Virgin Media has pledged to offer internet download speeds of at least 1Gbps to 15 million homes by the end of 2021.
Are there any off the shelf routers that will do a full 1Gbps throughput with openwrt?
I'd rather avoid a separate x86 board with switch if i can...
you will get better offer in 2021, if you don't have 1Gbps then wait until you have it.
There are some routers that can give 1Gbps thru NAT. In 2021 there will be more.
LOL , it's not happening in 2021 , it starts soon and will be completed by the end of 2021.
VM always takes that long to do 15 million customers with their upgrades as they can't post that many self install modems all at once. 
Supposedly, devices based on the MT7621 chip can do it?
More likely IPQ40xx, IPQ86xx, Marvell.
Supposedly, OpenWrt supports hardware flow offloading on Mediatek chips and not Qualcomm chips, so no.
There is no need for hardware offloading on mentioned devices to get 1Gbps.
The router/modem they will be supplying is rumored to have a Intel Puma 7 chipset what ever that is...
Plenty of all-in-ones can or will be able to support routing with NAT through a dedicated ASIC. Very few, if any, can handle 1 Gbps through the CPU. There is quite a difference between merely routing 1Gbps and doing something with it on the way.
A successor to arguably the worst DOCSIS chipset ever. For a start, see https://www.theregister.co.uk/2016/12/03/intel_puma_chipset_firmware_fix/ -- though patched, this was never fixed.
LOL , another cock-up to go along with Brexit.
mt7621 can definitely do it - I’ve had gigabit fibre (900 down / 450 up) for nearly 3 years, and have been using an ubiquiti er-x to drive it (initially with edgeOS, but using openwrt for the last 18 months). To local speedtest servers I see high 800s (which is about right as for various reasons my ISP chooses to use pppoe and hence a bit of overhead)
er-x/er-x-sfp won’t drive full symetric gigabit (due to hardware design choices) but other mt7621 platformns (mikrotik hexS etc) should be able to do better
Can confirm. My Dir-860l (mt7621) does gbit NAT. It won't do SQM/QOS on top of it though. For that x86 is the only logical choice currently AFAIK.
Unless things have changed i've seen lots of complaints in the past about bad 5G support in linux with slow throughputs?
QCA ath10k works fine with throughput in the 500 mbps range (can't comment on MTK or Marvell as I haven't owned any)
Edit:
Real-world performance, with an iPhone and a couple iPads attached and at least one actively being used, 5 GHz link over 802.11s with batman-adv routing and VLAN tagging, iperf3 running on the EA8300, 2x2 routers. Range between routers is great enough that mobile phones and laptops won't connect over the same path (-65 to -75 dBm reported).
[ 5] 0.00-1.00 sec 11.9 MBytes 99.6 Mbits/sec 238 79.2 KBytes
[ 5] 1.00-2.04 sec 11.0 MBytes 88.5 Mbits/sec 0 153 KBytes
[ 5] 2.04-3.07 sec 11.0 MBytes 90.1 Mbits/sec 0 191 KBytes
[ 5] 3.07-4.00 sec 7.98 MBytes 71.9 Mbits/sec 52 141 KBytes
[ 5] 4.00-5.00 sec 10.8 MBytes 90.6 Mbits/sec 0 189 KBytes
[ 5] 5.00-6.07 sec 11.4 MBytes 89.0 Mbits/sec 99 107 KBytes
[ 5] 6.07-7.00 sec 10.4 MBytes 93.8 Mbits/sec 0 165 KBytes
[ 5] 7.00-8.03 sec 12.4 MBytes 101 Mbits/sec 0 204 KBytes
[ 5] 8.03-9.02 sec 10.6 MBytes 90.2 Mbits/sec 0 236 KBytes
[ 5] 9.02-10.05 sec 13.1 MBytes 106 Mbits/sec 169 139 KBytes
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval Transfer Bitrate Retr
[ 5] 0.00-10.05 sec 111 MBytes 92.2 Mbits/sec 558 sender
[ 5] 0.00-10.07 sec 110 MBytes 91.8 Mbits/sec receiver
So in real world terms still less than standard stock firmware on a consumer router. 
No, I get 300-500 mbps the way the most people test the devices. I get 200-300 Mbps from my laptop. I gave you real numbers from a long-range 5 GHz link.
Here you go, does this "600 Mbps" number make you feel better?
tx bitrate: 292.6 MBit/s VHT-MCS 6 80MHz short GI VHT-NSS 1
rx bitrate: 292.6 MBit/s VHT-MCS 6 80MHz short GI VHT-NSS 1
Same link, "600 MBit/s connection" (2x2 devices)
I quote actual throughput over a path that most devices can't even connect over, not link rates.
Yeah, sure QSDK drivers might be a bit better in the lab. After all that's where reviews are done. That's about the only place you can ever run 160 MHz bandwidth, which is what the manufacturers advertise. Manufacturers don't advertise, "Oh, if you actually try to run 802.11n on 2.4 GHz in an urban setting, you might get 2 Mbps throughput, if you're lucky."
Yeah, right "1200" for an EA6350 or "AC3200". It's a number, that's not a measure of throughput. It's just marketing hype.
I didn't see any meaningful throughput difference between the stock firmware and the OpenWrt firmware on an IPQ4019, either at moderate or at long range.
I could give a $%& what a router does in the same room.
Sadly as a newbie and someone with learning difficulties i didn't understand 90% of your post or any reason why i would want something to make me feel better? 
All i understood was that which will be half of what VirginMedia 1Gbps is upgrading too so it looks like they will get lots of complaints.
Virtually no 802.11ac router can achieve 1 Gbps throughput with most mobile/portable clients under even laboratory conditions.
The highest modulation rate on an 80 MHz channel is 433.3 Mbps. 160 MHz channels are effectively unusable in anywhere but very remote areas (which often have military or weather radar installations, which also prevent use of 160 MHz channels). With a two-stream client, under ideal conditions (which doesn't happen with all-in-one routers as the antennas are too close together to be completely uncorrelated), that would be a modulation rate of 866 Mbps. Mobile clients that support three streams are very rare. Now you need to include the effects of the time slots taken up by back off, management and beacon frames (from both the APs and the clients), the overhead of the 802.11 framing itself, the overhead of TCP, and you're down to around 2/3 of that, somewhere in the 500-600 Mbps range for a dual-stream client.
None of this has anything to do with firmware. It is the limitations of the protocols themselves.
A rough rule of thumb is you can get half of the modulation rate as throughput in a "quiet" environment.
Have neighbors? Even a hundred meters away? Yes, their APs chew up your time slots as well. People walking past with cell phones? Yep, they send out beacon requests and every AP within range sends out a beacon frame. All of those chew up time slots. Poof, no more "theoretical limits" for you! Oh, yeah, beacon frames are sent out at the lowest rate supported by the AP, so it's even more painful than at first glance.
Even with a three-stream client, you'd be lucky to get over 800 Mbps throughput in a real-world environment.
Then comes the reality check -- even though you've got "1 Gbps" on your ISP link, what consumer-intended source can actually come close to that?
British ISP's have been fined before (by our ofcom) for over exaggerated speeds so they will have to be very clear when they release this product and the new puma based superhub to expect less than half the speed if i understand what your saying correctly...