CMZ64GX3M8A1600C9 - half a module died after 10 days of different tests

Hello. I bought a matched 64GB kit from Corsair, and after 10 days of testing, one of the modules (half a module, actually) seems to have died. I'd like to know your opinion about the reasons why this might have happened. I basically would like to know if "too much" testing can physically kill the memory. The system is the one in my profile.

This time I was using HCI Memtest (version of memtest that runs on windows), because in my previous build, when I was mixing two different sets (something you shouldn't do), memtest wasn't able to find the error, but hci was always able to detect problems in a few minutes or seconds. My hypothesis is hci may stress the memory controller harder. Nevertheless, the programs used and their effectiveness are not important in this situation as I never managed to finish all the tests I had in mind.

First I tested the memory with everything on auto and xmp profile.

Then I tested the memory and cpu again, this time with xmp profile and the cpu overclocked at 4.4Ghz (normal speed is 3.6Ghz). Vcore was set to +0.015. It's higher than that when left on auto. I was trying to find the minimum stable vcore for that speed, which is not a high overclock at all.

The ram was never overclocked and its voltage was never changed. I always used the xmp profile.

Temperature is fine, PSU voltages are in spec.

The tests were:

-First test

4 days: hci memtest (PASSED). Config: xmp profile, everything else auto

-Second test:

1 day: prime95 small FFTs (PASSED). Config: xmp profile, cpu overclocked at 4.4Ghz vcore +0.015

-Third test

1 day: prime95 Blend (PASSED). Config: same as before

-Fourth test

4 days. hci memtest (FAILED after 72h). Config: same as before

When the fourth test failed, I got tens of thousands of errors in the log for about 1 hour, and then a BSOD. When I rebooted I found out the ram detected was 60GB instead of 64GB. One stick is shown as "abnormal" in the EFI. It doesn't work regardless of the slot. The EFI correctly reports the abnormal stick wherever you place it. The rest of the memory is fine.

I removed all the ram and only left this abnormal stick. In this situation, only 4GB are detected, which is not correct as it's size is 8GB.

It seems the stick is faulty, and I'd like to know if I should RMA the kit. It simply doesn't look like there is anything else I can do. I am worried because I don't understand what happened, so I don't know if I there is anything I can do next time to improve the situation. Or if testing so much is just "bad", and I shouldn't do it, because it just feels like testing killed it, and that doesn't make much sense.

I'd like your input on the following hypothesis:

1- The ram has always been faulty. Maybe the heat sink wasn't properly attached on one side, and after testing for so long, the elevated temperature finally killed half of the stick. The temperature simply made it fail early. Which is good, because this way I won't have any nasty surprises later on.

2- The motherboard malfunctioned and killed half of the stick. I have no evidence for or against this, and I don't know if it's even possible.

3- The PSU killed half of the stick. It's 6 years old and I've never had any problems with it. I have no idea if it could kill half a module, while leaving the others untouched.

4- A long stress test increased the ram temperatures too much, and half a module died. The ram was not overclocked, voltage wasn't modified. The test was certainly longer than usual, although with so much ram it's just how it goes if you don't want to reduce the number of passes. It seems plausible, but I don't know how to monitor the ram temperature, nor what its safe value should be. A clarification on that would be appreciated.

5- A long stress test used the computer "too much", so "something" happened, and the memory broke. This seems crazy to me, both because it was only a few days, not years, and because other people use their hardware all the time to compute all kind of intensive simulations. I just don't understand why I should use the hardware "enough, but not too much", provided the temperatures/voltages are good. If anyone thinks this makes sense, please explain your reasons.

6- Undervolting vcore killed half of a stick. I don't know if it makes sense, probably not. I'm quite sure this didn't happen and it's probably impossible under normal circumstances.

Thanks for your help.

- The ram has always been faulty. Maybe the heat sink wasn't properly attached on one side, and after testing for so long, the elevated temperature finally killed half of the stick. The temperature simply made it fail early. Which is good, because this way I won't have any nasty surprises later on.

Highly unlikely, they dont generate much if any heat at all anyway. And they wont untill really overclocked.

2) Aint gonna happen

3) Aint gonna happen

4)Aint gonna happen. Again they dont generate any real heat. Barely warm to the touch. Extensive testing isn't an issue either.

5 and 6 are a no go too.

You just had half a stick fail. It happens every now and then. The links on the left to request a replacement set! :)

Thanks you very much. I will RMA as you suggested.

I would like to add that the memory burns if you touch it when you are testing with hcimemtest. I have no idea of the temperature, but it burns. I also tried other sets and they also get really hot when testing with this program. I just thought it was normal or maybe that my skin was too sensitive. This was after several hours, so I was worried about the temperature after days. Anyway, I will replace them and won't repeat the test, just in case.

Thanks you very much. I will RMA as you suggested.

 

I would like to add that the memory burns if you touch it when you are testing with hcimemtest. I have no idea of the temperature, but it burns. I also tried other sets and they also get really hot when testing with this program. I just thought it was normal or maybe that my skin was too sensitive. This was after several hours, so I was worried about the temperature after days. Anyway, I will replace them and won't repeat the test, just in case.

your just thin skinned:D:

seriously tho,memory will get rather hot when loaded up...