Most reviews I have seen, like SmallNetbuilder and CNET show this to have the highest in terms of 5Ghz range/performance of the tested routers. 2.4Ghz is decent. My personal experience is in line with this as well.
I placed my router on top floor on the ground and with 5Ghz one floor below the router I get 60 MB/s (480+ Mbps) Down / 40-45 MB/s (320-360Mbps Up). In basement I get around 40 MB/s (320 Mbps) Down. The testing was done using a Dell Latitude E6230 with an Intel 7260ac adapter upgrade and a Netgear ReadyNAS 524X NAS.
Interestingly enough the R7800 even outperforms the R9000 in 5Ghz even though it uses the same WiFi chipset. I tested the R9000 for Netgear.
Hi. I currently have ddwrt on this router and want to switch to LEDE because I'm facing a strange issue where in just the 5 Ghz radio crashes while 2.4 Ghz radio continues fine. Log analysis suggested that it is a crash in the firmware. .Anyone else facing this issue on LEDE ? If not, I would want to try hnyman build. Thanks for the help.
Also, can I flash this directly over dd-wrt or should I flash back to stock before I get LEDE on it ?
Hi. I installed the official 17.01.2 build, and then upgraded to the hnyman's 17.01.2 builds from the official build using the sysupgrade tar file I found in hnyman's dropbox folder. While everything started off fine, I face a weird issue. Whenever I restart the router, all my settings are gone and it goes back to my default settings, thus I lose all my wireless settings. Infact, the router password I set is also gone. Same happens when I restore the settings I have saved to file. When I try to apply them, they seem to be applied, but the router restarts and goes back to default settings. Any ideas on how to fix this ?
Yeah, I know ;). I have subscribed to his tree as well. I don't have any IPQ806x to test. I'm waiting for the IPQ4019 related changes to show up. It can't be too long.
For both of your questions, monitoring is easy with LuCI statistics, which has functionality for both of those measurements. Just install LuCI statistics, plus the needed collectd modules: collectd-mod-cpufreq and collectd-mod-thermal
If you want more manual native Linux kernel stuff, advice below.
Earlier in this same thread, from this message onward:
You can manipulate the speed properties just like for other normal Linux kernels. But the manipulation possibilities are rather limited. Mostly you can adjust the scaling governor, is it "performance" = always max, or "ondemand" = changes according to the CPU load (=the default in ipq806x).
Example e.g. in CPU frequency scaling driver for mvebu (WRT3200ACM etc.)
Yeah it's kind of hardware nat for ipoe on wired connections. It offloads nat routing with switch inbuilt capabilities. It's not the hardware nat as in stock, as its limited to wired connections and ipoe. And it's still work in progress, it crashes occasionally.
@hnyman btw I've compiled that fast nat driver against k4.9, seems to be working fine. I'm updating driver sources with qsdk commits at the moment.
Sure. If you manage to make the needed patches to clear commit(s), I can test it. I have so far not invested time in looking to the matter, as the "original" fastpath changes from gwlim were a mismatch of fastpath plus everyhing else, and the approach from musashino was better but still looked unnecessarily complicated. If you have cleaned it up, that is great.
It also sounds promising if blogic gets his hardware NAT changes ready.
But I don't have that speedy gigabit connection that I would really benefit from those changes, so I have been lazy
I may be able to give it a shot over my gigabit connection. I can build my flavor if it will be git clone-able or if a working build is made available. Working meaning it won't brick the router. Wife won't be so forgiving if the internet dies and if I have to spend time to fix it.
You don't necessarily need a real Gigabit provider line to test the NAT speeds as it might be hard to organize a saturation from the real Internet and test full-duplex with it.
An alternative setup would be 2 PCs, each with a Gigabit Ethernet port (all modern hardware has them). The "server" PC, connected to the WAN port of the router, will need to run a DHCP server and iperf in the server mode, and the "client" PC, connected to one of the LAN ports, will run the client mode of iperf. The DHCP server on the server PC will need to be configured to provide an IP from a different subnet than router's own LAN. Afterwards tests can be done iin both directions (half- or full-duplex). The easiest way to organize this is to boot a live Linux distro on both PCs.
I understand your post. I also know how to test without an actual gigabit internet connection. I'm just making a point that use cases may affect performance differently. If I'm never stressing my wired or wireless enough I'm likely to claim that it is stable. Conversely, if I'm constantly pushing maximum throughput through all my interfaces, I'm more likely to see bottlenecks/crashes that arent visible to others.
By default, yes, only some 19 MB is available after flashing the router: to preserve the possibility to revert back into a fully functional Netgear OEM, the LEDE firmware is only stored to the 30 MB "ubi" partition and the 70 MB "Netgear" partition is left intact and marked read-only..
But you can compile an image that uses a larger portion of the flash:
Is this still a case?
I'm looking for a new router for 250/20 Mb/s connection to use with SQM layer_cake. I'm considering the Netgear R7800 but those new information are making me want to reject this device. Can anyone clear this up for me, please?