Hi,

Nemo1985 wrote on 2026-06-17, 14:00:

My guess is that the chip to change would be the one next to C1, but what passive components I should test to check if any of them is damaged?

Usually and looking at the pics there doesn't seem to be any passive components, traces or solder connections that show any visual damage, and the passive components on these sticks are not really have a tendency to fail, obviously you can measure them for shorts or such, but I find it unlikely for these to be damaged.

As from the memtest results it seems that only particular chips are damaged on these sticks, if you are very good at soldering you might be able to make one functional stick out of the two, this kinda defeats the purpose of a twin kit though.

Sometimes DDR sticks showing errors can be made working if under-clocked, however with more than 1000 errors, the under-clocking would be most likely need to be significant to fix the issue(even if it does), but this kinda defeats the purpose of high performance DDR sticks.

If you already about to write these sticks off as a loss, then you can try to also increase the voltage, rarely that can fix some memory cells to work in the chips, however this is both unlikely and has the risk of damaging the still working chips, but if they are about to be scrapped anyway there is no harm in trying. However if increased voltage does work, I recommend putting a heat-sink on these chips, as voltage increase will increase the heat generated by these chips.

I also know someone who believes that exercising RAM chips with constant 1 and 0 writes sometimes can regenerate them, never worked for me though, so I think he believed some kind of urban legend or a once in a blue moon kind of thing...

what voltage are you running them at?

I don't have any specific knowledge about these particular chips, but it seems somewhat unlikely that both sticks would be defective.

I remember struggling with memory compatibility on some nForce4 boards back in the day too. I think the most problematic was an Abit AN8 32X I was tinkering with in ~2012-2013, but that ended up having lots of bad caps later on too, which may or may not have been related. It would pass memtest with some RAM but not others, even though all worked fine on other systems.

Do you have any other DDR-400 systems you could test the BH5 RAM in?

I would check on another motherboard first. Preferably not Athlon 64.

Thank you all for the advices.
I did some further tests: give more voltage won't make them more stable, quite the opposite, the only almost working setting is v2.72, if I push more voltage they don't boot or give garbled screen in the bios already (I tried to push to v3), which is somewhat counterintuitive for a Winbond BH5 kit.
Apparently they work fine at stock frequency 333 mhz with 2-2-2-5 T1, but there are no way to make them work at 400 mhz.
I really don't know what to do, the model is: Kingston KHX3000/256 with declared voltage of v2.5

I agree with the posts above. A64 and BH5 was never a great combo.

You want an nForce 2 board with high vdimm.

Athlon 64 might not like that weird bank organization.