Nov 6, 2019
9
1
15
System:

i7-4790k
Z79X Gaming 7 Mobo
32g Kingston 1866c DDR3 RAM (4x8g)
RTX 2070 Super
Corsair H100iGTX Liquid Cooling
Windows 10
Thermaltake Toughpower XT Gold 1475w PSU

I have had this setup for about 5-6 years now (except the GPU which I recently upgraded to) and never had much of an issue. Though I think now the MOBO might be <Mod Edit> itself.
About two years ago I ran into a RAM issue where my computer was in a startup loop and it just would not boot. I took the last stick of ram out and boom I was back up and running. I thought "oh dodgy stick" threw it to the side and shrugged my shoulders. For the next two or so years I had no issues running it with the three sticks... up until the last month or two when I started getting very frequent crashes. The crashes were Kernel-Power 41 in the eventlogs, so just flat shutting off, not shutting down. No dumps were being created at all and I started trying to diagnose it. Everything I tried was not successful. I ran memtest, prime95 (both flagged 0 errors or showed nothing concerning), uninstalled drivers with DDU and reinstalled latest, checked the seating of all my components, ran chkdsk on all my disks and repaired one of them.. none of what I have been doing has solved anything.

Now despite running Memtest twice over two nights and not getting any errors I noticed that Rimworld would consistently bring out this behaviour in my PC and it made me think, hey maybe there is something wrong with my RAM. Weird because all other games I play i wouldn't get as consistent crash behaviour.
Anyway, I started playing musical chairs with my remaining three sticks of RAM and the three slots they are currently using and have found that all my sticks of RAM are fine... BUT

a) Slots 1 and 2 can have any stick in them or combination of sticks in them and they will run fine.

b) If I place any stick inside slot three or four though on its own, I get the same startup loop I got two years ago when my first stick of RAM died, which tells me that these DIMM slots are kaput??? (It also tells me that the stick of ram i tossed to the side two years ago and have now lost was actually probably still fine)

c) If I have a stick in slot three and one other stick in one of the good slots my pc will startup and run but I will get these crashes frequently (they are seemingly random, though Rimworld seems to trigger it) but not a startup loop.

d) If I put a stick in slot four I get the loop regardless if any of the good slots have a stick in them.

SLOT 1 and 2 FINE
SLOT THREE KAPUT ON ITS OWN AND UNSTABLE WHEN PAIRED WITH A SLOT 1 or 2
SLOT FOUR COMPLETELY UNUSABLE WITH ANY CONFIGURATION

I was getting ready to bite the bullet and replace my MOBO, RAM and CPU all at once to essentially have an all new rig to go with the recently acquired 2070 as I'm now at the point where I can't alter any of those without having to upgrade the others anyway. But I was really hoping to wait out a while longer for DDR5 to be released and a new generation of MOBOS to support it. I don't want to drop AUD$1500+ now to have DDR5 and the next set of Mobos to drop in two years time, that would suck. I mean I already screwed myself over by getting the 2070 super not long ago for the 3000 series to just come in like a wrecking ball and make me hate my life.

I guess what I am hoping for is a miracle and that there may be something I can do to breathe life into those dead dimm slots and I'm hoping someone on here has a solution or knows of anything I could possibly do but I'm dreaming aren't I? Can DIMM slots <Mod Edit> themselves like this? What are the chances of two slots dying?
Reality is I'm probably going to have to replace my MOBO which will mean upgrading everything now right?

Any input or advice would be super appreciated. I'm no whizz with computers, capable yes, but there's so much I don't know.
 
Last edited by a moderator:
Solution
Have you ever replaced the processor in the cpu socket?
If so, you might have bent a pin in the process.
Bent pins have unpredictable symptoms, but ram slot issues seem to be one of them.

Were all of the 4 original sticks from the same matched kit?
Unmatched kits can cause strange symptoms.

Since you have a Z motherboard, you should have the ability to adjust ram voltage.
Try adding a bit more voltage, that sometimes makes things more stable.

If memtest86 can complete a few passes with no errors, Your ram is probably ok.
I have no clue why errors could be triggered by rimworld.
Perhaps there is a patch or setting at issue.

Lastly, consider running with 16gb since that seems to be stable.
I don't think I would invest in ddr3 ram if...
....

Any input or advice would be super appreciated. I'm no whizz with computers, capable yes, but there's so much I don't know.

Having tested all those memory configurations I'm going to assume you've re-seated the DIMM's enough to rule that out as a probable cause.

Each DDR4 DIMM socket has 288 solder joints. Solder joints are well known to develop cracks over time due to things like vibration and repeated thermal expansion/contraction. (4x288) or 1,152 opportunities for a crack and since the DIMM sockets are connected together in a daisy chain all you really need is one intermittently bad joint to affect more than one memory location. Motherboard mfr's have pretty good process quality but these consumer grade boards are hardly going to be built to high-reliabilty aerospace standards.

A crack is hard to find even under magnification. But if this is the culprit you could go through all the solder joints on the mobo backside, touching them with a soldering iron to reflow the solder. It takes time and patience, and a steady hand.

And then there's also the CPU socket's solder joints that run to memory locations. They're probably ball-grid array joints and even more susceptible to cracking. People have had luck reflowing those in an oven (famously on some Nvidia GPU's). But that's pretty much one of those 'either this or the bin' solutions.
 
Last edited:
  • Like
Reactions: SonicMan
Have you ever replaced the processor in the cpu socket?
If so, you might have bent a pin in the process.
Bent pins have unpredictable symptoms, but ram slot issues seem to be one of them.

Were all of the 4 original sticks from the same matched kit?
Unmatched kits can cause strange symptoms.

Since you have a Z motherboard, you should have the ability to adjust ram voltage.
Try adding a bit more voltage, that sometimes makes things more stable.

If memtest86 can complete a few passes with no errors, Your ram is probably ok.
I have no clue why errors could be triggered by rimworld.
Perhaps there is a patch or setting at issue.

Lastly, consider running with 16gb since that seems to be stable.
I don't think I would invest in ddr3 ram if you are contemplating a motherboard change.

As to ddr5, I would not worry too much. when it comes, it will come.
From a performance point of view the change from ddr3 to ddr4 at equal speeds was not much.
 
  • Like
Reactions: SonicMan
Solution
Having tested all those memory configurations I'm going to assume you've re-seated the DIMM's enough to rule that out as a probable cause.

Each DDR4 DIMM socket has 288 solder joints. Solder joints are well known to develop cracks over time due to things like vibration and repeated thermal expansion/contraction. (4x288) or 1,152 opportunities for a crack and since the DIMM sockets are connected together in a daisy chain all you really need is one intermittently bad joint to affect more than one memory location. Motherboard mfr's have pretty good process quality but these consumer grade boards are hardly going to be built to high-reliabilty aerospace standards.

A crack is hard to find even under magnification. But if this is the culprit you could go through all the solder joints on the mobo backside, touching them with a soldering iron to reflow the solder. It takes time and patience, and a steady hand.

And then there's also the CPU socket's solder joints that run to memory locations. They're probably ball-grid array joints and even more susceptible to cracking. People have had luck reflowing those in an oven (famously on some Nvidia GPU's). But that's pretty much one of those 'either this or the bin' solutions.
That's some good ideas for a last ditch effort before turfing it! Thanks. Like you said there is no way I can see any cracks lol but if it comes to a last resort this might be a great thing for me to try. I do have a soldering iron but didn't think of this.
 
Try reseating the CPU if you haven't tried yet and check the CPU socket for bent pins. Look at the pins from more than one angle to get a good look at them. Sometimes CPU pins can shift out of place if the system is bumped or there is a lot of vibration in the case. A shifted or bent pin can cause a memory channel to stop function. Slots 1/2 are probably Channel A and 3/4 are Channel B.
 
Have you ever replaced the processor in the cpu socket?
If so, you might have bent a pin in the process.
Bent pins have unpredictable symptoms, but ram slot issues seem to be one of them.

Were all of the 4 original sticks from the same matched kit?
Unmatched kits can cause strange symptoms.

Since you have a Z motherboard, you should have the ability to adjust ram voltage.
Try adding a bit more voltage, that sometimes makes things more stable.

If memtest86 can complete a few passes with no errors, Your ram is probably ok.
I have no clue why errors could be triggered by rimworld.
Perhaps there is a patch or setting at issue.

Lastly, consider running with 16gb since that seems to be stable.
I don't think I would invest in ddr3 ram if you are contemplating a motherboard change.

As to ddr5, I would not worry too much. when it comes, it will come.
From a performance point of view the change from ddr3 to ddr4 at equal speeds was not much.
I haven't replaced the CPU no and the RAM was all purchased in the same kit when I first had the machine built.
Adjusting voltage is something I haven't tried hmmmm I might look into this thank you. I run Rimworld HEAVILY modded which is what made me think that maybe it was a ram issue, I think I run in the ballpark of like 70 mods lol.
I think I will just run 16gb for now, the main problem is that I use the PC for a lot of 3d modelling, texturing and rendering so it's obviously not going to be optimal.

I'm certainly not going to buy anymore DDR3 for my current rig as I think my sticks are fine, and I did find the old stick I threw aside. It was put back in the box it came in and I tucked it away with some other stuff so that was cool to find. Turns out that stick runs fine in the right slots just like I was thinking, so I'd bet the sticks are A-OK.

I didn't realise there wasn't much difference between DDR3 and DDR4. Is it possible to upgrade my MOBO to something that can handle a nice next gen cpu for my 2070 but still be compatible with my current DDR3 sticks? That would actually be a pretty good option. hmmm.