pi-rat Posted December 18, 2019 Posted December 18, 2019 Hello, What kind of speed can I expect from the ethernet on the BPM1? It's told to be 1Gb/s, but I am not even getting near that speed: wget https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.4.5.tar.xz ... linux-5.4.5.tar.xz 100%[==============================================================================================>] 104,37M 4,43MB/s in 24s 2019-12-18 22:38:33 (4,35 MB/s) - »linux-5.4.5.tar.xz« gespeichert [109435988/109435988] BPM1 and the router are set to 1000Mb/s, and the download line works at more then 100Mb/s. Downloading the same file on my PC shows a considerably difference: linux-5.4.5.tar.xz 100%[==============================================================================================>] 104,37M 15,4MB/s in 6,9s 2019-12-18 22:43:45 (15,2 MB/s) - »linux-5.4.5.tar.xz« gespeichert [109435988/109435988] Is 4MB/s an average speed for the BPM1?
Igor Posted December 19, 2019 Posted December 19, 2019 8 hours ago, pi-rat said: It's told to be 1Gb/s Use search above. We discusses about M1 Bananas years ago and since hardware hasn't changed, numbers are the same. Theoretical speed is close to 1GB (use iperf to measure, not a download from external servers) but utilisation is something else. Also on protocol level. Even its an old horse, its still better than RPi's 3 "gigabit": https://www.jeffgeerling.com/blogs/jeff-geerling/getting-gigabit-networking
vlad59 Posted December 19, 2019 Posted December 19, 2019 Nope the Bpi M1 can do much more, see here : Your test file is not so quick to download : root@minus:~# curl -o /dev/null https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.4.5.tar.xz % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 100 104M 100 104M 0 0 13.0M 0 0:00:07 0:00:07 --:--:-- 13.4M root@minus:~# curl -o /dev/null http://test-debit.free.fr/1048576.rnd % Total % Received % Xferd Average Speed Time Time Time Current Dload Upload Total Spent Left Speed 32 1024M 32 332M 0 0 93.6M 0 0:00:10 0:00:03 0:00:07 93.6M^C The slowing factor is often the disk.
vlad59 Posted December 19, 2019 Posted December 19, 2019 I also did the test with wget (on a SSD) : wget https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.4.5.tar.xz --2019-12-19 09:17:14-- https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.4.5.tar.xz Resolving cdn.kernel.org (cdn.kernel.org)... 2a04:4e42:1d::432, 151.101.121.176 Connecting to cdn.kernel.org (cdn.kernel.org)|2a04:4e42:1d::432|:443... connected. HTTP request sent, awaiting response... 200 OK Length: 109435988 (104M) [application/x-xz] Saving to: ‘linux-5.4.5.tar.xz’ linux-5.4.5.tar.xz 100%[=========================================================================>] 104,37M 15,1MB/s in 6,9s 2019-12-19 09:17:21 (15,1 MB/s) - ‘linux-5.4.5.tar.xz’ saved [109435988/109435988]
pi-rat Posted December 19, 2019 Author Posted December 19, 2019 This is strange. Even if redirect the download to the great void with "wget ... -O /dev/null", making sure that no disk is used, the download speed does not exceed 4MB/s, while my PC downloads exactly the same file with 15MB/s. I tried changing the cables, I tried changing IP, I tried changing the router, but all to no avail. I have no explanation for this behaviour.
Igor Posted December 19, 2019 Posted December 19, 2019 29 minutes ago, pi-rat said: I have no explanation for this behaviour. Have you update u-boot loader? It's manual procedure -> armbian-config -> system -> install ... / update boot loader. Manufacturing quality can also be a problem on those cheap boards ... but first start testing network throughput with a correct tool: iperf Then proceed up.
pi-rat Posted December 19, 2019 Author Posted December 19, 2019 22 minutes ago, Igor said: armbian-config -> system -> install ... / update boot loader. The only "install" I see in menu "system" is "Install Desktop". If I select that one, it starts to install the full desktop package.
Igor Posted December 19, 2019 Posted December 19, 2019 It's first option under the system, then select update boot loader.
pi-rat Posted December 19, 2019 Author Posted December 19, 2019 The first submenu in "System" is "Freeze".
martinayotte Posted December 20, 2019 Posted December 20, 2019 The message is quite clear ! You must run "nand-sata-install" from SDCard, meaning that SDCard is actually the boot/root device...
pi-rat Posted December 20, 2019 Author Posted December 20, 2019 I know what this message means. But the system has already been installed to SSD and is running from SSD right now. So, if I understand you correctly, in order to update the boot-loader, I'll have to shut down the system, remove the SSD, boot from SDCard, and then I am to call the command "nand-sata-install" without any SSD attached. Wouldn't that try start the hole set up process again and wouldn't that fail because no SSD has been attached? Also, I don't quite understand, how updating the boot-loader might possible increase the max ethernet speed. This is the file system right now: Filesystem Size Used Avail Use% Mounted on udev 417M 0 417M 0% /dev tmpfs 97M 6,9M 90M 8% /run /dev/sda1 108G 27G 76G 27% / tmpfs 485M 0 485M 0% /dev/shm tmpfs 5,0M 8,0K 5,0M 1% /run/lock tmpfs 485M 0 485M 0% /sys/fs/cgroup tmpfs 485M 48K 485M 1% /tmp /dev/mmcblk0p1 7,2G 2,3G 4,8G 33% /media/mmcboot /dev/zram0 49M 22M 24M 49% /var/log tmpfs 97M 4,0K 97M 1% /run/user/123 tmpfs 97M 0 97M 0% /run/user/1000 And this is /boot: drwxr-xr-x 4 root root 4096 Dez 19 23:30 ./ drwxr-xr-x 22 root root 4096 Sep 17 17:26 ../ -rw-r--r-- 1 root root 0 Dez 18 22:06 .next -rw-r--r-- 1 root root 3276146 Nov 18 18:39 System.map-4.19.84-sunxi -rw-r--r-- 1 root root 202 Dez 19 23:30 armbianEnv.txt -rw-r--r-- 1 root root 0 Dez 19 21:45 armbianEnv.txt.out -rw-r--r-- 1 root root 1536 Jul 5 19:51 armbian_first_run.txt.template -rw-r--r-- 1 root root 4882 Jul 5 19:51 boot-desktop.png -rw-r--r-- 1 root root 230454 Jul 5 19:51 boot.bmp -rw-r--r-- 1 root root 3726 Dez 18 22:05 boot.cmd -rw-rw-r-- 1 root root 3798 Dez 18 22:05 boot.scr -rw-r--r-- 1 root root 163246 Nov 18 18:39 config-4.19.84-sunxi lrwxrwxrwx 1 root root 17 Dez 18 22:06 dtb -> dtb-4.19.84-sunxi/ drwxr-xr-x 3 root root 12288 Dez 18 22:05 dtb-4.19.84-sunxi/ lrwxrwxrwx 1 root root 17 Jul 5 19:50 dtb.old -> dtb-4.19.57-sunxi -rw-r--r-- 1 root root 6843382 Dez 19 17:19 initrd.img-4.19.84-sunxi drwxrwxr-x 2 root root 4096 Jul 5 19:51 overlay-user/ lrwxrwxrwx 1 root root 21 Dez 19 17:19 uInitrd -> uInitrd-4.19.84-sunxi -rw-r--r-- 1 root root 6843446 Dez 19 17:19 uInitrd-4.19.84-sunxi -rwxr-xr-x 1 root root 7392848 Nov 18 18:39 vmlinuz-4.19.84-sunxi* lrwxrwxrwx 1 root root 21 Dez 18 22:06 zImage -> vmlinuz-4.19.84-sunxi* ~$ uname -a Linux pi 4.19.84-sunxi #19.11.3 SMP Mon Nov 18 18:39:42 CET 2019 armv7l GNU/Linux As you can see, the system is running the latest kernel that was installed by armbian-config yesterday. Is it really still necessary to update the boot-loader? I am totally confused now, I really can't see any connection between the boot loader and the slow eth speed.
Igor Posted December 21, 2019 Posted December 21, 2019 20 hours ago, pi-rat said: I really can't see any connection between the boot loader and the slow eth speed. I know it is (do forum search for your education purposes), but I don't know what do you want? Your Banana perform as expected! RAW network speed is measured with iperf3 tool. Everything else includes protocol overhead, encryption, compression, ... where M1 will lost a lot since it has slow CPU. On 12/19/2019 at 10:50 AM, pi-rat said: It's told to be 1Gb/s Adjust your perception with reality and not the other way around Not with information which is presented by hardware manufacturer, sellers and amateur bench-markers.
pi-rat Posted December 21, 2019 Author Posted December 21, 2019 14 minutes ago, Igor said: ... but I don't know what do you want? That makes two of us. I was just trying to follow your advice: when I told of the slow download speed and the big difference compared to my PC, you ask me if I had updated the boot loader, and when that did not work, you ask me to run the SSD installation script. So, what are we talking about here?
Igor Posted December 21, 2019 Posted December 21, 2019 3 minutes ago, pi-rat said: when I told of the slow download speed and the big difference compared to my PC I can only give you few hints while fast scanning what you are saying. Updating everything is most universal hints which solves most of the problems most of the time. Since u-boot update procedure has been moved out from the automatic ... it has to be done manually and if script doesn't work you have to go to manual procedures which are written somewhere in the code. But it's not u-boot loader. You expected this board will match your PC speed. No it will not. Not even close, but you still can use it for many things. That's it.
Igor Posted December 21, 2019 Posted December 21, 2019 Your Banana speed: (Cubietruck has 2Gb of memory, the rest is the same) vs RK3399 on NvME (which is on/close to desktop PC) https://www.armbian.com/nanopc-t4/
pi-rat Posted December 21, 2019 Author Posted December 21, 2019 I know I can't compare the BP to my PC, and I don't want to. I am just wondering about this great difference when downloading the same file from the same server. As far as I know, downloading does not really require much CPU power, and if you redirect the data stream to /dev/null, it is also independent from any block device. I would understand it, if the difference was marginal or if the file came from two different servers or via two different routers or any thing like that. I agree to you, that most probably the reason for this difference in download speed has to be found outside the BP, and that's what I am trying to find out. But, like I told you, I switched the router, I switched the cables, but the difference always stays the same. And that has simply aroused my curiosity.
Igor Posted December 21, 2019 Posted December 21, 2019 2 minutes ago, pi-rat said: I am just wondering about this great difference when downloading the same file from the same server Measure speed between M1 any your PC. Run iperf3 -s (server) on your PC (there are clients for windows, mac, linux) and connect with iperf3 -c your_PC_IP ... then change client and server and report numbers.
pi-rat Posted December 22, 2019 Author Posted December 22, 2019 Here are my testing results. 1: "iperf -s" was run on the router: a) from PC to router: root@OpenWrt:~# iperf -B 1.1.1.1 -s ------------------------------------------------------------ Server listening on TCP port 5001 Binding to local address 1.1.1.1 TCP window size: 85.3 KByte (default) ------------------------------------------------------------ [ 4] local 1.1.1.1 port 5001 connected with 1.1.1.2 port 51174 [ ID] Interval Transfer Bandwidth [ 4] 0.0-10.0 sec 1.00 GBytes 859 Mbits/sec b) from BP to router: root@OpenWrt:~# iperf -B 1.1.2.1 -s ------------------------------------------------------------ Server listening on TCP port 5001 Binding to local address 1.1.2.1 TCP window size: 85.3 KByte (default) ------------------------------------------------------------ [ 4] local 1.1.2.1 port 5001 connected with 1.1.2.2 port 43496 [ ID] Interval Transfer Bandwidth [ 4] 0.0-10.0 sec 622 MBytes 522 Mbits/sec 2: "iperf -s" was run on the BP and on PC: a) from router to PC: root@pc:~# iperf -s ------------------------------------------------------------ Server listening on TCP port 5001 TCP window size: 128 KByte (default) ------------------------------------------------------------ [ 4] local 1.1.1.2 port 5001 connected with 1.1.1.1 port 60463 [ ID] Interval Transfer Bandwidth [ 4] 0.0-10.0 sec 1.07 GBytes 917 Mbits/sec b) from router to BP: root@bpm:~# iperf -s ------------------------------------------------------------ Server listening on TCP port 5001 TCP window size: 85.3 KByte (default) ------------------------------------------------------------ [ 4] local 1.1.2.2 port 5001 connected with 1.1.2.1 port 51657 [ ID] Interval Transfer Bandwidth [ 4] 0.0-10.0 sec 725 MBytes 607 Mbits/sec The names have been changed to protect the innocent... ;) Unfortunately, iperf3 is not available for OpenWRT. But I hope the differences between the two version will not be too big to spoil the results.
Igor Posted December 22, 2019 Posted December 22, 2019 Everything is normal. I would not bother with a bit lower speed on one side ... Orangepi Win (64bit/2g/faster), since its by accident on my desk: Iperf to server shows 937 Mbits/sec, other direction 850 Mbits/sec 2G file: wget from local web server 37.5MB/s (SD card) wget from local web server 81.5MB/s (memory) wget https://cdn.kernel.org/pub/linux/kernel/v5.x/linux-5.4.5.tar.xz 54.7MB/s (memory). My dl speed is 500Mbps. (my desktop and same server via 1Gb network = 104MB/s)
pi-rat Posted December 22, 2019 Author Posted December 22, 2019 Well, that little difference between my PC and my BP is nothing I bother about. But I am still very curious about what cuts the download speed to half, when both machines download the same source. It feels as if the download server does not really want to give its data to the BP. ;)
Recommended Posts