OK, thanks. I'll try again in a few snapshots, may be it will work better or will wait for robimarko or someone with SFP+ to comment on what's needed for that to work.
What switch are you planning to use for SFP+ by the way? I was also looking for something small and so far the only decently looking was Mikrotik CRS305-1G-4S+IN, but I'd consider other options.
Is it safe to sysupgrade from the Google Drive image to the firmware-selector snapshot? I think someone said they bricked it but I'm not sure if that was the cause?
worked well for me. I did have it do a factory reset on flasing though. or more specifically, I ticked the "keep settings" button off before the firmware was flashed.
That someone was probably me
From what I can tell, DON'T.
Apparently the partitions are not compatible and that will break the update process.
I'll also say that I have only tested going from robimarko's firmware (aka firmware selector snapshot) to remitor's (aka the google drive image), not the other way round. Since Gin sais it worked for him, my post here might be obsolete. - Just saying, the safest route is go to Stock, then to official.
My network is exclusively RJ45, with the 89Xs SFP+ port being the first of it's kind in my home, and two SFP ports on my current switch, unused since no device.
I noticed, and posted about it here already - but iirc snapshots tend to not include lucy by default, so it's not a big deal I guess. Just load it from the repo, simple as that.
Ah, I see. I'm interested in building 10 Gbps network (mixed fiber and DAC cables for lower power consumption than RJ45). So looking into some switches that are home / desktop grade, not those big industrial ones.
What's good about RT-AX89X itself is that it has an SFP+ port and can work as a WiFi access point even if you have a separate router which can handle full 10 Gbps traffic well (since RT-AX89X itself can't as was pointed about above). At least as long as it will be possible to make SFP+ work
For router I was thinking about something like DEC750.
That's the non upstreamed Qualcomm component for routing? Why is upstream kernel not good enough for that? Sounds like everyone needs efficient network so it's weird some non upstream components are still needed for it.
Interesting. What's good about DEC750 is that it has direct support from opnsense developers and whole official firmware is basically FreeBSD based and open source to begin with, so it's pretty neat. Plus it's AMD Ryzen embedded which is interesting (it's powerful, fanless and compact). OpenWRT can run there probably too.
But if RT-AX89X can be made to route 10 Gbps properly, then separate router won't be needed.
Sorry for off-topic. I do plan to use RT-AX89X with OpenWRT, just not sure if in full router capacity or as WiFi AP in addition to a non WiFi router.
Thanks, I think I get the idea, it's becasue of the separate accelerator chip which relies on non upstreamed code (Network Sub System). I.e. routing happens not on the CPU but on that chip which doesn't yet have proper upstream support apparently. I guess something like DEC750 gets away without the need for such chip due to having a way more powerful CPU.
It probably will never be upstream (Qualcomm decision is to make it open source) but currently there is NSS-wifi build (full NSS offloading support) that is completely working. I currently use it on my QNAP (it has two 10G ports) for over a year. Zero CPU usage while routing and full NSS wifi offloading. Absolutely stable and it's only up to you if you want to try it.
But I've just found out that when I connect Mikrotik SFP module to the AX89X SFP+ port 1G link is established and normal data transfer is possible.
That's with latest Asus stock firmware although they stated only 10G link is possible. Is it an error or just an outdated info.
As shown in this post further up, I’ve been looking into how to get this horrifyingly high temperature of 96C (205F) under control. Especially since the inside of the router started to smell suspiciously similar to burnt electronics, and the backplate got really hot aswell.
This is the tool I use to read out temps on the router:
From what I did observe while monitoring temps, this doesn’t seem to ever happen. The CPU keeps hovering around 65C, without a fan. This by itself would be safe and not an issue.
But as in my previous post, whatever gpio008 is, gets really hot, and nothing triggers the fan to kick in. That is not a sensor error, the temps are that hot!
After editing settings in /etc/control/alpine-fan-control to have a higher drv_speed_min, set tmp_sens to “gpio008” and fan_cont to “fan”, I got the fan to run and cool down the hardware sufficiently:
85 is still high, but acceptable. Also the backplate is much cooler now. I can even touch it without burning myself.
I do recommend to swap the original fan, as it is a bit noisy, and I get why people might want to turn that thing off I installed a Noctua 90x14 and even on full power it is dead silent, so now I enjoy safe hardware and a quiet room.
I don’t know if the previous software can be updated by cli, I did not find settings for it. But I do want to encourage to update the controller accordingly. I assume that other owners will be in a similar situation, and I want to prevent damage from occurring due to high temperature. The web interface will by default not inform you, so it is easy to go unnoticed.
So what exactly is the part that makes that gpio sensor so hot? And was it under load or during idle operation? In my minimal test it never went so high. I swapped the fan to Noctua NF-A9 5V but didn't change any hwmon settings, it's all stock.