Because each site is creating the results they want to get a different angle out.
On THAT particular review session they gimped quite a lot of things:
They also used 2400MHz memory for all platforms, unusually low especially since we know the Intel chips can get along with much faster memory speeds than Ryzen.
No such thing as a fair test. You get the results you set it up for.
Fact is, Intel support officially 2400Mhz, AMD support officially 2666Mhz. Mobo makers include higher support for AMD and Intel than the officially stated support, but ultimately 1Mhz beyond 2400Mhz for Intel and 2666Mhz for AMD is overclocking. So in fairness, if you buy a Dell system, you aren't going to find 3000Mhz memory are you, you'll get what is officially supported, 2400Mhz. For AMD you will get up to 2666Mhz.
For instance Alienware sell 4 computers, three come with 2400Mhz, one actually comes with 2666Mhz.
If you're testing stock against stock, that should include at most the maximum memory speed a company claims to support at stock, not what a motherboard adds as an overclocking option for one platform but not the other.
It's also rather clear that A, a lot of AMD motherboards will run 3000Mhz just fine, B, memory support is improving every week and C, new platforms will always take a while to get up to speed.
The only remotely fair way to test is to give both systems either their max supported, or even the same memory speed. Under normal situations, lets say a gtx 1080 using maxed out settings a 1080/1440p, then overclocking memory won't make much if any difference. If you're running lower IQ/res settings specifically to put the load on the CPU more heavily, then you are also going to run into memory limitations, removing them for one platform but not the other is absurd.
There were a quite silly amount of reviews that ran 2400-2666Mhz for Zen vs 3000-3200Mhz for Intel. If I was reviewing, I'd include pure stock tests, that means, max official memory speed, normal clocks, basically nothing at all changed. Then I'd run max comfortable overclock on both memory and cpu on both systems to get overclocked results. Showing overclocked memory for gaming on one system and calling it stock for Intel in one of the only scenarios that shows a difference in memory speed AND at settings people never actually use is so incredibly biased and misleading.
It's entirely fair to say, overclocked, Intel platform is far more mature and you might get 4000Mhz out of a super set of memory on Intel, but only 3200Mhz with the same set on AMD. It's entirely fair to use both those results comparing overclocked systems, that is what overclockers might find to be a real use case scenario and it's entirely valid. Though I'd also point out that it's extremely likely as with every new platform that over time AMD memory speeds will come up. But even for a site like Anandtech, 80+% of readers will still end up buying a OEM system and not overclocking and they won't get 3000Mhz memory with the system, they'll get a stock system with 2400Mhz memory.
Any site comparing stock to overclocked immediately loses credibility with me. I also hate it with GPUs, some site will benchmark a purely stock lets say RX480 against ONLY overclocked gtx 1060s and proclaim AMD dead. Those reviews should show stock 1060, stock 480 and the overclocked 1060 results. But leaving out the stock 1060 causes a much bigger gap and gives a very different impression to a reader.