BOSD every week on a server up 24h, maybe hardware degradation?

the same slot, the only pci express available. The mobos and CPU are equal, even RAM modules.
What can't cause memory violation?
Power/defective power supply?
Wrong contact of the CPU cooler?

There is also a BIOS malfunction that occours after the BOSD.. it freeze and doesn't boot again. It may be connected to the SATA system, *may*. The thing is when it occours I'm not there to see the monitor, and when I switch it on the HDMI is with "no-signal", maybe after minutes the BIOS could itself put the GPU is power safe mode? Then I can't see where it hangs.

So briefly there are two problems:
1) unstability
2) when unstability occours, BIOS isn't proceeding with boot after.

Thank you

Hi all!
I had some time to run Memtes86 and repeat it.. in 30 % of cases I noticed a RAM fault at the same address.. about position 2,81GB.. and only there! Then I copletely switched the RAM modules between servers to see if it was a RAM fault and it compared again. So, like someone said here, the problem is on motherboard! I still can't believe how motherboard can fail a specific ram location only.. for me it's a mistery!
Fortunately the board is under guarantee and I'll ask for a replace.

Also, I disabled the AMD IDE settings under BIOS, which created that "INIT" page on display during BIOS boot, and tonight the BOSD "a" appeared again, but the BIOS let it reboot.

There is not big sense in your words, sorry, a pin can't be assigned to that small area of ram only, if it is broken the problem will recursively reflect to the other areas of ram too. Plus, as I wrote, the problem doesn't compare at each memory scan, but appears only some times, less about 3 times every 10 test restart/system reboot, and with only some tests of Mem86, this is really confusing. With some tests I tested ram in that area for 40 minutes without revealing problems, we may suppose that bilions of data where written and read back in that "faulty" locations without issues, another time after a reboot as soon I started memtest86 I saw a scroll of 19 errors in a while..
Nothing broken here, it's a strange behaviour of northbridge, maybe connected with access times. In my opinion there is a parassite capacitance somewhere on the tracks of the board