lti wrote on 2024-09-25, 02:22:

It seems like Realtek learned to make decent Ethernet NICs after the RTL8139, but Intel is generally still better for more deman […]
Show full quote

It seems like Realtek learned to make decent Ethernet NICs after the RTL8139, but Intel is generally still better for more demanding stuff (as long as you didn't get the 2.5Gb chip). Their audio codecs have been decent (better than Conexant) for a while, but the actual limit of your PC's audio quality is the computer or motherboard pack-in audio "enhancement" software that randomly enables effects and can't be removed (if it isn't bundled with the driver, there's some kind of flag where Windows will reinstall it automatically).

I've never heard good things about Realtek WiFi, but they aren't the worst around. You could have that Atheros USB dongle I mentioned earlier.

eton975 wrote on 2022-11-11, 03:25:

Not OP but I do know that on the Intel Gigabit CT (82574L-based) ones, typically the faked Delta isolation transformer has printed rather than embossed text with the Delta logo. On the I210-T1 and i350s, etc the Intel logo itself on the card PCB may be missing or look wonky, but from my understanding fakes do exist with the proper or very well faked logo, so this is no guarantee of a genuine card.

I didn't know that anyone would make counterfeit 1Gb NICs. My idea of a counterfeit NIC for all this time was an old card relabeled as a much faster one. I've seen mini-PCIe WiFi cards labeled "Intel AX200" on Amazon (real AX200 cards are M.2 only), and they cost the same as a real AX200.

In 1997, the first Ethernet adapter that I used was Realtek RTL8029 based. It wasn't anything special, but it worked and was reliable. I used it for about 2 1/2 years . It worked for me under Windows 95, Windows NT4, DOS (packet driver) and probably Linux at some point (memory is sketchy on that one).

AFAICR, the Realtek RTL8129 was the one which was really not that great . The RTL8139, while not great either, was already an improvement and, in machines with fast-ish CPUs, it was not really problematic as a client NIC.

Just installed a D-Link DWA-582 wifi card with the RTL8812AE chipset on my 980DE3/U3S3 system with the FX-6300, which is running Debian 13 on the 6.12.48 kernel. It has had a few hiccups but seems considerably more stable than the first RTL8812AE-based card that I had in my brief testing of it. Download speed is mediocre at about 60Mbps, upload is decent at 80Mbps. Ping fluctuates depending on load.

Maybe the 'rtl8821ae' driver has been improved since the last time it was used, or the first RTL8812AE card was just a lemon that I was unlucky to buy due to chance?

I will need to do some testing in Windows as well.

Windows testing seems to be fine, *as long as* WPA3 is disabled and WPA2 or older is used. If WPA3 is enabled on my Netgear Nighthawk RAX36S router, Linux can connect to its SSIDs, but Windows will say it cannot connect, even with the latest D-Link driver from the website.

Under Debian the 5GHz band has some hiccups in the first 30-60 seconds of operation if gnome-system-monitor is open and on the 'Resources' tab, if a speedtest.net internet speed test is being run. gnome-system-monitor stutters when this happens and sometimes there is a socket error in the upload test under these conditions. After a few minutes (link training possibly) this issue seems to disappear.

Overall, a substantial improvement compared to my 2021 testing with the old Rotanium PCE-AC1202 RTL8812AE card I used to have as well as the TP-Link TL-WN881ND card with the RTL8192EE chip. The question is then: is the improvement due to a newer kernel and drivers, or did I just get unlucky with 2 dud lemon PCIe cards in a row all those years ago? What does everyone here think?

I think it's better drivers and possibly luck of the draw. Got a lot of dropouts testing on Ubuntu 17.04 with the DWA-582.

I helped a driver developer called "Bitterblue Smith" some months ago with a register dump of the RTL8812AE for the rtw88 driver. They fixed some issues with antenna selection code thanks to my help with that, meaning that that driver can actually take full advantage of 2-antenna cards using these chipsets.

I really hope that that WiFi driver's RTL8812AE/RTL8821AE/RTL8723xE support gets upstreamed to the mainline Linux kernel soon. With this special driver I am getting pretty decent performance and the 5GHz band actually works correctly. The trouble is, I had to manually install and compile this using DKMS for Debian 13 and I suspect many newer Linux users will see this and get frustrated quickly. I have to commend the late Larry Finger on finally helping push an effort forward to fix all this.

It's honestly sad that it took roughly 12 years since launch to get these chips working right. What a waste of time...

eton975 wrote on 2026-08-07, 11:22:

I really hope that that WiFi driver's RTL8812AE/RTL8821AE/RTL8723xE support gets upstreamed to the mainline Linux kernel soon.

Have any patches been sent upstream?

dr_st wrote on 2026-08-07, 11:31:

eton975 wrote on 2026-08-07, 11:22:

I really hope that that WiFi driver's RTL8812AE/RTL8821AE/RTL8723xE support gets upstreamed to the mainline Linux kernel soon.

Have any patches been sent upstream?

Good question. A quick google doesn't seem to reveal anything upstreamed in the specific context of the RTL8812AE/RTL8821AE for the rtw88 drivers, but I might not be looking in the right places. This is what I searched up on the kernel.org mailing lists. There does seem to be activity around the RTL8723BE/RTL8723DE chips though...

Correct me if I'm wrong of course.