Summary
The objectives of this exercise were to compare network performance and power supply effects of Raspberry Pi 4 computers activated using PoE (Power over Ethernet) and normal power supplies, then to consider options for remote control to run programs on the Pi when the only cable connected supplied power via an Ethernet cable.
The main Raspberry Pi used was fitted with a PoE HAT that has a fan and converts Ethernet supplied power to that required to operate the Pi. Additional hardware obtained was unit that injects power on to the cable and one that converts the power in place of the HAT. The main cables used were combinations of three 48 metres CAT 6 (30+10+8) and 50 metres (30+10+10), the latter being a thin one (CAT 5?).
LAN Tests - These comprised using my LanSpeed benchmark and Burnindrive program. The former measures writing an reading speeds of large files, random access and small files and the latter creates files using numerous data patterns that are checked for data comparison failures on reading. The tests were run using normal power supplies, then using PoE, in both cases using 48 and 50 metre cable combinations.
LAN Tests Normal Power - Except where the thin cable was used, large file data transfer speeds clearly demonstrated Gigabit performance, perhaps a little slower with the long cable. The thin cable lead to a reduction in speed to 100 Mbps. In all cases, as expected, performance of random access and small files was awful. No data comparison errors were recorded during the Burnindrive tests, but thin cable tests were much slower, again.
LAN Tests PoE - Without the thin cable, performance could be a little slower than with normal power supplies, and there were no data errors. Including the thin cable, there were all sorts of issues, failing to run properly or the program crashing.
PoE High Power Consumption CPU Stress Tests - I ran my floating point and integer CPU stress tests that were known to consume the most power, using eight threads. These were run for 10 minutes and my CPU MHz, voltage and temperature monitoring program was run at the same time. Four tests of each were run using normal power, 48m thick cable, 50m including thin cable, with PoE HAT, and 50m with the alternative power converter. With the fan in use, there were only moderate increases in temperature and no CPU MHz throttling, with performance essentially constant and the same on all tests.
PoE CPU Stress Tests With USB 3 Hard Drive and USB 3 Flash Drive - It became clear on running these tests, that careful consideration of USB connection and activity might be needed with PoE. Using the 50 metre cable combination, with the thin one, caused crashes of stress testing program, without any USB activity, and also with the long thick cable, when attempting to use the hard drive.
Tests were carried out using normal power without and with hard disk activity, then using PoE, via the 48 metre thick cable combination, without and with flash drive data transfers. At least, these tests were all successful, with 100 MB/second data transfers having little impact on CPU performance (with USB activity for less than half of the time).
One Wire PoE Terminal Remote Control - Fully functionable Raspberry Pi Terminals were enabled on remote hardware. The Pi only had a single wire connected, an Ethernet cable with PoE, but without the power injector unit connected to an Ethernet hub, forcing WiFi intercommunication. Different ssh connection methods were required, using Windows 7, Windows 10, Linux (Raspbian and Ubuntu) and an Android smart phone. Screenshots from these, displaying various benchmark results and monitoring details, are provided.
The alternative PoE power converter was used for the Ubuntu test, for connection to a fanless Pi board. Two terminals were opened to run a CPU stress test and monitor MHz and temperature, using one thread, then eight. As expected, the latter suffered from MHz throttling, and the former at constant speed with moderate increases in temperature. This was to demonstrate that the cheaper, fanless PoE option could be adequate for some applications.
One Wire VNC Viewer Remote Control - Appropriate VNC Servers and Viewer software was installed on two Raspberry Pi 4s, a Windows PC and an Android smart phone, with a Server Pi having the same one wire connection as above. VNC Viewer provides remote desktop displays and full functionality. Using this facility, I ran my Raspberry Pi OpenGL benchmark with selection and dynamic display on a Windows monitor and a CPU benchmark with activities controlled by a remote Raspberry Pi. Finally, the Android phone was connected to monitor my JavaDraw benchmark, already running on the Server Pi. Screenshots are provided. as before.
Conclusions - At least, for the particular equipment used here, Power over Ethernet can be used over long cables (48 metres in this case), with little communications performance degradation and providing adequate power. Operation from remote computers and smart phones, with access to the same local network, is easy to apply, with realtime observation of normal output displayed by programs. Failure free usage can depend on avoiding the use of lower quality cables and such as USB connections that impose high, even transient, power demands.
Go To Start
General
I have a Raspberry Pi 4 cooled by the fan in a PoE HAT, initially unaware what PoE was all about. The fan does a great job for cooling purposes, as I reported at ResearchGate in
Raspberry Pi 4B Stress Tests Including High Performance Linpack.pdf,
but is rather expensive for use as just a fan.
Details of the PoE HAT are available in
a raspberrypi.org publication.
This clarified (to me) that the power to operate the Pi can be supplied over an Ethernet cable, but needs to have remote power-sourcing equipment available. What was not made clear is that the Pi can be used at a distance, with power and communication suppled by a single wire, or maybe just power, if WiFi is working. PoE standards indicate a distance limit of 100 metres.
The power source provides a high voltage, like up to 50 volts, for efficient transmission over a distance, the PoE HAT converting this to the required 5 volts at 2.5 amps.
In my ignorance and looking for inexpensive hardware, I bought a DSLRKIT Gigabit Ethernet Active PoE Splitter, costing £9.90 from Amazon, but this only provides the HAT voltage conversion and Ethernet connection. Tests have been run (or attempted) including this, instead of an HAT, but such devices should not really be used unless approved by raspberrypi.org.
Then I bought the real thing, a TP-LINK TL-PoE150S PoE Injector Adapter, including 48 volts, 0.5 amps power supply. This supports Gigabit Ethernet with one port in and one out, costing £16.99 from Amazon (I suppose that this should also be officially approved).
Tests were run using short and long cables measuring data transmission speeds and trying to indicated voltage limitations using stress testing programs, then using remote control to check single wire operation.
LanSpeed Benchmark Next or Go To Start
LanSpeed Benchmark - Normal Power, Short Cable
Network speeds are measured using my LanSpeed benchmark. This measures writing and reading MB/second for large files, milliseconds random reading and writing time of small blocks for a range of file sizes, plus average writing and reading performance of 200 small files, in terms of MB/second and milliseconds per file.
Commands
Following is the parameter to mount the remote file path, where the public folder was pre-allocated, then the execution command. In this case, the remote system was a PC using Windows 7.
Note, M = 1048576 with the MB parameter for large files doubling for a second set. Maximum specified MB size is 1023 as 2048 produces a segmentation error, using Raspbian 32 bit Operating System, where maximum file size is 2 GB - 1 (parameter would be less than 1,023.9999995232).
sudo mount -t cifs -o dir_mode=0777,file_mode=0777 //192.168.1.68/d/ray /media/public
./LanSpeed MB 1000 FilePath /media/public
Log File
Following is a copy of the text log file, saved in the same folder as the benchmark execution file. This is for the Raspberry Pi based program transferring data to/from the Windows 7 based PC via a short cable from the main hub. The same performance details are displayed as the test is running.
Large file data transfer speeds clearly demonstrated Gigabit performance. Note the slow results for small files and increase with file size.
#####################################################
LanSpeed RasPi 1.0 Mon Apr 13 16:15:23 2020
Selected File Path:
/media/public/
Total MB 266240, Free MB 157082, Used MB 109158
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
1000 111.56 108.91 108.29 110.60 92.45 108.48
2000 110.73 113.53 113.81 99.21 104.85 102.32
Random Read Write
From MB 4 8 16 4 8 16
msecs 0.007 0.011 0.052 0.93 1.30 0.98
200 Files Write Read Delete
File KB 4 8 16 4 8 16 secs
MB/sec 1.16 2.33 4.14 1.28 2.26 5.03
ms/file 3.53 3.51 3.96 3.20 3.62 3.26 0.307
End of test Mon Apr 13 16:18:45 2020
Variations
1000 111.56 108.91 108.29 110.60 92.45 108.48
1000 112.53 110.51 110.89 103.18 101.20 110.89
1000 111.65 111.42 109.67 109.77 99.56 100.98
msecs 0.007 0.011 0.052 0.93 1.30 0.98
msecs 0.006 0.011 0.127 1.05 0.83 0.83
msecs 0.006 0.009 0.275 1.44 0.84 0.89
ms/file 3.53 3.51 3.96 3.20 3.62 3.26 0.307
ms/file 4.51 3.07 3.27 1.97 2.03 2.22 0.140
ms/file 3.59 3.83 4.16 2.80 2.93 3.03 0.186
Performance Monitor
The following are results from the sar -n network performance monitor. This can be used following installation of the sysstat package. The command executed was sar -n DEV 1 10, for 10 1 second periods and average, entered after the start of the write and read speed displays. The tx and rx kB/second recordings are essentially the same as those produced by the benchmark.
The main other measurements are for packets per second, with packet sizes around 1480 Bytes, return 66 Bytes.
eth0 rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil
Write 4258.10 78338.00 283.48 115821.38 0.00 0.00 0.00 94.88
Read 76744.60 33972.50 113341.35 2207.15 0.00 0.00 0.00 92.85
LanSpeed Long Cables Next or Go To Start
LanSpeed Benchmark - Normal Power, Long Cables
Cat 6
I have a 30 metre Cat 6 cable connected from the main hub, round the house, to another room, I extended this with two more cables to 48 metres. Examining the resultant log file, below, indicated that the only real change, from the short cable test, was on reading large files, more sample appearing to be slower. This is confirmed by the other two examples shown, and variations in the network performance results.
#####################################################
LanSpeed RasPi 1.0 Tue Apr 14 19:18:26 2020
Selected File Path:
/media/public/
Total MB 266240, Free MB 159280, Used MB 106960
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
1000 110.51 113.30 111.52 101.75 82.53 93.44
2000 114.09 113.78 113.21 91.33 86.31 89.50
Random Read Write
From MB 4 8 16 4 8 16
msecs 0.006 0.009 0.152 1.15 1.11 1.00
200 Files Write Read Delete
File KB 4 8 16 4 8 16 secs
MB/sec 1.24 2.34 4.42 1.55 3.34 5.22
ms/file 3.31 3.51 3.70 2.65 2.45 3.14 0.164
End of test Tue Apr 14 19:21:59 2020
1000 112.49 111.73 113.20 98.30 85.79 89.54
1000 110.25 113.55 111.59 102.73 94.26 97.00
eth0 rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil
Av Write 2974.60 78172.50 201.73 115578.00 0.00 0.00 0.00 94.68
Av Read 60009.60 16616.70 88632.31 1080.54 0.00 0.00 0.00 72.61
Sanple Read Per Second Range
19:19:04 80461.00 36840.00 118840.38 2387.14 0.00 0.00 0.00 97.35
19:19:09 48503.00 3765.00 71638.97 250.32 0.00 0.00 0.00 58.69
Thin Cable
I replaced the extension 8 metre cable with a 10 metre thin cable (Cat 5?) and tests repeated. As shown below, this lead to maximun network speed of 100 Mbps. A test was also run just using this thin cable, where resuklts were virtually identical, including the odd rxmcst/s, or number of multicast packets received per second. I have no idea why.
#####################################################
LanSpeed RasPi 1.0 Tue Apr 14 20:07:48 2020
Selected File Path:
/media/public/
Total MB 266240, Free MB 159282, Used MB 106958
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
100 11.55 11.51 11.50 11.26 11.68 11.64
200 11.68 11.63 11.66 11.67 11.72 11.72
Random Read Write
From MB 4 8 16 4 8 16
msecs 0.023 0.903 0.592 1.26 1.20 1.21
200 Files Write Read Delete
File KB 4 8 16 4 8 16 secs
MB/sec 1.05 1.94 3.29 1.08 1.93 3.23
ms/file 3.90 4.23 4.99 3.80 4.24 5.08 0.320
End of test Tue Apr 14 20:10:55 2020
eth0 rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil
Write 3231.50 7966.60 209.27 11776.81 0.00 0.00 0.20 96.48
Read 8045.70 3962.70 11889.11 258.28 0.00 0.00 0.00 97.40
Data Validation Long Cable Next or Go To Start
Data Validation Long Cable
The results reported here are from running burnindrive2 stress testing program. Burnindrive writes four files, using 164 blocks of 64 KB, repeated 16 times (164.0 MB), with each block containing a unique data pattern. The files are then read for MMM minutes, on a sort of random sequence, with data and file ID checked for correct values. Then each block (unique pattern) is read numerous times, over SSS seconds, again with checking for correct values. MMM and SSS are runtime parameters, along with the filepath. Further information, including data pattern hexadecimal values, can be found in my ResearchGate report
Raspberry Pi 4B Stress Tests Including High Performance Linpack.pdf.
The remote system was mounted as for LanSpeed, again to a PC using Windows 7. The run time command is shown below, followed by results, over 21 minutes.
The way that the program works does not provide high speed data transmission. In this case, around 53 MB/second writing and 30 MB/second reading.
Command - ./burnindrive2 Seconds 4 Minutes 10 FilePath /media/public Log 30
###############################################################
Linux Storage Stress Test for ARM v2.0, Tue Apr 14 23:24:01 2020
File size 164.00 MB x 4 files, minimum reading time 10.0 minutes
File 1 164.00 MB written in 3.17 seconds
File 2 164.00 MB written in 3.15 seconds
File 3 164.00 MB written in 3.07 seconds
File 4 164.00 MB written in 2.96 seconds
Total 12.35 seconds, Elapsed 12.35 seconds
Start Reading Tue Apr 14 23:24:14 2020
Read passes 1 x 4 Files x 164.00 MB in 0.41 minutes
Read passes 2 x 4 Files x 164.00 MB in 0.80 minutes
Read passes 3 x 4 Files x 164.00 MB in 1.15 minutes
Read passes 4 x 4 Files x 164.00 MB in 1.53 minutes
Read passes 5 x 4 Files x 164.00 MB in 1.89 minutes
Read passes 6 x 4 Files x 164.00 MB in 2.26 minutes
Read passes 7 x 4 Files x 164.00 MB in 2.61 minutes
Read passes 8 x 4 Files x 164.00 MB in 2.97 minutes
Read passes 9 x 4 Files x 164.00 MB in 3.32 minutes
Read passes 10 x 4 Files x 164.00 MB in 3.68 minutes
Read passes 11 x 4 Files x 164.00 MB in 4.03 minutes
Read passes 12 x 4 Files x 164.00 MB in 4.40 minutes
Read passes 13 x 4 Files x 164.00 MB in 4.75 minutes
Read passes 14 x 4 Files x 164.00 MB in 5.10 minutes
Read passes 15 x 4 Files x 164.00 MB in 5.46 minutes
Read passes 16 x 4 Files x 164.00 MB in 5.81 minutes
Read passes 17 x 4 Files x 164.00 MB in 6.18 minutes
Read passes 18 x 4 Files x 164.00 MB in 6.55 minutes
Read passes 19 x 4 Files x 164.00 MB in 6.90 minutes
Read passes 20 x 4 Files x 164.00 MB in 7.25 minutes
Read passes 21 x 4 Files x 164.00 MB in 7.60 minutes
Read passes 22 x 4 Files x 164.00 MB in 7.96 minutes
Read passes 23 x 4 Files x 164.00 MB in 8.31 minutes
Read passes 24 x 4 Files x 164.00 MB in 8.67 minutes
Read passes 25 x 4 Files x 164.00 MB in 9.03 minutes
Read passes 26 x 4 Files x 164.00 MB in 9.38 minutes
Read passes 27 x 4 Files x 164.00 MB in 9.75 minutes
Read passes 28 x 4 Files x 164.00 MB in 10.13 minutes
Start Repeat Read Tue Apr 14 23:34:22 2020
Passes in 4 second(s) for each of 164 blocks of 64KB:
2080 1920 1920 2080 2000 1940 2000 2020 1980 2000 2120
2160 1760 1960 1960 1940 2000 2000 1900 1880 1960 2100
1880 1920 2100 2040 1840 1960 1960 2060 2060 1900 2100
2080 2000 2080 2180 2020 2280 2240 2120 2160 2160 2000
2100 1920 2040 2020 2060 2180 2080 2000 2200 2060 1960
1980 2060 2040 1980 2080 1980 1820 1960 1820 2180 1840
2100 2280 1820 2200 1940 2100 2220 2000 2000 1780 1800
2100 1980 1860 2080 2080 1880 2000 1900 2060 1960 2100
2040 2120 1880 1940 2160 2160 2000 2060 2140 2180 1880
2140 2200 1980 2120 2140 1960 2020 2140 2060 2120 1920
2060 2160 1960 2160 2220 1980 1940 1920 1980 2080 1980
2040 2160 2140 1980 2120 2160 2180 2040 2080 2020 2020
2060 2120 2020 2100 2160 2100 2060 2120 2060 2080 1980
1920 2060 2080 2160 1960 2120 2160 2100 2160 2060 2200
2220 2260 2160 2180 2240 2120 2000 2220 2060 2200
335660 read passes of 64KB blocks in 10.99 minutes
No errors found during reading tests
End of test Tue Apr 14 23:45:21 2020
Data Validation Long Thin Cable Next or Go To Start
Data Validation Long Thin Cable
This Burnindrive test was run via the 50 metre cable combination, including the 10 metre thin one, using the same command provided on the previous page. The slow results, introduced by the thin cable, were again shown, but somewhat slower than before, due to the program overheads, at 9.7 MB/second writing and 8.8 For each of the reading sessions.
Following the logged program results is a summary of details provided from running Perfmon, performance monitor, on the Windows 7 based PC. Besides network traffic, the selected parameters included PC disk input and output. Network writing and reading speeds are indicates as somewhat faster than recorded by the program, but could include unrelated activity. Calculated packet sizes were around 1500 Bytes data transfer and 70 acknowledgement.
The Perfmon statistics included a number for packets under 12 headings covering Controller, Interface and Pseudo, Discarded and Errors, Outbound and Received. All were recorded as zero.
###############################################################
Linux Storage Stress Test for ARM v2.0, Thu Apr 16 14:27:14 2020
File size 164.00 MB x 4 files, minimum reading time 10.0 minutes
File 1 164.00 MB written in 16.93 seconds
File 2 164.00 MB written in 16.99 seconds
File 3 164.00 MB written in 16.96 seconds
File 4 164.00 MB written in 17.08 seconds
Total 67.96 seconds, Elapsed 67.96 seconds
Start Reading Thu Apr 16 14:28:22 2020
Read passes 1 x 4 Files x 164.00 MB in 1.26 minutes
Read passes 2 x 4 Files x 164.00 MB in 2.51 minutes
Read passes 3 x 4 Files x 164.00 MB in 3.75 minutes
Read passes 4 x 4 Files x 164.00 MB in 5.00 minutes
Read passes 5 x 4 Files x 164.00 MB in 6.24 minutes
Read passes 6 x 4 Files x 164.00 MB in 7.49 minutes
Read passes 7 x 4 Files x 164.00 MB in 8.74 minutes
Read passes 8 x 4 Files x 164.00 MB in 9.98 minutes
Read passes 9 x 4 Files x 164.00 MB in 11.23 minutes
Start Repeat Read Thu Apr 16 14:39:35 2020
Passes in 4 second(s) for each of 164 blocks of 64KB:
580 560 580 580 580 580 580 560 560 560 560
580 580 580 580 580 580 580 560 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 580 580 560 580 580 580 560 580 560 560
580 560 580 560 580 580 580 580 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 560 580 580 580 560 560 580 560 580 580
560 580 580 580 580 580 580 560 580 580 560
580 580 580 580 580 580 580 580 580 580 580
580 580 580 600 580 580 560 580 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 580 580 580 580 580 560 560 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 580 580 580 580 580 580 580 580 580
94700 read passes of 64KB blocks in 11.16 minutes
No errors found during reading tests
End of test Thu Apr 16 14:50:45 2020
Windows 7 Summary Perfmon Logging
Bytes Bytes Packets Packets Output Disk Rd Disk Disk Wr Disk %CPU
rec/sec sent/sec rec/sec sent/sec Q length Bytes/sec Rds/sec Bytes/sec wrts/sec Util
Write 10534034 214239 7055 3047 0 1948 0 8148232 13 10
Read1 244962 9622502 3368 6449 0 9219947 142 239715 4 8
Read2 240384 9703990 3391 6502 0 9277668 146 16760 4 8
LanSpeed PoE Power Next or Go To Start
LanSpeed Benchmark - PoE Power, Long Cables
The first results below are for a speed test using the 48 metres of long cables, with power supplied by the remote TP-LINK TL-PoE150S PoE Injector Adapter. Performance was essentially the same as that using the normal power supply, except writing large files appeared to be slower. Details of later tests are shown, identifying that the degraded performance was irregular. The sar network monitoring details were for the first test, also identifying slow writing.
#####################################################
LanSpeed RasPi 1.0 Fri Apr 17 11:49:50 2020
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
1000 89.03 90.89 88.84 112.83 103.04 108.71
2000 89.43 88.61 88.33 104.78 100.79 105.57
Random Read Write
From MB 4 8 16 4 8 16
msecs 0.007 0.008 0.102 0.94 0.84 0.81
200 Files Write Read Delete
File KB 4 8 16 4 8 16 secs
MB/sec 1.08 2.40 4.99 1.51 4.41 6.14
ms/file 3.81 3.41 3.28 2.71 1.86 2.67 0.296
End of test Fri Apr 17 11:53:30 2020
Later
1000 112.50 110.38 113.18 112.83 111.74 112.62
1000 112.42 112.16 109.86 109.71 95.00 103.44
eth0 rxpck/s txpck/s rxkB/s txkB/s rxcmp/s txcmp/s rxmcst/s %ifutil
Write 4135.70 62161.70 273.68 91792.31 0.00 0.00 0.00 75.20
Read 73410.80 32399.50 108454.45 2100.40 0.00 0.00 0.20 88.85
Thin Cable
The Power over Ethernet speed tests, including the thin cable, did not impress. The first one, using the 10 metre thin cable plus 40 meter thick cables, failed to read any long files. Then, the next run, combined with the 30 metre CAT 6 cable, produced unacceptable slow reading speeds. A third test, provided expected long file writing and reading performance, but was slower on most random and small file tests.
Note that these problems were not noted during earlier PoE tests, using just the 10 metre thin cable.
#####################################################
MB Write1 Write2 Write3 Read1 Read2 Read3
100A 11.53 11.51 11.53
100B. 11.52 11.51 11.52 3.40 4.32 3.45
All 10m thin + thick - 100A 40m, B 30m, next 8m
LanSpeed RasPi 1.0 Fri Apr 17 15:10:31 2020
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
100 11.56 11.57 11.53 11.63 11.63 11.66
200 11.68 11.65 11.67 11.65 11.69 11.66
Random Read Write
From MB 4 8 16 4 8 16
msecs 0.019 2.962 3.711 1.31 1.77 1.76
200 Files Write Read Delete
File KB 4 8 16 4 8 16 secs
MB/sec 0.73 1.91 3.23 1.23 1.06 2.41
ms/file 5.61 4.28 5.08 3.34 7.71 6.81 0.731
End of test Fri Apr 17 15:13:40 2020
PoE Data Validation Long Cable Next or Go To Start
PoE Data Validation Long Cable
This was slightly slower, on reading, than the run using normal power, reading four files less in 13.1 seconds, and more than 6500 blocks less in 11 minutes.
Windows Perfmon results are also provided, confirming that they were again slightly higher than program writing and reading speeds.
Unlike using normal power, when Perfmon was used, constant values of Family Controller\Packets Received Discarded were reported over the whole period, for all long cable tests.
###############################################################
Linux Storage Stress Test for ARM v2.0, Sun Apr 19 10:24:24 2020
File size 164.00 MB x 4 files, minimum reading time 10.0 minutes
File 1 164.00 MB written in 3.09 seconds
File 2 164.00 MB written in 3.17 seconds
File 3 164.00 MB written in 3.01 seconds
File 4 164.00 MB written in 3.00 seconds
Total 12.28 seconds, Elapsed 12.28 seconds
Start Reading Sun Apr 19 10:24:36 2020
Read passes 1 x 4 Files x 164.00 MB in 0.42 minutes
Read passes 2 x 4 Files x 164.00 MB in 0.77 minutes
Read passes 3 x 4 Files x 164.00 MB in 1.16 minutes
Read passes 4 x 4 Files x 164.00 MB in 1.52 minutes
Read passes 5 x 4 Files x 164.00 MB in 1.90 minutes
Read passes 6 x 4 Files x 164.00 MB in 2.26 minutes
Read passes 7 x 4 Files x 164.00 MB in 2.61 minutes
Read passes 8 x 4 Files x 164.00 MB in 2.97 minutes
Read passes 9 x 4 Files x 164.00 MB in 3.35 minutes
Read passes 10 x 4 Files x 164.00 MB in 3.72 minutes
Read passes 11 x 4 Files x 164.00 MB in 4.11 minutes
Read passes 12 x 4 Files x 164.00 MB in 4.48 minutes
Read passes 13 x 4 Files x 164.00 MB in 4.84 minutes
Read passes 14 x 4 Files x 164.00 MB in 5.19 minutes
Read passes 15 x 4 Files x 164.00 MB in 5.59 minutes
Read passes 16 x 4 Files x 164.00 MB in 5.97 minutes
Read passes 17 x 4 Files x 164.00 MB in 6.33 minutes
Read passes 18 x 4 Files x 164.00 MB in 6.71 minutes
Read passes 19 x 4 Files x 164.00 MB in 7.10 minutes
Read passes 20 x 4 Files x 164.00 MB in 7.47 minutes
Read passes 21 x 4 Files x 164.00 MB in 7.81 minutes
Read passes 22 x 4 Files x 164.00 MB in 8.17 minutes
Read passes 23 x 4 Files x 164.00 MB in 8.55 minutes
Read passes 24 x 4 Files x 164.00 MB in 8.92 minutes
Read passes 25 x 4 Files x 164.00 MB in 9.35 minutes
Read passes 26 x 4 Files x 164.00 MB in 9.75 minutes
Read passes 27 x 4 Files x 164.00 MB in 10.13 minutes
Start Repeat Read Sun Apr 19 10:34:44 2020
Passes in 4 second(s) for each of 164 blocks of 64KB:
2100 2040 1760 1960 1900 1740 1980 2060 2080 2140 1880
2040 1860 2100 2160 2060 2040 2020 2020 2000 1900 1840
2020 2060 2120 1940 1800 2160 2140 2080 2140 2000 1860
2160 1860 2020 2200 1860 2000 2000 1720 2300 2140 2120
2020 2040 1980 2000 1880 2160 1880 2080 1980 1720 2140
2200 1900 2280 2100 2100 2080 1800 1880 1780 2120 2140
1800 2060 1620 2040 1920 2000 1860 1580 1900 2140 2140
2060 1920 2120 2220 2160 1800 2000 1820 1780 2020 1940
2240 2020 2080 2000 1840 2080 1980 1860 1820 1840 2100
2020 1980 2100 1860 1860 1960 1960 1700 1940 1740 2120
2020 2200 2180 2240 1840 2180 2080 2160 2200 2160 2160
1780 2020 2100 1660 2060 1820 2020 1920 2100 2140 2140
1960 2020 2300 1960 2020 2060 2160 1980 2080 2160 1900
1900 2100 2140 2000 1980 2020 1980 2140 2160 1940 1920
1960 2100 2040 2160 2180 1980 1940 2180 2020 1820
329080 read passes of 64KB blocks in 11.00 minutes
No errors found during reading tests
End of test Sun Apr 19 10:45:44 2020
Windows 7 Summary Perfmon Logging
Bytes Bytes Packets Packets Output Disk Rd Disk Disk Wr Disk %CPU
rec/sec sent/sec rec/sec sent/sec Q length Bytes/sec Rds/sec Bytes/sec wrts/sec Util
Write 54933931 301095 36790 3540 0 0 0 17889800 29 12
Read1 410183 32125907 5383 21526 0 30665779 473 112951 4 13
Read2 415707 34214684 5415 22924 0 32660559 508 30468 4 12
PoE Data Validation Including Thin Cable Next or Go To Start
PoE Data Validation Including Thin Cable
Again these tests failed to run properly with some long cable combinations, writing far too slowly at total lengths of 50 and 40 metres, as shown below. Reading was found to be happening, but pitifully slowly. A successful run was achieved at 18 metres. Then, as apparent from the running times and Windows Perfmon results, this was slightly slower than the 50 metre test, using normal power supplies.
###############################################################
File size 164.00 MB x 4 files, minimum reading time 10.0 minutes
40m + 10 m thin
File 1 164.00 MB written in 51.04 seconds
File 2 164.00 MB written in 54.63 seconds
File 3 164.00 MB written in 54.96 seconds
File 4 164.00 MB written in 49.84 seconds
30m + 10 m thin
File 1 164.00 MB written in 32.96 seconds
File 2 164.00 MB written in 34.74 seconds
File 3 164.00 MB written in 32.21 seconds
File 4 164.00 MB written in 32.59 seconds
###############################################################
8m + 10m thin
Linux Storage Stress Test for ARM v2.0, Sun Apr 19 12:44:00 2020
File 1 164.00 MB written in 16.96 seconds
File 2 164.00 MB written in 16.96 seconds
File 3 164.00 MB written in 16.91 seconds
File 4 164.00 MB written in 16.99 seconds
Total 67.82 seconds, Elapsed 67.82 seconds
Start Reading Sun Apr 19 12:45:08 2020
Read passes 1 x 4 Files x 164.00 MB in 1.29 minutes
Read passes 2 x 4 Files x 164.00 MB in 2.55 minutes
Read passes 3 x 4 Files x 164.00 MB in 3.81 minutes
Read passes 4 x 4 Files x 164.00 MB in 5.06 minutes
Read passes 5 x 4 Files x 164.00 MB in 6.32 minutes
Read passes 6 x 4 Files x 164.00 MB in 7.58 minutes
Read passes 7 x 4 Files x 164.00 MB in 8.84 minutes
Read passes 8 x 4 Files x 164.00 MB in 10.10 minutes
Start Repeat Read Sun Apr 19 12:55:14 2020
Passes in 4 second(s) for each of 164 blocks of 64KB:
580 560 580 580 560 520 560 580 580 580 580
580 580 580 580 580 520 560 580 580 580 580
580 580 580 580 520 580 580 580 580 580 580
580 580 580 580 580 580 580 580 580 580 580
580 580 580 580 580 520 580 580 580 580 580
580 580 580 580 580 580 520 580 580 580 580
580 580 580 580 580 580 580 580 540 580 580
580 540 580 580 580 580 580 580 580 580 580
580 560 560 560 580 560 520 580 580 580 520
520 580 580 580 580 580 520 580 580 520 580
540 520 580 580 580 580 580 580 580 580 580
580 580 560 580 580 580 560 580 580 560 520
580 580 580 580 580 520 580 580 580 580 580
580 580 540 580 580 580 580 580 580 580 580
560 580 560 580 560 540 580 580 540 580
93820 read passes of 64KB blocks in 11.20 minutes
No errors found during reading tests
End of test Sun Apr 19 13:06:26 2020
Windows 7 Summary Perfmon Logging
Bytes Bytes Packets Packets Output Disk Rd Disk Disk Wr Disk %CPU
rec/sec sent/sec rec/sec sent/sec Q length Bytes/sec Rds/sec Bytes/sec wrts/sec Util
Write 10449522 213390 6999 3038 0 1427 0 9848739 18 9
Read1 238376 9524994 3354 6382 0 9090131 140 36065 4 8
Read2 239729 9577087 3374 6417 0 9157249 145 16583 3 8
Floating Point Voltage Check Next or Go To Start
PoE Floating Point Voltage Check
For a check that PoE voltage appeared to be adequate, CPU stress tests, that had been identified as those with the highest current demands, were run. For floating point it was MP-FPUStress using 8 threads, 1280 KB, 32 Operations Per Word. The RPiHeatMHzVolts2 program was run at the same time to record CPU MHz, Voltage and temperatures. These particular tests were run with an operational HAT fan, for steady state conditions.
The tests were run with power supply provided as indicated below, those with 50 metre cables included the 10 metres thin cable. All ran continuously with the CPU at 1500 MHz. Different CPU voltages were indicated, but these were also consistent. The last results shown are from where the power is obtained via the Ethernet Active PoE Splitter, not the PoE HAT.
Average and maximum recordings are shown at the end, effectively indicating the same performance and temperatures for all tests.
Normal Power PoE Thick 48m PoE 50m mixed PoE not HAT 50m
MHz 1500 1500 1500 1500
Volts 0.8472 0.8455 0.8490 0.8472
Seconds MFLOPS CPU PMIC MFLOPS CPU PMIC MFLOPS CPU PMIC MFLOPS CPU PMIC
°C °C °C °C °C °C °C °C
Start 20168 47 46 20123 49 47 20102 47 45 20254 49 47
10 20168 57 55 20213 59 55 20102 58 55 20254 58 55
20 20276 60 55 20216 61 57 20180 61 57 20258 61 57
30 20556 61 57 20605 62 59 20307 61 57 20245 61 58
40 20316 61 57 20181 61 59 20343 61 57 20556 61 58
50 20212 60 57 20276 62 59 20579 61 58 20198 61 58
60 20074 61 57 20190 62 59 20310 61 58 20291 62 58
70 20210 60 57 20157 60 59 20228 61 58 20280 60 58
80 20265 60 57 20616 62 59 20165 60 58 20212 60 58
90 20573 60 57 20297 61 59 20179 61 58 20321 60 58
100 20228 59 57 20342 62 59 20222 62 58 20543 60 58
110 20285 60 57 20238 61 59 20580 62 58 20291 60 58
120 20263 60 57 20240 61 59 20215 61 58 20234 60 58
130 20235 60 57 20549 61 59 20247 60 58 20222 61 57
140 20222 59 57 20272 62 59 20194 60 57 20216 61 58
150 20289 59 56 20173 62 59 20356 60 57 20279 59 57
160 20565 58 57 20114 61 58 20238 60 57 20191 61 58
170 20214 59 57 20303 60 58 20287 61 57 20571 60 58
180 20299 59 57 20581 62 59 20271 62 58 20339 60 58
190 20203 59 56 20124 61 59 20215 61 58 20279 59 58
200 20244 58 57 20218 60 59 20219 61 58 20295 59 57
210 20084 58 55 20160 61 58 20354 61 58 20116 60 57
220 20411 59 55 20484 61 59 20254 62 58 20351 60 56
230 20463 59 55 20399 62 59 20238 62 58 20395 60 58
240 20247 59 55 20249 61 59 20176 61 58 20555 60 57
250 20321 59 55 20184 61 59 20612 61 58 20229 60 57
260 20245 58 55 20270 62 59 19988 61 59 20140 60 58
270 20289 59 55 20595 61 59 20336 61 58 20298 60 57
280 20339 59 55 20285 61 58 20336 61 59 20118 59 57
290 20280 58 55 20142 60 59 20291 62 58 20194 59 57
300 20296 58 55 20251 61 59 20256 61 58 20620 60 57
310 20324 58 55 20298 62 59 20543 61 58 20079 60 57
320 20311 58 55 20273 60 59 20349 62 58 20270 60 57
330 20256 59 55 20126 61 59 20310 61 58 20146 59 58
340 20320 59 55 20078 61 59 20117 60 58 20185 60 58
350 20250 58 55 20264 61 59 20077 61 58 20172 60 57
360 20570 58 55 20252 61 59 20233 60 57 20085 59 57
370 20299 59 55 20598 61 59 20334 61 57 20549 59 58
380 20153 59 55 20326 61 59 20297 61 58 20391 59 58
390 20214 58 55 20125 61 59 20626 61 58 20196 60 57
400 20288 59 55 20276 61 59 20214 60 58 20157 60 57
410 20273 58 55 20564 61 59 20234 60 58 20270 60 57
420 20544 58 55 20174 62 59 20290 61 58 20269 60 58
430 20268 58 55 20243 61 59 20320 61 58 20319 59 57
440 20225 59 55 20178 61 57 20211 61 57 20559 60 58
450 20223 58 55 20233 61 59 20634 61 58 20262 60 58
460 20304 59 55 20587 61 59 20292 61 57 20219 60 58
470 20266 58 55 20165 61 59 20303 61 58 20222 60 58
480 20220 58 55 20297 61 59 20197 61 58 20275 59 58
490 20591 57 55 20191 61 58 20104 61 59 20319 60 58
500 20222 57 55 20173 61 58 20244 60 58 20575 60 58
510 20216 57 55 20618 60 59 20573 60 58 20248 60 58
520 20207 58 55 20167 61 59 20209 60 58 20136 59 58
530 20225 58 55 20247 61 59 20184 61 57 20238 60 58
540 20149 58 55 20230 61 59 20320 59 57 20289 60 58
550 20460 58 55 20477 62 59 20227 60 57 20325 60 57
560 20236 58 55 20362 60 59 20215 61 58 20212 59 57
570 20236 59 55 20166 61 59 20238 61 58 20556 60 58
580 20117 58 55 20089 61 59 20348 62 58 20243 60 58
590 20118 58 55 20092 60 59 20228 61 58 20310 59 57
600 20312 58 55 20611 61 59 20177 60 59 20112 59 57
Average 20283 58 56 20287 61 59 20279 61 58 20287 60 57
Maximum 20591 61 57 20618 62 59 20634 62 59 20620 62 58
Integer Test Voltage Check Next or Go To Start
PoE Integer Test Voltage Check
The integer stress test used was MP-IntStress, again with 8 threads, 1280 KB and power supply arrangements. CPU MHz and voltages were also the same as above. Temperatures were not as high as during the floating point tests.
Performance was little different, over all the runs. Temperatures were also the same, except the last test, where there were indications that the PoE HAT voltage conversion might lead to an increase.
Normal Power PoE Thick 48m PoE 50m mixed PoE not HAT 50m
MHz 1500 1500 1500 1500
Volts 0.8455 0.8455 0.8490 0.8472
Seconds MB/sec CPU PMIC MB/sec CPU PMIC MB/sec CPU PMIC MB/sec CPU PMIC
°C °C °C °C °C °C °C °C
Start 22231 48 46 21580 46 44 21840 46 45 22291 42 39
10 22201 55 51 21052 55 51 21659 54 51 22172 50 46
20 22089 56 51 21223 55 51 21466 54 51 22238 52 46
30 21580 56 51 21392 56 51 21567 54 51 21688 52 48
40 21093 55 51 22002 56 52 21433 54 51 22032 53 48
50 20678 56 52 21213 55 51 21253 55 51 22256 54 50
60 20990 56 52 21440 56 52 21345 55 51 22180 53 50
70 20372 56 52 20951 56 51 21719 55 52 22223 54 49
80 20708 56 52 21630 56 53 21525 56 52 22184 54 49
90 21311 56 52 21722 56 53 22230 55 53 22158 53 49
100 21075 56 52 21005 56 53 22233 55 52 22265 52 49
110 22049 57 52 20092 57 53 22011 57 53 22299 54 49
120 22012 56 52 21580 56 53 21495 57 52 22238 53 49
130 21663 56 52 21639 56 53 20992 55 52 22124 54 50
140 22161 56 52 21756 56 53 21149 56 53 22266 53 49
150 21469 56 52 20782 56 53 21538 55 52 22235 53 49
160 21122 56 52 21128 56 53 21992 57 53 22242 54 49
170 21865 55 51 21276 56 53 21764 56 53 22241 52 49
180 22135 56 51 20859 56 53 21543 56 53 20933 53 49
190 21366 55 51 20812 56 53 21131 56 52 20918 53 49
200 22267 55 51 22068 56 53 21433 56 52 20322 53 49
210 21514 56 52 21147 55 52 22185 56 52 21007 54 49
220 22156 55 51 21052 56 52 21751 55 53 21043 52 49
230 21583 55 51 20699 55 52 22256 57 52 20905 53 48
240 21204 56 51 19376 55 52 22338 56 52 21346 53 48
250 20525 56 51 21277 56 52 22268 56 52 21463 51 47
260 21098 54 51 21437 55 52 21460 56 52 22236 51 47
270 21218 54 50 20804 56 52 21759 56 52 22049 51 47
280 21418 54 50 21507 56 52 21830 57 53 22231 52 48
290 22362 54 51 21875 55 52 22349 56 53 22277 53 48
300 22218 54 51 21099 57 53 22390 56 53 22298 53 48
310 22224 54 51 20999 56 52 22415 55 52 22296 52 49
320 22266 55 51 21265 55 52 22024 57 53 22236 52 49
330 21657 56 51 21269 56 52 21657 56 52 20896 53 49
340 21941 55 51 21169 54 51 21313 57 53 21131 52 49
350 22015 55 52 20568 54 51 21578 57 53 20844 53 49
360 21029 56 52 21208 54 51 21592 56 52 21067 54 49
370 21154 55 51 21473 54 51 22199 56 53 20914 53 49
380 20550 55 52 21379 54 51 22335 56 53 21044 53 50
390 21051 55 52 20510 54 51 21289 55 52 20963 53 49
400 20925 55 52 20684 54 51 21662 56 51 20588 53 49
410 20097 56 52 21536 55 51 21790 55 51 20626 54 49
420 21885 54 52 21516 56 52 21630 54 51 22252 53 49
430 21645 56 52 21621 56 52 22345 54 51 22174 54 48
440 22238 56 52 21690 56 52 21987 54 51 22285 53 49
450 22058 55 52 21765 55 52 22390 54 51 22239 53 50
460 21830 55 52 21437 56 52 22374 54 51 22189 53 50
470 22183 56 52 22045 56 52 21409 54 51 21749 53 49
480 22242 56 51 20792 57 52 21063 55 51 21644 53 50
490 21881 56 52 20685 57 52 21405 54 51 21276 52 48
500 22377 55 52 21607 56 52 21724 55 51 21048 52 47
510 22270 57 52 20840 55 52 21704 54 52 20846 52 48
520 21065 54 51 20626 56 52 21967 56 52 20792 51 47
530 21099 54 50 20614 55 52 21210 57 52 20916 53 48
540 20931 54 51 21488 56 52 21832 56 52 21043 53 49
550 21271 54 51 20826 56 52 22413 56 52 20873 53 49
560 21321 55 51 19965 56 52 21199 56 52 22275 52 50
570 21206 55 51 21075 56 52 22195 56 52 22245 53 49
580 20819 55 51 20834 56 52 22342 56 52 22096 53 49
590 22059 54 51 21931 55 52 22093 56 52 22280 52 50
600 21598 55 51 21477 57 52 22344 55 52 22136 53 49
Average 21551 55 52 21186 55 52 21793 55 52 21694 53 48
Maximum 22377 57 52 22068 57 53 22415 57 53 22299 54 50
PoE Stress Tests With USB Drives Next or Go To Start
PoE Stress Tests With USB 3 Hard Drive and USB 3 Flash Drive
It became clear on running these tests, that careful consideration of USB connection and activity might be needed with PoE. Using the 50 metre cable combination, with the thin one, caused crashes of stress testing program, without any USB activity, and also with the long thick cable, when attempting to use the hard drive.
Following are details of successful tests, using normal power and PoE with the USB drives connected, then normal power plus a hard drive benchmark and PoE whilst benchmarking the USB 3 flash drive.
There was not a lot of difference in the results, including with the impact of USB data transfers transfers (indicated as #), in both cases averaging more than 100 MB/second for much of the time, with 500 and 1000 MB files.
Normal Power Normal Power +HD PoE 48m PoE 48m +USB 3 Sick
MHz 1500 1500 1500 1500
Volts 0.8490 0.8490 0.8490 0.8490
Seconds MFLOPS CPU PMIC MFLOPS CPU PMIC HD MFLOPS CPU PMIC MFLOPS CPU PMIC FD
°C °C °C °C °C °C °C °C
Start 20442 47 45 19968 46 45 19992 50 47 20154 47 47
10 20335 58 55 20283 57 54 20260 60 56 20279 57 55
20 20425 60 57 20112 60 56 20341 62 58 20346 61 57
30 20564 61 55 20542 62 57 20121 65 61 20151 64 59 #
40 20401 61 59 19310 61 58 # 20122 64 61 17594 64 59 #
50 20311 62 59 19105 61 58 # 20242 65 61 17387 63 59 #
60 20376 62 59 18475 61 58 # 20063 65 61 17330 64 61 #
70 20451 62 59 18859 63 58 # 18663 64 59 17735 64 61 #
80 20270 62 59 18989 61 59 # 20178 63 61 16920 64 61 #
90 20346 62 59 18882 61 58 # 20229 63 61 16996 64 61 #
100 20649 63 59 19095 62 58 # 20193 63 61 17036 63 59 #
110 20351 63 59 18663 61 58 # 14925 62 58 17150 64 59 #
120 20365 62 59 18515 61 58 # 19801 62 59 16785 64 61 #
130 20285 62 59 19395 60 58 # 20183 63 60 17525 64 61 #
140 20337 62 59 19934 62 59 # 20334 63 59 17489 64 61 #
150 20295 61 59 20250 61 58 # 19297 63 61 18130 64 61 #
160 20410 61 59 20442 62 59 20148 64 61 20253 65 62 #
170 20546 62 59 20467 61 59 20151 64 61 19864 64 62 #
180 20345 60 59 20162 61 59 20174 64 61 19843 65 62 #
190 20307 61 59 20171 63 59 20239 64 61 19936 65 63 #
200 20325 62 59 20081 62 59 20113 64 61 19971 66 63 #
210 20221 63 59 19024 63 59 # 20277 64 62 20009 66 63 #
220 20388 62 59 18710 62 59 # 20431 63 62 19758 65 63 #
230 20601 61 59 18622 62 59 # 20186 63 62 19881 66 63 #
240 20344 62 59 19118 61 58 # 20106 64 62 20402 65 63
250 20415 62 59 18755 61 59 # 20558 63 62 20229 66 63
260 20346 62 59 18576 61 59 # 20102 64 61 20335 66 63
270 20332 62 59 18533 63 59 # 19574 64 62 20333 66 63
280 20317 62 59 17994 62 59 # 19983 65 62 20343 66 63
290 20526 60 59 18237 62 59 # 19715 65 62 20230 66 63
300 20467 62 59 19920 61 59 # 20006 65 62 20223 66 63
310 20390 62 59 20280 63 59 19849 65 62 20536 66 63
320 20321 61 59 20222 62 59 19904 65 62 20387 66 63
330 20430 61 59 20255 61 59 20018 65 62 20286 66 63
340 20406 62 59 20091 61 59 19669 65 62 20265 65 63
350 20273 62 59 20596 61 59 20060 64 62 20306 65 63
360 20464 62 59 20282 61 59 20078 64 62 20292 65 63
370 20546 62 59 20264 61 59 20042 64 62 20169 66 63
380 20457 61 59 20102 62 59 19940 65 62 20198 65 63
390 20280 61 59 20147 62 59 19972 65 62 19930 65 63
400 20371 62 59 20640 61 59 20037 65 62 19219 65 63
410 19258 62 59 20187 61 59 20055 64 62 20027 65 63
420 19449 62 59 20153 62 59 20067 64 62 20023 65 63
430 20234 61 57 20224 61 59 19997 64 62 20203 66 63
440 20334 62 59 20250 60 59 20131 64 62 19635 66 63
450 20007 62 59 20298 62 59 20338 64 62 20225 66 63
460 20189 62 59 20214 62 59 19861 65 62 20337 66 63
470 20296 62 59 20115 62 59 19811 65 62 20554 66 63
480 20242 61 59 20295 61 59 20178 64 62 20331 66 63
490 20174 62 59 20298 62 59 20210 66 62 20232 66 63
500 20227 61 59 20454 62 58 20221 64 62 20311 66 63
510 20586 62 59 20206 62 59 20230 65 62 20101 66 63
520 20341 62 59 20240 62 59 20576 65 62 20236 66 63
530 20276 63 59 20268 61 59 20130 64 62 20299 65 63
540 20191 62 59 20637 61 59 20162 65 62 20651 65 63
550 20298 61 59 20156 61 59 20225 65 62 20327 65 63
560 20224 63 59 20244 62 59 20223 65 62 20247 66 63
570 20485 61 59 20273 61 59 19534 64 62 20210 66 63
580 20415 62 59 20196 62 59 19891 64 63 20336 66 63
590 20223 62 59 20420 61 59 19249 64 62 20214 67 63
600 20198 61 59 20479 61 59 20248 64 62 20443 66 63
Average 20323 61 59 19798 61 58 19961 64 61 19625 65 62
Maximum 20649 63 59 20640 63 59 20576 66 63 20651 67 63
PoE PuTTy Remote Control Windows 7 Next or Go To Start
Terminal Remote Control - One Wire PoE WiFi Only - Windows 7 and PuTTy
These tests were carried out with only the Ethernet 48 metre cable combination connected to a Raspberry Pi 4 with a PoE HAT, but without the cable being plugged in to the main hub. These were from a PC using Windows 7 with WiFi communication as the only available Pi connection. The first was from opening the Pi network folder on the PC and clicking on LanSpdx86Win.exe, saved on the Pi. The second one was after entering the Pi’s IP address in the PuTTy application and running the Pi LanSpeed benchmark using the fully functional Terminal.
Windows Benchmark From Raspberry Pi 4 - On opening Widow Network access to my remote Raspberry Pi, and clicking on LanSpdx86Win.exe, a Command Prompt window appears on the PC. The benchmark then runs, displaying results as they are calculated. Following are details obtained via Print Screen.
Note that performance can vary in all test functions, across the different platforms used. Connecting to my main hub, 5 GHz WiFi from a smartphone was indicated as running at 131 Mbps, with another, at 2.4 GHz, achieving 72 Mbps. These would be around 14.4 and 8.0 MBytes/second, respectively. Here, WiFi measured speeds of large files are all less than 10 MB/second.
Raspberry Pi Benchmark From PuTTy on Windows 7 - The next measurements are from using PuTTy, on a Windows 7 based PC, that requires the Pi’s IP address to be entered. This provides a Linux Terminal with all the usual copying and earlier commands to be used again. In this case, commands selected were cd to the appropriate folder, mounting the PC destination folder and execution of the program. Again, Print Screen was used to produce the following details.
Note, performance of this benchmark, running on the Pi, is somewhat slower than than from the Windows program, on all test functions.
PoE Terminal Remote Control Windows 10 Next or Go To Start
Terminal Remote Control - One Wire PoE Windows 10 & Phone
In order to carry out these tests, the longer cable combination could not be used, power being supplied via an 8 metre CAT 6 Ethernet cable, again with no data connection, to force the use of WiFi. In order to execute the benchmark on the Pi, Windows PoweShell had to be used, along with an ssh command (ssh pi@192.168.1.????). Then the usual set up and run commands could then be selected to produce the following results, copied using Print Screen, again.
Bearing in mind that is the same benchmark program run on the Pi, from Windows 7, performance writing and reading large files was faster, but other test functions essentially the same.
Phone Screen Shot - An Android phone was used for monitoring purposes, after installing the Termius app. Setup for this requires the same ssh details indicated above. In this case, the sar -n DEV network monitoring tool was used to confirm WiFi transmission speeds. Below is a copy of a screenshot from the phone, for one sample covering twelve seconds. Details from The LanSpeed log follow, starting four seconds before sar. During this period, a total of around 8 MB/second was recorded by sar, essentially the same as that shown in the benchmark log.
LanSpeed RasPi 1.0 Sun Apr 26 20:19:36 2020
MBytes/Second
MB Write1 Write2 Write3 Read1 Read2 Read3
8 8.06 7.81 8.17 8.35 8.09 7.39
16 7.10 8.12 8.51 7.85 7.42 7.59
Non-HAT PoE Terminal Remote Control Linux Next or Go To Start
Terminal Remote Control - One Wire PoE No HAT or Fan, Linux
These tests again included power over the 8 metres cable, but this was supplied to the Pi from the Ethernet Active PoE Splitter, through the normal Pi board supply socket. No cooling fan was connected. This time, the client system was a PC running Linux Abuntu, connecting via two Terminals using an ssh pi@192.168.1.???? command.
One terminal was used to run and display the Floating Point stress test, using one then 8 threads, the other terminal running the program that measures CPU MHz, volts and temperatures. Below are copies or extractions obtained using Print Screen.
Using a single thread, CPU MHz and recorded MFLOPS were effectively constant over the 10 minutes run, with only moderate increases in temperature. The 8 thread test lead to temperature increases that gave rise to CPU MHz throttling, with corresponding reductions in measured MFLOPS. These tests suggest that fanless one wire working could be possible, to provide constant performance, with light the CPU loading.
Terminal Remote Control To Pi64 Next or Go To Start
Terminal Remote Control One Wire PoE to Pi64 From Pi32
This demonstrates, using ssh from a Raspberry Pi using Raspbian (pi@raspberrypi), to connect to a different Pi with 64 bit Gentoo (demouser@pi64), running a 64 bit program there and displaying the results locally./
PoE VNC Viewer Windows Next or Go To Start
VNC Viewer One Wire PoE Windows From Pi
VNC Viewer was installed on a Windows based PC to display a remote desktop that would have been produced on a Raspberry Pi. Initially, the desktop was not displayed under Windows. As reported, when no display is connected, a change is needed in raspi-config, Advanced Options, Resolution, where one has to be selected from the list, instead of the Default setting.
VNC Viewer and Server were installed on the Raspberry Pi. The resolution chosen was deliberately smaller than possible, to allow more visible activities on the Windows screen.
To Start the activity, logging on is via VNC Connect, in the bottom left hand corner below. This then displays the remote desktop window. In this case, a Terminal was opened (bottom right), where normal Linux curser operations retrieved paths and execution statements. The latter were arranged to execute my OpenGL benchmark, rotating kitchen test function and run it for a while. When running, time and performance details are displayed in the test window title bar and on the Terminal. A clearer copy of part of the Terminal results is shown below. The main display screenshot was captured using PrtScr.
PoE VNC Viewer Other Pi Next or Go To Start
VNC Viewer One Wire PoE Pi From Other Pi
The VNC applications were installed on a second Raspberry Pi, to check that remote desktop could be accessed. In this case, I ran one of my benchmarks that measures floating point performance, using an increasing number of threads. The local and enclosed remote Raspberry Pi desktops are show below and a clearer view of the Terminal contents below.
VNC Viewer Android Mobile Phone Next or Go To Start
Android Mobile Phone VNC Viewer
This exercise did not involve PoE but was used to demonstrate the use of VNC Viewer on an Android phone. Following are sections of screenshots showing the installed app icon, VNC Connection facility and a display of my JavaDraw benchmark, running on the Raspberry Pi (well part of it). In this case, it was more of a spying activity, as the benchmark was started on the Pi. This must raise data protection issues.
Below is a copy of the JavaDraw log file saved on the Raspberry Pi.
Java Drawing Benchmark, May 1 2020, 13:42:12
Produced by javac 1.7.0_02
Test Frames FPS
Display PNG Bitmap Twice Pass 1 638 63.76
Display PNG Bitmap Twice Pass 2 847 84.67
Plus 2 SweepGradient Circles 826 82.58
Plus 200 Random Small Circles 549 54.86
Plus 320 Long Lines 234 23.34
Plus 4000 Random Small Circles 63 6.28
Total Elapsed Time 60.1 seconds
Operating System Linux, Arch. arm, Version 4.19
Java Vendor Raspbian, Version 11.0.5
Roy Longbottom May 2020
1960 to 2020 - Celebrating 60 years involvement in
computer reliability, performance and stress testing
Go To Start
|