I have to disagree on one small note:
For a travel router most hotspots are going to be 2.4Ghz; so the dongle, with its better rx/tx is best used as the client and the Pi can use 5Ghz (or 2.4 but 5 does not go through walls well so less snooping in on it) in the hotel.
I should have checked: the 3 has only 2.4.
3B has only 2.4GHz (and 100M ethernet), while 3B+ is dual band (but also horrible 5GHz WiFi, with so-called gigabit ethernet running on USB2 which gives about 330Mbps max.)
OK let's go back to CF-953AX, I grabbed another RPi 4B to test, updated EEPROM with another SD card first, go back to OpenWrt and using the CF-953AX stock firmware (a bit old), it's kind of more stable, but it might still crash occasionally without reason (no obvious system log showing). Again, updating to the latest CF-953AX firmware will not allow the system to boot. But when it's working, I can sustained 600-630Mbps, sometimes hitting 700Mbps, speed is OK but still too unstable as AP.
So looking at all different trial, I don't recommend CF-953AX as AP at the moment, need to find a good way to update firmware of dongle first (I just noticed that on RPi's OS the CF-953AX has newer firmware loaded so maybe next time I have to copy it out)
Thanks.
That is a lot of additional work and information.
I realize you know this but might not have considered it: If you plug in a monitor it will show the bootup crash.
Thanks again. 
Or you could boot and then plug it in and see what happens.
That seems like the most likely reason for the failure. I bought a dongle from the Canakit people who sold me the Raspberry Pi 4. It has the same RaLink chip as the Eastech dongle. There's no model number on it but it's tiny like those bluetooth dongles that come with a wireless mouse. It's a 2.4G only USB 2.0. It works to receive wifi but it won't work at all as an access point to transmit wifi so it couldn't use it in my travel router project. When I tried to use the dongle to receive and the built-in radio to transmit the system just hung and didn't work at all forcing a re-imaging and my having to start all over - even when configured correctly. But of course it probably wasn't configured correctly due to my human error - so I gave up on it.
Wait:
There are decent wifi cards?
Some chipsets will do AP mode so only one card is needed, if we can find it.
The alfa will do AP mode. Software AP mode but AP none the less.
I used EDUP EP-AC1605 / ALFA Network AWUS036ACM as 5ghz wifi AP for months. It didn't get hot at all and never had problems.
Forget to mention more about the IO-DATA WHG-AC433US, this one is kind of weird when I tried to use under Linux (not limited to RPi). First it has 100% same outlook as the same IO-DATA branded WN-AC433UK, according to DeviWiki it's MT7610U, and from Windows I can see that WHG-AC433US is also 802.11ac WiFi loading a Ralink driver, I suspect they are actually the same but just different USB ID (04bb:0955).
Eventually I tested with following after installing kmod-mt76x0u:
echo 04BB 0955 > /sys/bus//usb/drivers/mt76x0u/new_id
It's recognized by the system immediately, proved that it's also MT7610U based USB dongle, I don't know if there will be any chance this USB ID got put back to linux kernel upstream, so now I have to make the above command part of the startup script in order to have it working.
Just got another USB WiFi, friend helped to get from China, it claims to be MT7612U based (model BL-R7612RD3), which is correct. Under OpenWrt it shows as AUKEY USBAC1200, however I couldn't find any matching from search.
But the very first test is not that positive, the mt76x2u driver recognizes it easily, however on my RPi 4B the AP crashes almost immediately when I pass traffic on it.
Thu Jan 25 18:22:59 2024 kern.warn kernel: [ 527.994568] xhci_hcd 0000:01:00.0: WARN: TRB error for slot 2 ep 7 on endpoint
Thu Jan 25 18:22:59 2024 kern.err kernel: [ 528.001851] mt76x2u 1-1.1:1.0: tx urb failed: -84
Thu Jan 25 18:22:59 2024 kern.warn kernel: [ 528.006607] xhci_hcd 0000:01:00.0: WARN waiting for error on ep to be cleared
Thu Jan 25 18:22:59 2024 kern.err kernel: [ 528.013770] mt76x2u 1-1.1:1.0: tx urb submit failed:-22
Above log flooding in syslog, and apparently it's related to USB, I searched a bit and looks like many people reflecting same issue on Raspberry Pi 4 with various USB devices, not sure if this dongle has trouble working with RPi 4B. I will test again with NanoPi or older RPi to see how it goes later.
Note: This dongle is USB 2.0 based, so 2T2R 866Mbps can only be displayed link rate, also I can feel it a bit warm even without loading.
If it is USB2 and it claims 1200Mbps we know that math won't work.
USB2 is 480Mbps. 1d+ 1d-. I do not care what it reports. I'd unplug it, cut it open with a dremel and look at the chip itself.
If it is wasting USB2 power into heat... ...I dunno. Is it metal?
Be careful with it, it smells of fake sd cards. I'd put it on a hub and hope it frys the hub before the pi.
Thanks for trying it!
Well this is disappointing.
I tested with an old laptop with Linux, confirmed the status before putting to the RPi4
Just now I moved this stick to NanoPi R2S, running in AP mode, this time there is no problem, AP mode runs fine, several Waveform Bufferbloat/Fast.com/Ookla Speedtest showing it's running good, temperature didn't rise up during on load.
But to my surprise, this dongle doesn't show me obvious performance increase when compared to Asus USB-AC51, I understand that USB 2.0 max. 480Mbps will be blocking the USB WiFi to attain it's max. speed, but from previous result of Asus USB-AC51, I wish this dongle can give me at least 300Mbps but no....tested many times with device sitting very close to the dongle, download speed max. 160Mbps, upload max. 230Mbps (most of the time it's just 200Mbps), I would say I definitely go for Asus one because it's a lot smaller in terms of the size.
This is promising...
Funny how life works.
I guess this won't be the issue, the device itself is USB 2.0 only, and the processor on NanoPi R2S is RK3328 which is ARM A53 with 1.5GHz based clock, it's not slow, the CPU itself should be able to handle all those with just single core (in fact the system loading is very low), so the CPU affinity probably won't help.
When I have time tomorrow, will try out NanoPi R4S (all NIC are from SoC/PCI, while the R2S LAN port is actually an USB RTL8153B NIC) to see how it goes.
But I've seen tests on some other dongle types MT7612U WiFi, probably due to bad antenna design so the performance isn't good, maybe my dongle belongs to this type.
Just now I took RPi 4B with latest snapshot to give it a try on the previous MT7612U dongle.
Now with kernel 6.1.x, the former mentioned USB issue has gone, which matches what was observed by folks in Raspberry Pi forum, they also find that kernel 5.15.x on Raspberry Pi OS having problems when dealing with various USB devices.
I did speed test several times without issue (before it was immediately stuck when I put traffic on it) then AP crashed, I guess it could be due to the nature of "snapshot"
Just to confirm, using older Raspberry Pi with my mt7612u dongle has no issue even with 23.05.2 release, so apparently the USB controller on RPi 4B is really picky about devices (I've also seen people getting trouble with some USB-SATA controller as well)
And to my surprise, I guess there was an update to hostapd recently? After my test with mt7612u, I plug back my COMFAST CF-953AX and now I see 6GHz available!!
Tested with my Android phone which has WiFi 6E, confirmed it's connecting with 6GHz without problem, of course my older Pi has only 100M LAN so it's only a showcase now, and I noticed that it doesn't have previous AP crashing when I put traffic on, will move back to NanoPi R4S for more tests soon.
Has anyone tried Amazon's best seller?:
Or this, "Amazon's Choice?"
You can search those models on internet to check what chipset they are using, if it's Realtek then you can skip (well, most of them are Realtek)
Both of those are Realtek. Skipping.
