I am looking for a router with HW flow offload (currently only mt7621) and 4x4:4 n+ac WiFi.
I have compared the Table of Hardware with WikiDevi to find one without success.
Therefore a couple of questions:
Is there any other target / subtarget than mt7621 which has HW flow offload (even active development for such counts)?
Is HW flow offload worth it? One might take current and future WAN speeds in Germany (0.1 - 1Gbps down; 10% - 40% up) Home network (5 - 20 client devices).
Is ARM v7/v8 favorable over mt7621?
Is 4x4:4 WiFi worth it? One might take client device (phone and notebooks) support into account.
and of course is there a router with (future) OpenWRT support with mt7621 (HW flow offload) and 4x4:4 n+ac WiFi?
These questions are probably controversial and hard to answer in such an general scenario. Nevertheless I hope to get some (personal) feedback / thoughts also in regards to future proofness (coming 3-5 years). How does WiFi 6 / 802.11ax comes into play.
Lets try not to look to far into the future (our all beloved magic glass sphere) and focus on current state and until the end of the year.
There are a couple of other candidates (targets/ devices) which would provide the hardware to implement h/w flow-offloading, but none of those have a functional implementation yet. While a driver might emerge for some of them in the future, I wouldn't base my decision on faint hopes.
This is a good question, yes mt7621 can do routing, NAT and firewalling (simple rules only) at almost 1 GBit/s linespeed, but the actual CPU speed of the SOC is rather to the lower end of the spectrum. This means that any tasks beyond the hardware accelerated services can't cope with these requirements (prime example, VPN).
On the other hand mvebu (or x86_64) can do 1 GBit/s linespeed in software (easily), without a need for hardware acceleration - and these target platforms do provide significantly more power for additional tasks (e.g. VPN) than mt7621 as well. A weaker point of mvebu however would be the wireless side (mwlwifi), which might still be good enough for your needs - and on x86_64 you will need an external/ dedicated AP for all practical (and economical) intents and purposes anyways (e.g. ipq40xx/ ipq806x).
Personally I'd say yes, for the simple reason that all vendors (including Mediatek, see mt7622 (ARMv8)/ mt7623 (ARMv7)) have switched to ARM for their high-end devices. These days you'll only find mips in cheap entry level devices or legacy models.
Your phone won't support 4x4 (most likely 1x1 or 2x2, depending on the price category), your notebooks (aside from really expensive high-end ones) will do 3x3 at best, USB cards will only do 1x1 (or 2x2 with crippled antennas (too small/ too close to each other, effectively dropping you down to 1x1 as well)), 'normal' PCIe cards won't do more than 2x2 either, but there you can find (accordingly priced, you're quickly in the upper 2 or beginning 3 figure range) 3x3 or 4x4 models (antennas and pigtails becoming an issue). The only devices which might directly support and profit from 3x3 or beyond are probably tri-band mesh systems or top-end extenders.
However in terms of range and effective throughput over the distance, more antennas and MIMO do offer some advantages (as 802.11ax is already demonstrating), but don't expect too much from this either (wireless chipset/ driver and antenna/ pre-amp/ rf design quality matter just as much or even more, but you will see better solutions in the higher price ranges).
The only option supporting this (from the hardware point of view) would be mt7621 with mt7615 WLAN. While mt7615 has just been merged into mt76 a couple of weeks ago, these drivers are still very young… Right now, from a purely functional point of view, 2x2 ipq4018 will probably beat them hands down in terms of range, throughput and general driver maturity (this might change, if drivers evolve sufficiently in the future - but I wouldn't bet on it today)…
I'd look for at least dual-core, >=32 MB flash, >=256 MB RAM, those specs will be able to perform the tasks (WAN speeds) they're capable of doing now for the next couple of years and aren't likely to lose support within the next 5 years either.
802.11ax does improve the situation again (even if you only have 802.11ac clients), but currently none of these devices/ targets are supported by OpenWrt (there seems to be some initial work on ipq807x and ath11k, but I wouldn't expect that to become ready within the next couple of months).
This is a good strategy, but an answer is only possible if you narrow down the specification of your expected WAN target margins (speed/ throughput) considerably, as you have many more options (including reasonably affordable ones) for the lower regions of that spectrum - while >500 MBit/s limits you to mvebu or x86_64 (or mt7621 with the limitations of hardware flow-offloading, but not much more) from the get go.
E.g., if you're currently in the 100-200 MBit/s WAN speed range, without an immediate (next ~12 months) forecast to go significantly beyond that, you can get away with ~50-70 EUR ipq4018 (802.11ac/ wave2/ Mu-MIMO/ 2x2) devices - and re-check the market once you do actually need more than that (and will probably end up cheaper over all). In the 350-400 MBit/s WAN speed range, ipq8065+qca9984 (802.11ac/ wave2/ Mu-MIMO/ 3x3) for ~160-180 EUR would be the next contender (under the same conditions).