bugger

[kalniel]

Memtest86 is really great these days. But damn.. Explains why Forza Horizon 5 crashed so regularly..

[Jonj1611]

That's bad luck man. Not sure how many sticks you have, is it just one stick affected or all? Possible to be the ram slot?

[Hoonigan]

Ooft, that's rough! A lot of DRAM manufacturers offer lifetime warranties so you might be ok. Fingers crossed.

[Rob_B]

ouch, upgrade time!

[Iota]

I hope it's just a bad stick of ram, not the ram slot itself. Corsair are pretty good at warranty stuff from my experience.

[Percy1983]

Is that with XMP turned off, have you tried upping voltage a little.

[DanceswithUnix]

That's beyond turning it off and on again, but worth re-plugging the ram and see if it was a dirty contact.

Edit: I can't see how many errors you are getting there, I guess if it is a small cluster then that's the silicon but if it is just a ton of errors then it might be a contact.

If it was DDR5 then it might have silently fixed that for you. Still not sure how I feel about that, I like my ECC to tell me of problems.

[kalniel]

Good point on the contact - I've had the RAM sat in there for about four years so doubt anything could have wriggled in but I'll check that first - alongside which stick it is (though for purposes of warranty they're a pair so will both have to go back if so).

It is a ton of errors - I gave up after 400 or so (though I did retest on different CPU cores just to rule out an issue there, same result). The only other thing making me thing it's not contact is the data coming back is only slightly out - I'd have thought (naively not knowing anything about this) that loss of signal would be rather more catastrophic.

Initially was on XMP including 1.35V, saw errors, so switched to JEDEC reference speed/timings/volts and saw just as many errors.

[kalniel]

Yeah, one stick has errors in all slots. Other stick has no errors in any slot. That's a RAM failure

[DanceswithUnix]

- I'd have thought (naively not knowing anything about this) that loss of signal would be rather more catastrophic.

Academic I know, but...

The problems are all in one column of the "Actual" data. There's 2 data bits there that aren't always returning what they should out of the 64.

Sometimes:
F becomes E, so 1111 is reading back as 1110
F becomes 7, so 1111 is reading back as 0111
0 becomes 8, so 0000 is reading back as 1000

It looks like two of the connections to one of the chips have gone flakey. If there were no warranty, I would suggest reading up on how to reflow a broken GPU in your oven and trying that technique as it my well just be bad solder joints that have failed with heat cycling and they would reflow just fine. But if you have a warranty to lose, just send them back.

[Output]

I feel your pain, it's certainly not a fun experience.

Sadly I had to replace my own Corsair Dominator Platinum DDR4-3200 RAM a couple of months ago, after it would have been running for a little over 3 years (so its 36-Month warranty was already expired by then), due to it suddenly causing OS crashes and related instability out of nowhere, with disabling XMP not making any difference and both sticks showing errors in Memtest86 (I didn't bother letting it run for long to find out how many after it showed the first few errors though, as those few combined with the other issues were enough confirmation already).

It had even corrupted my firewall installation, which had its main process trying to eat up all the memory even when the RAM was replaced, resulting in a need to reinstall that too.

While I had considered choosing Corsair again, a search seemed to suggest that their RAM is considered overpriced and inferior to other brands these days.

Given the options I could see, and the fact there was a free trial of Prime available to get the replacement ASAP, I chose Kingston Fury Renegade 32GB (2x16GB) DDR4-3600 CL16 Kit.

[Saracen999]

Good point on the contact - I've had the RAM sat in there for about four years so doubt anything could have wriggled in but I'll check that first ....

Just as a quick aside on that, I had pretty much that with a video card (a couple of decades ago well before cards weighed half a ton and gravity alone could drag them loose, and not on a machine that was ever moved).

I removed the card, reinserted it and all was fine.

I can only imagine that it was years of tiny vibrations gradually worked it loose, tiny bit at a time. It's only a theory but I can't think of anything else. So .... likely? Nope. But possible? Maybe.

[DanceswithUnix]

I can only imagine that it was years of tiny vibrations gradually worked it loose, tiny bit at a time. It's only a theory but I can't think of anything else. So .... likely? Nope. But possible? Maybe.

Thermal expansion/contraction stresses can add up over the years too.

[Saracen999]

Yeah, could be. I couldn't think of a way of narrowng it down back then, and even less so looking back. I was just glad reseating the board worked. In those circumstances, it's worth a try, but far from sure to work. I got lucky. On that occasion. Not so much when something (I suspect PSU failure) blew physical chunks out of chips on 3 HD controllers, wrecked a £1000+ workstation graphics board and fried about half the machine. I was just glad the other half remained unfried.

My office had that nasty, acrid electronic burning stench fir about a month though. Not fun.

[UseItNow]

I've had those kits, well severall with differing speeds and I recall them being with lifetime warranty via Corsair, so you could enquire if you still have then, though I never had a kit of them fail on me but have just recently had a Dom RGB kit fail on me, first kit of RAM to do so since some Crucial crazy speed DDR2!