That is exactly why i said that the “idea” / “thought” of me was wrong.
Though, it really depends on the amount of discharge and charge between the transitions.
So it might not work as intended, or it might.
This is dependant on CPU/BIOS capabilities - See Passmark table:
https://www.memtest86.com/compare.html#fn5
1 Like
Didn’t really have time to order new parts yet but I wanted to report that on my other asrock board: b450m Steel Legend + 3700x ; Ecc is functioning and the reports are not masked by PFEH - so it’s disabled. (There is no PFEH setting in the bios).
1 Like
So I want a new system with ECC and VFIO. So far it’s looking like I’ll be going for an enthusiast X570 motherboard. It’s been about a decade since I’ve done overclocking or dug around in those kinds of BIOS screens so maybe I’m missing something here but…
…what about undervolting the ram or overclocking it out of spec to generate bit errors?
Personally I’d feel a lot more comfortable doing that to validate the motherboard over shorting RAM pins.
Edit: this might work as a test on a server motherboard without as many BIOS options? Just get the slowest ECC compatible memory you can and try to run it at a higher available frequency? ehhh unless you have the memory sitting around already this is probably going to be a nightmare of compatibility. BuildZoid’s youtube “Rambling about DDR4 chips and PCBs” talked some about his problems of just popping in unvalidated memory that had a different DDR3 layout the manufacturer hadn’t tuned for.
@CyberTrog I agree, the easiest way is to lower voltage until ram is unstable enough. Lowering some timings like CL or RFC to the limit might also do it !
I recently pushed some CL22 3200 micron E-Die ECC UDIMMs to 3666 CL16 @ 1.40V and ecc errors proved very helpful to fine tune them (by not having to spend countless hours doing memtests !).
Reporting is working correctly on my board (asus proart x570, with default bios options) and, when unstable, WHEA 47 single-bit memory errors are reported in windows system logs. Some tools like OCCT memory stress tester are also able to pick them up which is great !
@CyberTrog @Ivanovitch_k
If OC/undervolt works then yes, it’s an easier way to setup.
But that will work only in some hardware/bios combinations.
Some boards’ memory training will be aggressive enough that the edge between stable and not-posting can be very thin.
And OC may prove only that the board is reporting errors if they happen.
But not if the board won’t rapport them.
To elaborate:
If the board you are testing on is not showing errors no matter how long you tweak the OC settings then you really didn’t prove anything. It may be that the OC is stable and errors aren’t happening but the board would report them if they did.
Alternatively the errors are happening and are being corrected silently without any reports (PFEH)
You just can’t be sure which one is true.
Besides it takes time to balance the OC (assuming you have OC options in the BIOS).
Shorting pins is fast and takes out the guesswork.
2 Likes
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.