Yup to all that, no tests are run at the same time try to eliminate as many variables as possible. But odd how consistently wired devices and the AT&T gateway are in the upper 900's and symmetrical while the router is today all over the map, from the teans to the 700's and very asymetrical in terms of up/down results. Either AT&T is playing shennanigans or something really funky is going on in the AX88. My speedtest results have been all over the map today and on some sites have recieved a message about a malformed packet and then the site reloads. I'm going to replace the Ethernet cable betwwen the router and AT&T gateway, I'll be amazed if that's it but have been around the block enough times to where I often quote the General in WarGames when they go try to bring up games on WOPR...
Actually replacing both cables, router to gateway and gateway to ont...
UGH, figured it out the hard way. Trend Micro! HW reset router and nodes, several times, methodically testing along the way. Slowly and methodically doing a speedtest with the native tool on the router. I got my symetrical +/- 940Mbps back until I enabled a Trend Micro feature. I was going to change the cables but after sleeping on it, if the PC and modem are doing great, why the cables...
Turning on Bandwidth Monitor results started to get wonky, assymetrcal and dropped 100-200Mbps on average. Then turning on AIProtection made things even worse, Speedtest results all over the map, sometimes not completing, sometimes in the single digits, dropped packets, though the AT&T gateway did not report any dropped packets
. How did this happen? So early in this thread, seeing the messages on the topic, I turned on the Bandwidth monitor, just because I wanted some detail on usage. That's when the problems started. Then when I disabled/uninstalled Skynet and got back some performance, I enabled AIProtection. Digging the hole I was in even deeper. The odd piece in all of this is that I didn't notice any real changes in the way CPU was being consumed when all this mess was going on. But after I load this router back up the way it was, I'm going to run through my testing again, what I suspect is that the CPU utilization peaks are going to be lower and of shorter duration when testing without Trend Micro features.
The three good things that are coming out of this are 1) starting clean with 388.2 beta, getting rid of all the nvram variables left by prior releases and the bouncing back and forth between 388 and 386 families for WiFi stability and 2) my technique for add a AIMesh node worked flawlessly (wired and wireless). My trick is to HW reset the node before it can be discovered and added, did the process a few times through the testing I was doing. Even after it being added if I reset the router then only way to re-add them was to HW reset just before starting the discovery process (WPS method), worked every time. If doing a wired backhaul, set the router up first for wired backhaul before resetting the node(s) and launching discovery (connecting them up correctly of course, WAN on node to LAN on Routher)
The third issue was with Site Survey causing me to think I had interference with 36/160 and me moving to 40/160 after hitting rescan a few times in a row caused the 5Ghz radio to move to 149/80. Since I did a reset, the default is Auto for channel, and back on 36/160... There's someone nearby on 36/80, a bar, but 40 is wide open, so back on 40/160 I'lll go. The hitting of the rescan repeatedly (when complete) did not cause a channel change like it was doing before. After many attempts of the sequence, the channel/bandwidth did not change.
One other thing I discovered by accident as I was switching the port the cable coming from the router into the gateway uses. When I did a client discovery I found some of my WiFi clients on the list even though I have both the 2.4Ghz and 5Ghz radios disabled on it and made sure the SSID are not visible. Not at all sure how that happens, because it shouldn't with the radios off/disabled.
