Thanks for your answer. What do you mean with low network performance? My only issue is that I have a download speed of ~50 Mbit although it should be 100 Mbit. And with a 7362SL with identical 500MHz single core processor, it gives me 100 Mbit. So I guess the core of the router is not the issue, or am I wrong?
AVM leverages offloads pretty sggressively, well possible that indet current OpenWrt the CPU has to do more...
If this would be DLM related the sync speed wpuld be closer to 50/10, but sure your ISP might limit you yo 50/10 but thst would be the same independent of modem.... Maybe look at the output of top or htop while running a speedtest, this would show if you are CPU limited....
Thanks again for the clarification. However, in the related discussion, the "Data Rate" and "ATTNDR" differ from each other. However, for me, both are beyond 100 Mbit, and still, my router only gives me 50 Mbit. Or am I missing the point again?
Thanks, yes. So we see in upload direction 37.7 Mbps out of a sync of 42.463 which is only:
42.463 * 64/65 * ((1500-8-40-20)/(1500-8+34)) = 39.234 -> 10037.7/39.234 = 96.090 %
that is pretty OK. the Download however:
110.140 * 64/65 * ((1500-8-40-20)/(1500-8+34)) = 101.77 -> 10062.1/101.77 = 61.020 %
that is more of a disappointment... But it could well point at CPU overload...
This is not really all that dynamic, the system tends to intervene at most once per day... and that happens at around 3:00 AM. More importantly it will result in lower sync rates that would then be more in line with the speedtest results. So I see no clear indication of DLM in action.
You can try but typically the operating ISP is very much not interested in letting resellers like 1&1 'play' with the physical link, they sell bitstream access to the resellers and expect them to just accept what is delivered... again the sync rates do not fit with a DLM intervention...
I agree your issue seems not to be DLM. So please have alook at the CPU load during a speedtest...
likely, configure htop to show the detailed CPU stats:
F2(Setup) -> in the first column Setup select Display options -> in the display options list click 'Detailed CPU time...'
that will show the time spend in softinterrupt context that is where cake does most of its work in... but that display in your screenshot already seems to imply that CPU 0 is running out of CPU cycles...
Sidenote when I tried to use a BT HomeHub 5A (similar MIPS hardware and lantiq xrx200 modem to your FB) I did not really manage to get it to do all I wanted it to do even at 50/10 Mbps. I later converted that device into a bridged Modem running OpenWrt but put all other router duties to a more modern/powerful router. In that configuration the modem delivered up to around 100/40 (unidirectionally no issue, but a mild throughput loss when running bi bidirectionally saturating loads, but for my use cases then 'good enough'). I since switched to a second hand FB7520 which no operates as bridged OpenWrt modem showing no issue even at full bidirectional saturation... I would guess the 7520/7530 quad core arm a7 SoC might even be up to operating as OpenWrt router, but I will not try that...
qdiscs like cake are not multithreaded, so can only run on a single CPU, it might be possible to push some processing (e.g. the upload qdisc) to CPU1, but I am not sure whether that works for lantiq on OpenWrt 22.
So from the legend (F1) I guess it is mostly in soft-irq .
In other words, does this mean that the 7412 is too weak? I'm a bit suprised because 1&1 says that one can use the 7412 for their 100 Mbit contract. Is there anything I could do to put some tasks from the first core to the second.
Yes that is what I expected, that CPU seems over taxed, what kind of stuff are you running on that poor little router?
As I said, AVM is quite aggressive in using offload engines (not sure the 7412 has one) and in optimizing its devices to eek out acceptable performance over the range of supported DSL standards/speeds.
Not much. I installed an adblocker which I already disabled right now. I also installed dslite, but I don't use it in my network config, so I guess that shouldn't be an issue, right?
Other than that, there is not much running besides a vanilla OpenWRT. If I don't make a speedtest, both cores are around 10-20%, so I guess it's not doing that much in the background.
Do you think my 7362 SL would perform better than the 7412 when I flash OpenWRT onto it? I mean it's the same 500 MHz, but it has a Lantiq XWAY VRX288, which might be a bit better. Or should I trust @slh 's comment that this router too won't achive the 100 Mbit and buy something other like a 7520?
VRX268 vs VRX288 does not make a tangible difference, but specifically the F!B7412 is known to be quirky (as mentioned, presumably suboptimal PCB of the analogue modem parts). I would expect the F!B7362SL to behave considerably better than the F!B7412.
Performance of all these VRX2x8 devices is roughly the same (at least that don't disable one core in favour of the voice core), without software flow-offloading around 85 MBit/s, with software flow-offloading enabled they will do for 100/40 MBit/s - without much of a margin, but still in a usable fashion. Don't expect sqm (as moeller0 was suggesting, around 50 MBit/s tops), nor VPN on the router (don't worry, it'll work for casual usage (e.g. road-warrior style), but don't expect any speed records) - and the RAM size severely limits the size of about (ad-)blocklists.
The ipq4019 based F!B7520/ 7530 is in another league (be careful, neither the F!B7530B, nor the F!B7530AX will ever be supported; Broadcom SOC and wireless).