.ignore-serial-consoles cfg80211.ieee80211_regdom=CN
pi@Pi-5:~ $
pi@Pi-5:/sys/class/gpio $ uname -a
Linux Pi-5 6.6.62+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.6.62-1+rpt1 (2024年11月25日) aarch64 GNU/Linux
===========================================================
To test memory stability ,I run 4 processes of memtester simultaneously . One of process report FAILURE at first run.
==================================================
pi@Pi-5:~ $ sudo memtester 1600M 2
memtester version 4.6.0 (64-bit)
Copyright (C) 2001-2020 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).
pagesize is 16384
pagesizemask is 0xffffffffffffc000
want 1600MB (1677721600 bytes)
got 1600MB (1677721600 bytes), trying mlock ...locked.
Loop 1/2:
Stuck Address : ok
Random Value : ok
Compare XOR : ok
Compare SUB : ok
Compare MUL : ok
Compare DIV : ok
Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : ok
Block Sequential : ok
Checkerboard : ok
Bit Spread : ok
Bit Flip : ok
Walking Ones : ok
Walking Zeroes : ok
8-bit Writes : |FAILURE: 0x57bd735cdc7eda8f != 0x57bd735ddc7eda8f at offset 0x000000002a229478.
16-bit Writes : ok
Loop 2/2:
Stuck Address : ok
Random Value : ok
Compare XOR : ok
Compare SUB : ok
Compare MUL : ok
Compare DIV : ok
Compare OR : ok
Compare AND : ok
Sequential Increment: ok
Solid Bits : ok
Block Sequential : ok
Checkerboard : ok
Bit Spread : ok
Bit Flip : ok
Walking Ones : ok
Walking Zeroes : ok
8-bit Writes : ok
16-bit Writes : ok
Done.
==================================================
When 4 processes running simultaneously ,Soc temp is high to 67 degree.
==================================================
pi@Pi-5:/sys/class/gpio $ sensors
pwmfan-isa-0000
Adapter: ISA adapter
fan1: 4697 RPM
cpu_thermal-virtual-0
Adapter: Virtual device
temp1: +65.0°C
rpi_volt-isa-0000
Adapter: ISA adapter
in0: N/A
rp1_adc-isa-0000
Adapter: ISA adapter
in1: 1.49 V
in2: 2.52 V
in3: 1.40 V
in4: 1.42 V
temp1: +51.4°C
==================================================
I have overclock pi5 to 2.9G,without over volt.Maybe it's the reason of FAILURE?But my pi5 runs well for sevral week.
I may test more(eg. No overcloking,disable NUMA) to find the reason.
- Attachments
-
- memtester.png
- memtester.png (45.36 KiB) Viewed 3476 times
Re: NUMA Testing. Is this way safe enough?
If memtester reports failure, then something is definitely broken. I agree it's likely the overclock went wrong.
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing. Is this way safe enough?
Yes, I'm sure the memtester failure is due to overclock and not NUMA.
Reduce the overclock, or add an over_voltage_delta setting until memtester passes.
Re: NUMA Testing
does NUMA against sched_ext, got system a little bit freeze.
Pi4/kernel-6.12/SDRAM_BANKLOW=3/scx_lavd.
Pi4/kernel-6.12/SDRAM_BANKLOW=3/scx_lavd.
- geerlingguy
- Posts: 585
- Joined: Sun Feb 15, 2015 3:43 am
Re: NUMA Testing
I got stable results for different benchmarks, testing on a couple different Pi 5s. I also tested overclocking at various frequencies, and saw a nice speed boost there too (taking back my single core record for Geekbench 6 of course, though I left some room for someone else to push it a bit further!):
https://www.jeffgeerling.com/blog/2024/ ... ram-tuning
https://www.jeffgeerling.com/blog/2024/ ... ram-tuning
The question is not whether something should be done on a Raspberry Pi, it is whether it can be done on a Raspberry Pi.
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing
I believe rpi-update is no longer needed. An apt full-upgrade should give you a numa supporting bootloader/firmware/kernel.geerlingguy wrote: ↑Mon Dec 02, 2024 7:02 pmhttps://www.jeffgeerling.com/blog/2024/ ... ram-tuning
Re: NUMA Testing
I getdom wrote: ↑Mon Dec 02, 2024 7:46 pmI believe rpi-update is no longer needed. An apt full-upgrade should give you a numa supporting bootloader/firmware/kernel.geerlingguy wrote: ↑Mon Dec 02, 2024 7:02 pmhttps://www.jeffgeerling.com/blog/2024/ ... ram-tuning
Code: Select all
pi@pi5crow:~$ dmesg | grep NUMA
[ 0.000000] NUMA: No NUMA configuration found
[ 0.000000] NUMA: NODE_DATA [mem 0x3f7fd2c0-0x3f7fffff]
[ 0.000000] NUMA: NODE_DATA [mem 0x7fffd2c0-0x7fffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0xbfffd2c0-0xbfffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0xffffd2c0-0xffffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0x13fffd2c0-0x13fffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0x17fffd2c0-0x17fffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0x1bfffd2c0-0x1bfffffff]
[ 0.000000] NUMA: NODE_DATA [mem 0x1ffb892c0-0x1ffb8bfff]
[ 0.000000] mempolicy: NUMA default policy overridden to 'interleave:0-7'
[ 1.886532] pci_bus 0000:00: Unknown NUMA node; performance will be reduced- geerlingguy
- Posts: 585
- Joined: Sun Feb 15, 2015 3:43 am
Re: NUMA Testing
Indeed! I've updated the blog post to note that.
The question is not whether something should be done on a Raspberry Pi, it is whether it can be done on a Raspberry Pi.
- geerlingguy
- Posts: 585
- Joined: Sun Feb 15, 2015 3:43 am
Re: NUMA Testing
Seeing your post, I also re-ran tinymembench, and found a similar regression—in my case memset going from 13+ GB/sec to 9.3 GB/sec. See my full test results here.Mikael wrote: ↑Wed Oct 23, 2024 8:03 pmRan some tests and things seem to be working well overall. Geekbench 6 is indeed faster than ever before. So is Google Octane v2. Passmark performs extremely good as well, except for the "Memory Write" sub test, which regresses from ~11500 MB/s to 8000 something. Sysbench memory read performance is also better than ever (very nice latency, both average and max), while the write test corroborates Passmark's result, showing a severe regression in write performance.
@dom - Here's how I'm running `tinymembench` for a reproducer:
Code: Select all
git clone https://github.com/rojaster/tinymembench.git
cd tinymembench && make
./tinymembench
The question is not whether something should be done on a Raspberry Pi, it is whether it can be done on a Raspberry Pi.
Re: NUMA Testing
Can numactl be used to change the memory allocation strategy program by program?geerlingguy wrote: ↑Mon Dec 02, 2024 9:33 pmSeeing your post, I also re-ran tinymembench, and found a similar regression—in my case memset going from 13+ GB/sec to 9.3 GB/sec. See my full test results here.Mikael wrote: ↑Wed Oct 23, 2024 8:03 pmRan some tests and things seem to be working well overall. Geekbench 6 is indeed faster than ever before. So is Google Octane v2. Passmark performs extremely good as well, except for the "Memory Write" sub test, which regresses from ~11500 MB/s to 8000 something. Sysbench memory read performance is also better than ever (very nice latency, both average and max), while the write test corroborates Passmark's result, showing a severe regression in write performance.
@dom - Here's how I'm running `tinymembench` for a reproducer:
The lower synthetic speeds don't result in other benchmarks running any slower though. It's just sad that not every number goes up and to the right :DCode: Select all
git clone https://github.com/rojaster/tinymembench.git cd tinymembench && make ./tinymembench
Re: NUMA Testing
I've got a Pi 5 8Gb and I've not done any testing for this so far so I haven't made any manual changes or installed anything for this other than keeping my Pi up to date. I'm running 12 Nov Bootloader.
When I run
dmesg | grep NUMA
I get:
[ 0.000000] NUMA: No NUMA configuration found
[ 0.000000] NUMA: Faking a node at [mem 0x0000000000000000-0x00000001ffffffff]
.
[ 0.000000] mempolicy: NUMA default policy overridden to 'interleave:0'
Does this mean it's working or do I need to do something else? I've tried searching for just 'interleave:0', but I could only find something where it says it would be 'interleave:0-7'?
Thanks.
When I run
dmesg | grep NUMA
I get:
[ 0.000000] NUMA: No NUMA configuration found
[ 0.000000] NUMA: Faking a node at [mem 0x0000000000000000-0x00000001ffffffff]
.
[ 0.000000] mempolicy: NUMA default policy overridden to 'interleave:0'
Does this mean it's working or do I need to do something else? I've tried searching for just 'interleave:0', but I could only find something where it says it would be 'interleave:0-7'?
Thanks.
Re: NUMA Testing
You still need to activate by adding a setting in the eeprom config. See e.g. Jeff Geerling's blog https://www.jeffgeerling.com/blog/2024/ ... ram-tuning
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing
Or even the first post of this thread!DirkS wrote: ↑Tue Dec 03, 2024 12:40 pmYou still need to activate by adding a setting in the eeprom config. See e.g. Jeff Geerling's blog https://www.jeffgeerling.com/blog/2024/ ... ram-tuning
Re: NUMA Testing
Apologies, I was a little confused what was needed now with the changes being incorporated into bootloaders etc. that was all.dom wrote: ↑Tue Dec 03, 2024 12:48 pmOr even the first post of this thread!DirkS wrote: ↑Tue Dec 03, 2024 12:40 pmYou still need to activate by adding a setting in the eeprom config. See e.g. Jeff Geerling's blog https://www.jeffgeerling.com/blog/2024/ ... ram-tuning
I'll add the line
Code: Select all
SDRAM_BANKLOW=1 Thanks for the fast replies.
- dom
- Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator - Posts: 8472
- Joined: Wed Aug 17, 2011 7:41 pm
Re: NUMA Testing
Yes. If it's not enabled on command line you can enable it for one process with:
Code: Select all
pi@pios:~ $ numactl --interleave=all head -1 /proc/self/numa_maps
5570550000 interleave:0-7 file=/usr/bin/head mapped=9 active=0 N0=2 N1=2 N2=1 N3=1 N4=1 N6=1 N7=1 kernelpagesize_kB=4
Code: Select all
pi@pios:~ $ numactl -l head -1 /proc/self/numa_maps
555f8b0000 local file=/usr/bin/head mapped=9 active=0 N0=2 N1=2 N2=1 N3=1 N4=1 N6=1 N7=1 kernelpagesize_kB=4
Re: NUMA Testing
I believe I'm part of the unlucky ones that are observing a regression here. Here's the before:
Here's the system description:
Now, let's perform a simple update to the kernel and EEPROM (without setting the SDRAM_BANKLOW):
So now we are in:
Running the benchmark again:
Still consistent. Now let's update the flag...
Reboot. Check for NUMA:
Run the benchmark:
Thoughts?
Code: Select all
sysbench memory --memory-block-size=1G --memory-total-size=20G --memory-oper=write run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)
Running the test with following options:
Number of threads: 1
Initializing random number generator from current time
Running memory speed test with the following options:
block size: 1048576KiB
total size: 20480MiB
operation: write
scope: global
Initializing worker threads...
Threads started!
Total operations: 20 ( 12.02 per second)
20480.00 MiB transferred (12305.51 MiB/sec)
General statistics:
total time: 1.6630s
total number of events: 20
Latency (ms):
min: 82.77
avg: 83.14
max: 83.63
95th percentile: 82.96
sum: 1662.71
Threads fairness:
events (avg/stddev): 20.0000/0.00
execution time (avg/stddev): 1.6627/0.00Code: Select all
> uname -a
Linux raspberrypi01 6.6.51+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.6.51-1+rpt3 (2024年10月08日) aarch64 GNU/Linux
> vcgencmd bootloader_version
2024年10月21日 15:27:49
version 951e1cc9d8b1d81c0ca1783a0634605616970bc3 (release)
timestamp 1729520869
update-time 1730230594
capabilities 0x0000007fCode: Select all
sudo apt update && sudo apt full-upgradeCode: Select all
> uname -a
Linux raspberrypi01 6.6.62+rpt-rpi-2712 #1 SMP PREEMPT Debian 1:6.6.62-1+rpt1 (2024年11月25日) aarch64 GNU/Linux
> vcgencmd bootloader_version
2024年11月12日 16:10:44
version 4b019946a06ea87c36d14fd6d702cc65fb458b9e (release)
timestamp 1731427844
update-time 1733529134
capabilities 0x0000007fCode: Select all
sysbench memory --memory-block-size=1G --memory-total-size=20G --memory-oper=write run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)
Running the test with following options:
Number of threads: 1
Initializing random number generator from current time
Running memory speed test with the following options:
block size: 1048576KiB
total size: 20480MiB
operation: write
scope: global
Initializing worker threads...
Threads started!
Total operations: 20 ( 12.03 per second)
20480.00 MiB transferred (12319.23 MiB/sec)
General statistics:
total time: 1.6611s
total number of events: 20
Latency (ms):
min: 82.39
avg: 83.04
max: 83.79
95th percentile: 82.96
sum: 1660.77
Threads fairness:
events (avg/stddev): 20.0000/0.00
execution time (avg/stddev): 1.6608/0.00Code: Select all
> sudo rpi-eeprom-config -e
SDRAM_BANKLOW=1
Code: Select all
> cat /proc/cmdline
... numa=fake=8 ...
Code: Select all
sysbench memory --memory-block-size=1G --memory-total-size=20G --memory-oper=write run
sysbench 1.0.20 (using system LuaJIT 2.1.0-beta3)
Running the test with following options:
Number of threads: 1
Initializing random number generator from current time
Running memory speed test with the following options:
block size: 1048576KiB
total size: 20480MiB
operation: write
scope: global
Initializing worker threads...
Threads started!
Total operations: 20 ( 8.37 per second)
20480.00 MiB transferred (8575.85 MiB/sec)
General statistics:
total time: 2.3869s
total number of events: 20
Latency (ms):
min: 118.76
avg: 119.33
max: 120.26
95th percentile: 118.92
sum: 2386.66
Threads fairness:
events (avg/stddev): 20.0000/0.00
execution time (avg/stddev): 2.3867/0.00
Jump to
- Community
- General discussion
- Announcements
- Other languages
- Deutsch
- Español
- Français
- Italiano
- Nederlands
- 日本語
- Polski
- Português
- Русский
- Türkçe
- User groups and events
- Raspberry Pi Official Magazine
- Using the Raspberry Pi
- Beginners
- Troubleshooting
- Advanced users
- Assistive technology and accessibility
- Education
- Picademy
- Teaching and learning resources
- Staffroom, classroom and projects
- Astro Pi
- Mathematica
- High Altitude Balloon
- Weather station
- Programming
- C/C++
- Java
- Python
- Scratch
- Other programming languages
- Windows 10 for IoT
- Wolfram Language
- Bare metal, Assembly language
- Graphics programming
- OpenGLES
- OpenVG
- OpenMAX
- General programming discussion
- Projects
- Networking and servers
- Automation, sensing and robotics
- Graphics, sound and multimedia
- Other projects
- Media centres
- Gaming
- AIY Projects
- Hardware and peripherals
- Camera board
- Compute Module
- Official Display
- HATs and other add-ons
- Device Tree
- Interfacing (DSI, CSI, I2C, etc.)
- Keyboard computers (400, 500, 500+)
- Raspberry Pi Pico
- General
- SDK
- MicroPython
- Other RP2040 boards
- Zephyr
- Rust
- AI Accelerator
- AI Camera - IMX500
- Hailo
- Software
- Raspberry Pi OS
- Raspberry Pi Connect
- Raspberry Pi Desktop for PC and Mac
- Beta testing
- Other
- Android
- Debian
- FreeBSD
- Gentoo
- Linux Kernel
- NetBSD
- openSUSE
- Plan 9
- Puppy
- Arch
- Pidora / Fedora
- RISCOS
- Ubuntu
- Ye Olde Pi Shoppe
- For sale
- Wanted
- Off topic
- Off topic discussion