I have built a desktop PC router which is running 23.05.5 x86-64 efi squashfs edition OpenWRT. Specs are as follows.
Intel i3 2100 3.1 Ghz dual core (SMT disabled)
1 x 2 GB DDR3-1333
4 port Intel Pro1000 PT PCIe 4x lan card
D-Link DIR-1320 wifi G PCI card (Qualcomm Atheros AR2413/AR2414 chipset)
4 GB USB 2.0 drive for OpenWRT
I have installed the following packages.
kmod-ath5k (wifi driver)
hostapd / hostapd-openssl
luci-app-sqm
luci-app-https-dns-proxy
luci-app-wifi-scheduler
The testing internet connection is 85 Mbps down / 8.1 Mbps up cable configured with 82000 down and 8000 up in SQM piece of cake. A wired desktop PC speedtest's 78000 down and 7500 up 100% stable with 11 ms latency.
The issue I am encountering is that no matter what I try, I cannot get any device on the WiFi G connection to download at anything more than 23 Mbps. I have used an iPhone 13 Pro Max, a Pixel 7 and an Xbox One console. Both phones have been less than 1 foot away from the antenna of the wifi card. Luci's wifi overview summary page shows the link operating at 54 Mbps while the receiving device only shows a maximum of 23 Mbps. I have the transmit power of the hotspot set to it's maximum (30 dbm / 1000mw). I have tried enabling and disabling WMM mode, no difference. I have tried enabling / disabling packet steering, no difference. I have tried disabling wifi security (no password / open network) to see if it was a problem with AES encryption performance, no difference. I even tried changing hostapd from the vanilla version to hostapd-openssl which is running now (have not reverted back to base hostapd), no difference. Channel is set to 6 with noise rating of 92 dbm. Not sure what else to try. Shouldn't WiFi G be capable of closer to the theoretical 54 Mbps if there is only 1 device 1 foot away from the antenna?
54Mbit is the PHY rate. PHY rate is the raw theoretical maximum that can be physically transmitter/received. Since WiFi's designed to work on a public, shared, unreliable medium, at least 50% of that raw transceiver bandwidth is already wasted in error correction mechanism (coding and modulation of transmitted frames), management frames to establish/maintain connection between AP and clients and lost/corrupted frame retransmissions. And then, there's protocol overhead for upper layers that further reduces what's left for data payload.
In conclusion no, you can't really squeeze much much more than 25Mbit/s of throughput out of 802.11g. If you hope for ~50Mbit/s, you need 802.11n or better.
Thank you for the information. At least now I know it's not a setting I've got wrong or something. I assumed there would be more available throughput than that when I planned to use this old WiFi G card I had laying around, might have to upgrade it after all.
These cards are two decades old, they have always been slow (still using three of them, albeit in old client systems that don't need much performance and are only powered up very infrequently). 16-25 MBit/s is the most you can get out of them under ideal circumstances - and the lower end of that range is more likely.
802.11ac and 802.11ax are really a noticeable performance explosion, 802.11n not so much (for many reasons).
G…it is actually this year 21years old, I would say it is actually the first usable meaningful wifi anyone had at home in the good old days or something like that. https://en.wikipedia.org/wiki/Wi-Fi
But the thing back then is that no one really expected anyone to have more than one single wifi laptop at best. There was no iot or smartphones back then. The mobile phone/cell phone to have was like Nokia 3310 and that had no wifi or internet either.
And now you have connected three devices to your single G connection and expected 54Mbit/s for all of them!?
The thing with modern wifi like ac and ax is not only higher frequency. You have beam forming and more bandwidth called QAM also so multiple devices can actually use the same AP and always expect max bandwith for all connected devices.