Well ironically enough, I used to work at company that builds custom computers by order for about hlaf a year, so 10 builds yeah in a slow week maybe.
After that I workd at IT for 3 years in a big company , I saw more computers and computer problems then you can imagine, including a lot of ram related issues.
Obviously I am not saying that it can't be a problem as I can only speak from my personal experience but again I never saw a motherboard with a new kit of ram not working because the ram is not validated by the manufacturer (and I can also tell you that mixing ram sticks from different kits also more often then not working just fine with at most some few tweaks)
If you want to know why that is, it's because the motherboards and rams are made to meet a given standard and it's the same standard across the board.
they just can't validate every kit, so your 10% incompatibility rate is extremely high
Cheers
This is enormously interesting, because for somebody whit that kind of experience you obviously live with the impression that this sort of incompatibilities is extremely rare, to the point of not being significant for the choice of a model of RAM sticks.
Obviously, this existence of industrial standards plays an important role in the production of any device, but that leaves us to question why the manufacturers of motherboard produce that sort of listing at all.
Recently, I payed some attention of the choices made by an online retailer of custom made computers as a friend decided to buy from them: cyberpowersystem.co.uk. And I have the impression after going through a whole list of their products, that they always take the RAM they use in their build on the QVL for the motherboard the customer chooses to use.
Those considerations (what you say here above, and why this retailer always seems to take the RAM from the QVL) led me to some head scratching, so I went back to my black board for some further considerations (as you have guessed, statistical analysis and probability calculation is my field).
Now I will consider this problem from the point of vue of a big retailer with an output of hundreds or thousands of computers. We can even imagine that it is in fact the production from all peoples who are likely to come and read, or even discuss on a forum like this one, even if, individually, they build only one machine.
Now, lets consider that the 10% I used for the first calculation is extremely high, and let consider 0.5% only.
But now, we decide to use 2 RAM sticks in each computer build (which is certainly more realistic).
So for each computer the probability of NOT seing an incompatibility is 0.995² = 0.99 (99%). So clearly this goes with your point of vue.
However, if we transpose this to building 1000 (one thousand) computers, the probability of NOT encountering an incompatibility is only 0.00004%.
So I think that these consideration allow to reconcile the 2 points of vue, by saying "yes, if you build a single computer the probability of not encountering that sort of problem is very large, however if you are a builder of numerous machines, you better take attention to these details, because their is a very big likelihood of encountering that sort of problem and you would be better off by using tested models of RAM to minimize the number of your problems"
The field of probability theory can be very confusing and counter-intuitive, and people generally struggle to transpose their daily experience to a probability in the general situation
Facebook
Twitter
Instagram