I've seen a few posts from a few years ago regarding throughput speed on the Nighthawk X6 R8000. All of the threads died out without much being figured out.
I've tried both the stock firmware (V220.127.116.11_10.1.67) and OpenWrt compiled from
I was seeing 400-450Mbp/s through the WAN to the ISP's speedtest under the OpenWrt image, and 900+ Mbp/s under the stock firmware.
In OpenWrt, I installed QoS just to make sure it was disabled. The posts were all suggesting QoS was playing havoc on the stream. I can tell you that it wasn't the limiting factor.
I enabled both software and hardware offloading.. This DID help, but only in getting me to the 400-450 mark over 300-350.
Anyone got any ideas? I'd much rather have OpenWrt on the router, but I can't justify the speed loss for it.
Is CPU maxed out when you run the speedtests? (
PID PPID USER STAT VSZ %VSZ %CPU COMMAND
9 2 root RW 0 0% 46% [ksoftirqd/0]
Not even close.. Though the
nic is showing 40-99%. But, since I KNOW I can get the speed in the stock firmware, I know the hardware can handle the link. For whatever reason, it can't under OpenWrt.
I've switched back to the ISP Provided Zyxel for now, but I'm not happy about it It does mean I can afford the downtime to get this right.
The r8000 is a dual-core device, so 46% CPU usage means that one of its cores is almost fully loaded.
I pulled the line stat.. In looking at it, this is probably what you wanted
CPU: 3% usr 6% sys 0% nic 74% idle 0% io 0% irq 14% sirq
Load average: 0.16 0.09 0.09 1/67 6204
Also, Packet Steering didn't seem to make a difference one way or the other, outside a normal variation from run to run.
Hey @jow or @anon50098793, do you have any ideas on this before I put it back to stock? I rdon't trust Netgear's firmware (https://www.zdnet.com/article/unpatched-vulnerability-identified-in-79-netgear-router-models/ but I can't run at half-speed or less, either.
i don't know much about that device... what are the soc / network chipsets? and what's in cat /proc/interrupts?
based on the stats and the previous feedback... my guess is that expectations may be too high... and / or like the others have touched on... newer codebase and / or unoptimised stack is revealing limitiations of the underlying hardware.
This is a known problem with openwrt and all broadcom routers, which nobody cares about, it seems. https://bugs.openwrt.org/index.php?do=details&task_id=2186&string=perf
You should use FreshTomato 2020.2 (.3 has new and unstable driver) and enable CTF in options.
Well, when on stock firmware, I get 900Mbps/240Mbps Down/Up.. On OpenWrt (built from
master), I get 400Mbps.. So I don't think it's the hardware that can't handle it, but the implementation.
OpenWrt's Nighthawk R8000 WikiDev
/proc/interrupts, I'll have to re-connect it to the network and get that for you.
no need... @a_guy pretty much provided the details in the bug report... so like he says... best to try and track down an alternative firmware with a "specific" codebase or just run stock.
Since the R8000 is Officially supported, shouldn't there be a maintainer? The has been "unassigned" and "very low priority" since early 2019. The whole BCM-chipset makes me leery of trying to poke at it myself, alone.. It's a PITA to get it to TFTP back to stock right
To solve this problem, Broadcom developed the proprietary ctf.ko module that watches in-system routing rules and implements NAT on its own. It results in much better performance (even up to 850Mb/s on BCM4706) while breaking things like QoS and advanced firewall rules.
Unfortunately ctf.ko is closed source and there is no open source alternative.
some details here
Yep. FreshTomato includes this ctf.ko module and it works. r8000 is supported by it. My old r6300v2 is also supported and I can get full gigabit WAN. FreshTomato is basically the only option for open source gigabit WAN but it doesn't support any of recent routers as it's stuck at 2.6 kernel.
This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.