I'm a little late to this thread (and this my first post), but I’ve been reading the informed commentary here for a couple years, but this is my first post. Here’s my situation and a couple questions:
My ISP provides 940Mbps symmetric (up/down) via fiber to my house, and their service is rock-solid. I have a Netgear R7000 (ISP-provided, I’ve cycled through three so far) that exhibits an annoying problem after 30-48 hours of use after a reboot/ power-up. Regardless of the stock firmware revision flashed into the R7000, after 30-48 hours of use, the data throughput (WAN to LAN, measured using the Ookla speed tool provided by my ISP, tuned for 1Gbit service) shows that my down/up speed goes from 896/927 Mbps down to 312/824 Mbps. When this occurs, I measure the rate from the OLT (optical line terminator) and I have a solid 940/940 Mbps down/up. The fix? Reboot/power-cycle the R7000, the data rate goes back to 896/927 Mbps, and all is well for another 30-48 hours. Lather, rinse, repeat; this behavior becomes rather tiresome after a while.
My question is this: Are there other folks here in the forum that have seen wired data throughput slowdown in the R7000 over time, or something similar?
Additional details: My R7000 is largely used to provide wired routing to a gigabit switch, which feeds the rest of the house via CAT5e cable. On the WAN port I have used both DHCP and static IP; I’m currently using a static IP due to a DHCP issue in the OLT. The wireless portion of the router is used to feed 5 old Airport Expresses which only support WPA/TKIP on 2.4GHz, and the 5GHz radio is set up with WPA2/AES for my iPad. I have disabled QoS, uPNP, don’t have any open ports, don’t use DMZ, no attached USB storage; all told, a very straight-forward system set-up. The wired devices on the network are a mix of 1000Base-T and 100Base-T, the usual stuff: computers, Roku boxes, TV, VOIP adapter, etc. (15 total).
My gigabit service was installed around 01Mar14, and I noticed this slowdown behavior within 2-3 days. In the intervening six months I have tried every version of the stock firmware publicly available, and after getting promoted to Level 2 tech support at Netgear, a few beta releases; all have failed to address this fundamental wired routing data slowdown problem. I kept detailed logs of data rates, firmware updates, and time to fail, and have forwarded these to Netgear, but they can’t/won’t replicate the failure, and ask me to run WireShark and generate more logs for them.
There is a bright side to all this: I have become an Asus fanboi. An engineer at my ISP, after looking at my logs and having a couple long discussions (I’m a retired engineer from the semiconductor industry, and we had some engineering experiences in common), loaned me an RT-AC68R, and that has been a revelation. Despite having a 200MHz slower routing core (BCM4708 vs BCM4709), the RT-AC68R provides better data throughput (929 Mbps down/940 Mbps up), and has been brutally reliable. As I said before, I don’t focus on wireless, but on wired routing, so I have been less sensitive to wireless throughput issues. I run on the Asus router when I’m not trying out new firmware on the R7000.
You might ask, “Why are you still fooling around with Netgear when you have a working Asus solution?” I thought I would stick with the problem until my warranty period expired, or Netgear L2 support folks gave up. I think the R7000 could be a dandy router (the BCM4709 certainly has the chops) if it just had decent firmware. Having gone through three R7000s, it’s not a single-unit hardware issue. I think this also points out that there aren’t many folks in Netgear’s user community that are routing gigabit data, or they would have fixed this by now. Also, if I was running under 300Mbps down/up, I would never have seen this issue pop up. At the end of the day, I can return the R7000 to my ISP and save the $5 a month rental fee, and in the mean time I have a relatively low cost toy to play with. I'm easily amused by these things.
With regards to the wired slow-down problem, I suspect there is an internal resource (register or software queue) that is not getting reset/cleared properly, and it takes 30-48 hours for it to fill up or get stuck, and that jams the little CTF packet accelerator state machine, dropping the data rate. When the code to drive the BCM4708/BCM4709 got written, the Asus guys got it right, and the Netgear guys didn’t. This really makes me wonder about the state of router testing at Netgear; you would think the high-end product would be throughly wrung out prior to shipping, at least when it comes to a basic function like wired routing. I have to say, this has really been an educational experience! Engineering truism: You don’t learn anything when things work correctly.
BTW, if anyone knows where I can find the detailed User’s Manual for the BCM4708/4709, please let me know. It looks like this is an NDA item from Broadcom, as all I have been able to find via Google is a block diagram and some marketing material. I’d like to see what’s inside the SOC, and get some idea about how it works.
That's my $0.02 worth on the R7000 vs. RT-AC68R opinion-fest. Thanks to all for any insight they can shed on this somewhat obscure and vexing problem.