A great and conceptually valuable article … although old and distant in time, it is still valid. There is no need to cheat, the numbers speak for themselves !!!
Has anyone tested this kind of RAM usage since then?
I will add a supplement in terms of hardware …
@anonymous_user_f86262ff1 … yes it’s true, you’re right.
With such a long article, the author probably misinterpreted the reference of this theory, probably not exactly what he meant.
However … many hardware savvy people believe that lower latency at lower clock rates will result in more bandwidth.
Meanwhile, there are other issues that affect final performance - frame capacity. One of them is synchronous or asynchronous cooperation with the clocking of the CPU processor … platform, bios and the very structure of the motherboard … a river topic that can be written endlessly
The hardware performance bottleneck has always been motherboards and some bios. Only recently have they exceeded 32 GB / s
Additionally, each manufacturer has a different motherboard, although the concept is similar for them. If you don’t know what’s going on, it’s always about voltage states and software bugs.
Everyone criticizes, but even the greatest, wisest, most specialized minds… even in groups they can’t handle it. Even they are unable to push through some of the unknowns and limitations.
Another important thing is what motherboard you have and what RAM you have. Even the same, they can be different… they can have very hard to spot errors
For example, there are reasons why some motherboards do not support certain memory like Kingston, GSkill, Corasair etc., although in theory everything should be fine.
Nobody really knows these reasons… but engineers, real specialists - in the plural, do what they can to solve such problems. Therefore, there are compatibility lists in the specification.
I’ve been through this many, many times in practice. This is the area where theories and plans are most often actually adapted… to what they really are. Too many variables and design complexities, no unambiguous repeatability of elements in relation to the expected voltage states … always gives a relative and not fully expected end result.
Apple is a good example. On a dozen or so motherboards, only one could fully work with an installed and working Leopard.
Sometimes it also happens that some damage or an underdeveloped component in the motherboard can cause errors that we misinterpret as a badly written program. Sometimes it can be the fault of the ATX power supply, sometimes the board, RAM, graphics card … and sometimes the software.
Once, a laptop user came to me - an engineer, chief specialist in the server room … because he couldn’t turn on wi-fi in his laptop. He was convinced that it was a damage or a defect to the network card … it took me about 20 seconds to fix the problem. It is easy to loop in the problems of the abyss and lose your vigilance …
Sometimes intuitively selected and configured equipment will work much more stable, more reliable and … almost as fast as those made to a decent maximum … giving you a few seconds of advantage. And just like you pee, drink something or chew a donut in front of the computer - in general … it’s worth thinking about !!!