My E6300 was at 3.4GHZ to 3.5GHZ and the Q6600 at 2.4GHZ - I also had the A6 3670K and a Core i3 2100 and a i3 3220. Mates with some of the later APU chips. I tried switching off HT and running them as pure dual cores - the A6 3670K still was preferable for me(that is with using a discrete HD5850 in both). The whole issue is not the per thread performance,but the fact the OS still needs to schedule tasks between the threads - 4 slower threads is still going to be generally smoother IMHO than two quicker ones(unless it is a mahoosive difference) and remember we are not looking at even double the speed per thread on the Intel chip. If you look at the more Intel friendly Cinebench software,R15 scores about 90ish on ST cores,and the G4400 is around 130ish,so around 45% faster on the Intel chip,but it has half the threads. It is going to have lower throughput.
Reviews are going to just run clean systems in a best case scenario - they are testing isolated CPU performance.
There are even extreme examples of that - a while back I built a rig for a mate who was running single threaded bioinformatics stuff. We ended up with a Phenom II X6 - we ran a few tests with other CPUs and the Intel chip was faster,but the issue is that in throughput terms it was worse(6 vs 4 instances),and he still had a usable rig if he was running a similar number of copies of the software,.and it was a consideration since these runs could last days at a time.
You make a good point about the HEVC encoding abilities of the G4000 series(I believe AMD Bristol Ridge will appear with it sometime this year),but remember there is still going to be a degree of load on the CPU,anyway and the G4400 has a lower number EUs when compared to the G4500. The issue I have is that most people are using these chips in a media setting as dedicated HTPC machines - the problem,is the father of the OP is using it as his primary machine. I still remember that video on a well known YT channel when someone tried streaming and gaming at the same time an FX8350 managed to do outdo a Core i5 3570K(!!),and they could not even believe it(even though it was repeatable),and they had never really bothered with an AMD chip and bought it on a whim. Then there is the other one on AT forums when a chap noticed if he pushed the IGP on a fews CPU including Intel dual core and Core i7 chips(IB or Haswell - can't remember which one),the CPU scores went down quite a bit,and the CPU was not throttling,and it was worse on the dual core he had(the A10 he had strangely did not suffer as much even though AMD has more issues with memory bandwidth with their memory controllers than Intel in theoretical tests).