MaxTheFast wrote:
Thanks for the new infos.
Well if the MB's manual is the MB's fingerprint I think I'd buy the MSI:
- 100 total pages VS asus's 30
- 20 pages about BIOS VS asus's 4
- 4 pages about OC VS asus's 0
Therefore as far as I see MSI'd be a better bet while ASUS'd be an almost blind choice, right?

Maybe I missed something because it's my first time so here're the links so you can check it out:
- ASUS
https://www.asus.com/Motherboards/PR...lpDesk_Manual/
https://dlcdnets.asus.com/pub/ASUS/m...WEB_060418.pdf
- MSI
https://www.msi.com/Motherboard/supp...RO#down-manual
http://download.msi.com/archive/mnu_exe/E7B86v2.0.zip

Of the two you've shown me, the MSI admits to "DRAM voltages"
in the BIOS section. While some of the features might be
interpreted as "enthusiast", they also serve to help people
running "stock". For example, on the machine I'm typing on,
one day out of the blue, I started getting RAM errors.
By bumping up Vnb (Northbridge) a couple notches, stability
returned. It seemed like perhaps the Northbridge was
suffering from electromigration or some similar kind
of parametric shift. That was about four years
ago, and the board is still running.

And that's one of the things we look for, via the BIOS
screens, is "features not documented via other means".

If the manual doesn't have a BIOS section, it doesn't
necessarily mean the features are missing. Maybe the
Asus happens to have DRAM voltage adjustment too. But
if the manual doesn't confirm it is present, how
do we know for sure ?

I can give another illustration. The feature "ECC" for protection
of memory, is a particularly slippery feature. Industry
"rumors" will say "yadda yadda this board will have ECC".
*But*, if you don't see a picture in the manual of the
ECC screen, virtually 100% of the time you will be ripped
off, and ECC won't be there.

On my current motherboard, the Northbridge was supposed to
have ECC. But there was some waffling about which kind
of RAM (chipset supports DDR2 and DDR3), and ECC might
not work with one of them. On some BIOS, the ECC screen
will simply not even appear, unless ECC RAM is installed.
So Paul forks out for some ECC RAM and... no BIOS entry.
Nada. It's not there.

This is why the manual is *so* important. If you had
a must-have feature, and the motherboard "contract"
doesn't show it's there, you really have no reason
to complain. It doesn't matter how many false industry
rumors there are, it's the manual which is your final hope.

The MSI manual has "A-XMP" which stands for "AMD version
of Intel XMP". This automatically operates the RAM at
the spec speed. If your processor isn't rated for that
speed, it might not be stable. But as long as you stick
with two sticks, for reduced bus loading, that
gives you some small chance it'll work.

Now, another smart thing you did:

HX432C16PB3K2/8

https://www.kingston.com/dataSheets/...16PB3K2_16.pdf

You've bought yourself two sticks of *single sided* RAM.
This gives absolutely the lowest bus loading, and makes
it more likely your RAM purchase will work with the A-XMP
switch. If TEAM corporation was making enthusiast RAM,
that's how they'd make it, with the single sided config.

If it doesn't work (like the RAM on my other machine
didn't, when the bus was heavily loaded), you can
still tune the numbers and "make" it work. No shame
in that. Plenty of options. We can't see the detailed
DRAM screen (which would have a ton of settings in it),
but since they claim to have a screen, there is probably
one present.

Some motherboards even have an "auto-overclocker", but
for the most part these are not all that good. In
some cases, they apply way too much voltage (the
equivalent of using a hammer). If they built a tuner
that really worked, it would need to run all night,
and it would need a non volatile storage to keep
the results between runs. Real hunting for stability
takes "time and small settings steps". In fact, you
purposely search for "instability" such as a RAM error
every ten minutes, as you "walk the edge of the settings".
The idea is, once you've constructed a curve for the
hardware, you then "back off a bunch", knowing that
now the error rate will be zero. Then you do a
Prime95 run (on Linux so your Window disk won't
get damaged) for 16 hours and prove it's good.

Paul