Need Help With DDR3 Memory Error Issues

[fullonshred]

Have an older HP desktop. YEARS ago pulled the 4gb that came in it and replaced with 4x4gb=8gb Gskill Ripjaws DDR3. Several years later added 2nd 2x4=8gb kit of same model of GSkill Ripjaws for 4x4=16gb total. Ran like a top for years. Upgraded to Windows 10. Ran like a top for more years.

A few months back I started to have some W10 crashes, freezeups etc. Did the normal stuff, update all drivers, scannow error check, hard disk checks, etc - all showed good. Finally crashed so bad I had to restore from a System Image backup. Did updates again and shortly the same problems began anew.

So ran Windows 10 memory Diagnostic, it reported all was well, but I was unconvinced. Eventually ran Memtest86+ for some hours and had many memory errors. So ran Memtest86+ 1 dimm at a time and found 2 dimms that were throwing errors, other 2 were not. Pulled the 2 bad dimms and everything is running fine on the 2x4-8gb Ripjaws that never threw errors.

Here is where it gets weird. The 2 "bad" dimms only threw errors in slots 1 and 3, and tested ok in slots 2 and 4, but computer would still crash after a day or 2 trying to run the "bad" dimms in the slots where they tested good.

With somewhat varied Memory Test results I didn't want to bother shipping the 2 "bad" dimms back to GSkill if they are not really bad/or might test good and are denied replacement by GSkill. So before filing a warranty replacement with GSkill I decided to test the 2 "bad" dims one at a time in another DDR3 based desktop I own. Neither "bad" stick has thrown even 1 error in this different computer. Ran 1 stick over night and am currently 6 hours into testing the second one, with Zero errors so far.

I sort of have an idea what maybe could be going on, but don't want to color anyone else's opinions/ideas about what is going on here.

I am interested in hearing from others what they think might explain these results.

Thanks for any help/ideas.

[bleys21]

Do both motherboards run the RAM at the same speed? I tend to overclock stuff, so I have run into issues with RAM that ran fine on one board (at a lower speed, due to mb limits) but failed in my main board at higher speeds.

Also, could be voltage related, although that wouldn't explain why it ran fine for years without problem, unless either the RAM or motherboard is getting tired. Are slots 1/3 and 2/4 running on different channels (i.e with separate memory controllers?)

Last question...which set of RAM failed? Was it the original set you bought, or the newer set?

[fullonshred]

Do both motherboards run the RAM at the same speed? I tend to overclock stuff, so I have run into issues with RAM that ran fine on one board (at a lower speed, due to mb limits) but failed in my main board at higher speeds.

Also, could be voltage related, although that wouldn't explain why it ran fine for years without problem, unless either the RAM or motherboard is getting tired. Are slots 1/3 and 2/4 running on different channels (i.e with separate memory controllers?)

Last question...which set of RAM failed? Was it the original set you bought, or the newer set?

Thanks for your reply.I think the current test system (Intel based) may be running it slower. Original system is AMD. I was never overclocking this ram though. It is ddr3 1600. I think intel maybe is running it at 1066. Will check tomorrow.

The slots are dual channel. Will have to check out controller situation tomorrow. IIRC the good pair of dimms ran clean in both channels but it was a while back so not 100% sure.

Have no way of telling which set of dimms are giving problems. Ram is identical and I have pulled them all and put back in multiple times so could even be a mix from both kits.

I was recently given a different brand of 2x4=8gb ddr3 1600 kit. I may stuff it into the two empty slots and see what happens. Will do a fresh system image first though just in case.

[t100d]

I can't remember what DDR3 means …

[bleys21]

Yeah, in that case, you may want to test the new RAM in the "bad" slots, and then test the shizzle out of it. If those fail as well, you've got a motherboard problem. If they're fine, at least one of those other DIMMS is bad, and you'll probably have to go back to square one to figure it out. Or, if you need another machine, take the bad RAM and put it in that other motherboard running at the slower clock rate, and it will likely work just fine...

Do both motherboards run the RAM at the same speed? I tend to overclock stuff, so I have run into issues with RAM that ran fine on one board (at a lower speed, due to mb limits) but failed in my main board at higher speeds.

Also, could be voltage related, although that wouldn't explain why it ran fine for years without problem, unless either the RAM or motherboard is getting tired. Are slots 1/3 and 2/4 running on different channels (i.e with separate memory controllers?)

Last question...which set of RAM failed? Was it the original set you bought, or the newer set?

Thanks for your reply.I think the current test system (Intel based) may be running it slower. Original system is AMD. I was never overclocking this ram though. It is ddr3 1600. I think intel maybe is running it at 1066. Will check tomorrow.

The slots are dual channel. Will have to check out controller situation tomorrow. IIRC the good pair of dimms ran clean in both channels but it was a while back so not 100% sure.

Have no way of telling which set of dimms are giving problems. Ram is identical and I have pulled them all and put back in multiple times so could even be a mix from both kits.

I was recently given a different brand of 2x4=8gb ddr3 1600 kit. I may stuff it into the two empty slots and see what happens. Will do a fresh system image first though just in case.

[fullonshred]

Yeah, in that case, you may want to test the new RAM in the "bad" slots, and then test the shizzle out of it. If those fail as well, you've got a motherboard problem. If they're fine, at least one of those other DIMMS is bad, and you'll probably have to go back to square one to figure it out. Or, if you need another machine, take the bad RAM and put it in that other motherboard running at the slower clock rate, and it will likely work just fine...

Do both motherboards run the RAM at the same speed? I tend to overclock stuff, so I have run into issues with RAM that ran fine on one board (at a lower speed, due to mb limits) but failed in my main board at higher speeds.

Also, could be voltage related, although that wouldn't explain why it ran fine for years without problem, unless either the RAM or motherboard is getting tired. Are slots 1/3 and 2/4 running on different channels (i.e with separate memory controllers?)

Last question...which set of RAM failed? Was it the original set you bought, or the newer set?

Thanks for your reply.I think the current test system (Intel based) may be running it slower. Original system is AMD. I was never overclocking this ram though. It is ddr3 1600. I think intel maybe is running it at 1066. Will check tomorrow.

The slots are dual channel. Will have to check out controller situation tomorrow. IIRC the good pair of dimms ran clean in both channels but it was a while back so not 100% sure.

Have no way of telling which set of dimms are giving problems. Ram is identical and I have pulled them all and put back in multiple times so could even be a mix from both kits.

I was recently given a different brand of 2x4=8gb ddr3 1600 kit. I may stuff it into the two empty slots and see what happens. Will do a fresh system image first though just in case.

Checked specs of the AMD Motherboard and with 16gb of ddr3 it downclocks from 1333mhz to 1066mhz, same speed the Intel system was running with each dimm as I tested them for many hours. Zero errors. I noticed that CPUZ showed I was in single channel mode in the HP AMD Desktop when it should have been Dual channel with the 2 good dimms. Could not find verification online but vaguely recalled that this MB maybe you don't skip a space for Dual Channel. First 2 slots are blue, second 2 slots are black, and when I moved the 2 "good" dimms both into Blue slots they now read as Dual Channel in CPUZ. Also noticed that 1 dimm was dual rank memory and the other was single rank. So apparently though the model number, name, heat sinks and everything looked identical one set was dual rank and 1 set was single rank. So after testing them I pulled out and marked the 2 "good" dimms, and as I was looking closely now I could see that 2 of the 4 (the dual rank dimms) were a little thicker, and with a flashlight I could see under the heatsinks that the thicker dimms indeed had chips on both sides (dual rank).

So - popped in all 4 dimms making sure to keep the single rank dimms in the black slots and the dual rank dimms in the blue slots. Have been running Memtest86+ for 6.5 hours and Zero errors so far. Will let it run all night and part of tomorrow and see what it shows in the afternoon. Hoping maybe I pulled a couple of dimms and got the 2 sets mixed in together while I was moving video cards around and got ranks in wrong slots and maybe that had something to do with the problem. Even if Memtest runs 18 hours perfectly, the real test will come as I run the system over several days. Time will tell.

I appreciate your suggestion about the memory possibly running at different speeds in the intel test system. That got me looking at CPUZ info and that showed me I was in single channel mode in AMD system and that I had a (working) pair of "mismatched" dimms, one being single rank and the other being dual rank. I hope maybe this is the key to solving the system problems and the "why" of my getting errors when testing the memory in the original HP AMD powered Desktop.

[LightWingStudios]

Updated your Board Bios to the latest?

[fullonshred]

Updated your Board Bios to the latest?

Yes. This is an old, but still quite functional system. Last BIOS Update is listed as 2013.

After 16.5 hours of Memtest86+ I now show 3 errors, Pass 3. Read somewhere Memtest86+ needs 8 passes to be conclusive. At this rate it will take another full 24 hrs.

May just test the 2 single rank dimms and if they pass muster stick with 8gb total. It runs fine on 8gb, but MOAR is better. : )

[bleys21]

Yeah, mixing and matching RAM has always caused me problems, especially on the server platforms, where they seem to be even more picky about RAM than the desktops.

Bummer you're still seeing some memory errors...I really thought that getting the pairs matched up would do the trick. Could be compatibility issues with the motherboard though...some mb's handle single and dual rank DIMMs differently. I know I've run into this on servers before, where if you use dual, you lose available slots. Not sure about consumer mb's, I installed my 16GB so long ago I don't even remember what the mb manual said about it, but then again, I only have 2 - 8 GB DIMMs installed, so I may not have triggered the issue...

[fullonshred]

Yeah, mixing and matching RAM has always caused me problems, especially on the server platforms, where they seem to be even more picky about RAM than the desktops.

Bummer you're still seeing some memory errors...I really thought that getting the pairs matched up would do the trick. Could be compatibility issues with the motherboard though...some mb's handle single and dual rank DIMMs differently. I know I've run into this on servers before, where if you use dual, you lose available slots. Not sure about consumer mb's, I installed my 16GB so long ago I don't even remember what the mb manual said about it, but then again, I only have 2 - 8 GB DIMMs installed, so I may not have triggered the issue...

Well the MB and computer ran great with all 4 dimms installed from December of 2015 up until about 6-8 weeks ago. At first I thought the problem was caused by a W10 Update (June Monthly IIRC) but all my other systems handled it just fine including an AMD system built from that very same time frame with same iteration of CPU and Gskill DDR3 Ripjaws 1333mhz memory installed. So probably never was the W10 update at all. To confuse issues even more I had also swapped in a newer larger graphics card roughly in the same time frame.

I may just test the 2 single rank dimms and if good go with them. I have read that all other things being equal, single rank dimms are a little faster. Probably not enough to tell the difference without synthetic testing though.

If memory testing didn't take so exceptionally long it would be a simpler matter.

[bleys21]

Oh, what kind of video card? Some of the high end ones pull some significant wattage from the PS. When I got a 1080ti card, I had to upgrade the PS from a 650w to a 1000w, as I was getting weird instability issues. And in theory, a 650w should be fine, and that PS is still working great in my son's computer with a 1660ti card. I'm wondering if the motherboard just isn't getting quite enough power to run all 4 DIMMs now with the new video card...

And yeah, memory/burn in testing sucks. I tend to overclock my stuff, so I use Prime95 to stability test the CPU (I've got an i7-7770k overclocked from 4.2 to 5ghz) Typically have to let it go 24 hours to make sure you don't have problems. Video card testing is a bit easier with the benchmark utilities....they tend to beat the shizzle out of the video card, and find any instability pretty quick

[fullonshred]

Oh, what kind of video card? Some of the high end ones pull some significant wattage from the PS. When I got a 1080ti card, I had to upgrade the PS from a 650w to a 1000w, as I was getting weird instability issues. And in theory, a 650w should be fine, and that PS is still working great in my son's computer with a 1660ti card. I'm wondering if the motherboard just isn't getting quite enough power to run all 4 DIMMs now with the new video card...

And yeah, memory/burn in testing sucks. I tend to overclock my stuff, so I use Prime95 to stability test the CPU (I've got an i7-7770k overclocked from 4.2 to 5ghz) Typically have to let it go 24 hours to make sure you don't have problems. Video card testing is a bit easier with the benchmark utilities....they tend to beat the shizzle out of the video card, and find any instability pretty quick

I moved in an old RX480 for a bit. Then swapped it out for a smaller GTX960 4gb model. It was a heckuva squeeze getting the 480 in that factory mid-tower with a micro-ATX board.

I have moved the 2 single rank dimms to slots 3 and 4 and have been running just them in Memtest86+ for 17 hrs (6 full passes) with Zero errors so I think they are solid in slots 3&4. Will switch them to slots 1&2 later and run same dimms overnight and see what happens. If I get errors then I will maybe have to figure there is something wrong with slots 1&2.

WRT the PSU, I have considered that. It is an Antec Earthwatts 500 that should have enough power for either card, but that PSU does have some years on it so it could maybe be a contributing factor.

[bleys21]

Oh, ok....when you said new video card, I was thinking of the situation I ran into, with a new high power-draw card. I don't think the RX480 should give you too much trouble, although the power draw max of that card is 150w. Not sure what CPU model you're running, but unless its a high end one, a 500w PS should be fine.

It is starting to sound like a bad memory controller for slots 1/2. If your test of the ram in those slots fails, I think you've got it nailed, and then at that point you're either happy with 8GB, or looking at a new motherboard...

[fullonshred]

Oh, ok....when you said new video card, I was thinking of the situation I ran into, with a new high power-draw card. I don't think the RX480 should give you too much trouble, although the power draw max of that card is 150w. Not sure what CPU model you're running, but unless its a high end one, a 500w PS should be fine.

It is starting to sound like a bad memory controller for slots 1/2. If your test of the ram in those slots fails, I think you've got it nailed, and then at that point you're either happy with 8GB, or looking at a new motherboard...

Tested in pairs, both pairs have passed 7+ tests of Memtest86+ in both pairs of slots. The test of all 4 dimms installed only threw 3 errors in 7 or 8 passes. I may just install all 4 again properly pairing the single ranks in one pair and the dual rank sticks as a pair and see what happens over time. I have this system backed up and I am not adding much of anything new to it these days other than Windows and AV updates. I could throw in a newer PSU but it is a little aggravating in that tight case.

Currents PSU is an Antec Earthwatts 500 with strong 12v rails and has been reliable with more demanding hardware than what is currently installed in this computer. But - it is probably around 10 years old.

[bleys21]

Yeah, give it a shot if you're all backed up, and are not really worried about the system. Worst case, if it gets cranky, yank 2 of the DIMMs and you're probably gold at 8GB...

[fullonshred]

[mention]bleys21[/mention] I appreciate your company and suggestions along this tedious and somewhat unnecessary journey. Your participation made it easier for me to keep going.

Two days ago I decided to try 1 last combo. Both dual rank dimms in channel 1, and both single rank dimms in channel 2.

48:22 later the system completed 9.5 passes in Memtest86+ with zero errors. I am calling it "good to go" unless/until further problems arise.

[bleys21]

Awesome! That's great news, sounds like a compatibility issue with the two sets of sockets, so sounds like you found the right combo...