Filling all 4 DIMM slots causes system crashes

[Palindrome]

Hi all, new user here. Been lurking for a few months though.

I recently assembled a PC at work, and it's been incredibly unhappy when I fill all 4 DIMM slots with memory modules. Populating all 4 slots causes the system to stutter, freeze, video output to cut in and out intermittently, and usually ends in the PC just restarting itself (I'm assuming it BSODs, however I cant verify this, since it happens when the video cuts out, and the next time I get video out, it shows the system at the motherboard splash screen). The CPU works completely fine and functions as expected when stresstested (Cinebench R20 on loop), same goes for the GPU (Unigine Heaven on loop). My main suspicion therefore is the memory, since it is not a 4x4GB kit, but rather 4 sticks of the same memory modules (essentially 4 "kits" of 1 stick each). They are the exact same model with the same frequencies, latencies and voltages across the board. Could it be that one stick is "bad"? Or that this 4x4 mix and match configuration doesnt work, despite the fact that the modules are, for all intents and purposes, identical to each other?

What I also found suspicious is that when booting into Windows and getting into task manager with 4 modules, it only shows 1.5GB of RAM in use after startup, which is suspiciously low imo. Booting into windows and checking task manager with 2 modules yields roughly the same amount of memory used as well. Open to any feedback and suggestions on how I might solve this issue, thank you in advance. Specs of the PC in question are as follows:

Ryzen 7 2700 non X, PBO Enabled, no other BIOS tweaks done
4x4GB Corsair ValueRAM 2400MHz (DOCP/XMP disabled)
Asus Prime B450M-A
MSI GTX1650 LP OC
Samsung 970 Evo Plus 250GB
SeaSonic 500W Bronze PSU (don't have the exact model on hand)

 

[Colddecked]

You'll have to test the ram and ram slots individually, see if one of them is bad. If they all seem good, you'll probably have to raise certain voltages... wish I could tell you off the top of my head which ones, i'm sure a memory overclocking guru will chime in.

You shouldn't have any issues with the ram not being a "kit" if they're all the same serial/part number that's perfect.

 

[HD64G]

SOC voltage is the main one to raise when populating 4 DIMM slots even at stock settings.

 

[maxfly]

You can also try bumping the memory voltage by as much as .50v (conservatively) to try and stabilize them. Im assuming stock voltage is 1.2v or 1.35v? If so 1.3 or up to1.45(if 1.35v is stock)wont hurt them.
If that doesn't work i would start by running 2x4gb in the primary slots at the rated speed, voltage and timings(with xmp, dbl check that they are set up correctly). Test them with memtest86+ or your most trusted memory testing program. If those check out as good, run the next 2x4gb the same way. If all 4 test as good you may have a bad dimm. If not you simply need to pull a module and retest until you find the bad one and rma it.
If all modules test good- Test the dimms (with the same single module)one by one with memtest. Until you get errors, or not.
If none of the above works or shows errors you may have bent cpu socket pins, paste in the socket or just a weak imc.

 

[Palindrome]

Thanks for all the suggestions so far, I will try those at work tomorrow. As for the CPU, it was an returned sample from a customer (bent pins, surprise surprise), however I repaired it. All pins intact, all as straight as I could get them. It doesn't slide in seamlessly, but just a light nudge and it slots into place. Will report back with my findings tomorrow, cheers all

 

[GorbazTheDragon]

the more sticks and the more ranks per channel the higher the memory controller load and therefore the higher the sensitivity to misconfigurations (by the motherboard/bios) or other problems (damaged traces or bad contacts) and the easier those things will cause instability.

Early DDR4 is usually also very fiddly.

 

[Palindrome]

I've now tried manually setting the memory voltage to 1.35V (up from default 1.2V on these), and set the speed to a fixed 2400MHz, which is their rated speed. This yielded no improvement, I was going to try bumping up the SoC Voltage, but after a failed boot, I went back into safe mode and was greeted by the following error message popup:
WerFault.exe - Error in Application "The application in 0x00007FFDEA3E456 pointed to memory at 0x00000000001802DC8AC. The process read could not be executed in memory. Press OK to solve this issue."
Could this be further confirmation of a bad DIMM? I guess my only choice is to run memtest now, will report back with those results.

 

[Palindrome]

MemTest86 found no errors with that 4x4GB configuration, it passed after around 4h with 0 errors. Fortunately I was able to find a different set of memory which also passed MemTest, and it also seems to work flawlessly in Windows. Very odd behaviour, but hey at least it works now.

 

[ThrashZone]

Hi,
Four hours is not enough time to test all 16gb
It should be allowed to run at least 6 passes with no errors at what ever speed you're having issues with.

 

[The red spirit]

SOC voltage is the main one to raise when populating 4 DIMM slots even at stock settings.

But it's just 2400 MHz memory, nothing super fast or anything. AMD website says that 2700 supports memory up to 2933 MHz. I'm assuming that it's at JEDEC spec too. As long as AMD is saying that, JEDEC spec RAM with frequency no higher than that should work perfectly fine without messing with overvolting. I would rather conclude that CPU's memory controller is defective.

 

[freeagent]

Did you clear the cmos at all when switching from 2x2 to 2x4 and back? Internal timings change when going to 4 sticks. If you neglected that step, it could be the source of your problems.

 

[Palindrome]

Thanks for the feedback, was not aware of either. I simply didn't have enough time to run multiple passes of MemTest, and no I did not think to clear the CMOS. That could've very probably helped.

 

[eidairaman1]

Get away from corsair

 

[Palindrome]

Get away from corsair

My own kit of Vengeance LPX at home has been running flawlessly for years now. I do avoid their coolers, fans and peripherals like the plague though

 

[Solid State Soul ( SSS )]

filling all dimm slots increases the load on the memory controller and reduce overclocking headroom resulting in problems with some high frequency XMP kits

 

[freeagent]

filling all dimm slots increases the load on the memory controller and reduce overclocking headroom resulting in problems with some high frequency XMP kits

Yes and no, depends on the CPU I think. My 5600X with 4 sticks can only do 1833 stable. With 2 sticks it does 2K 1:1 stable. My 5900X is not stable at 2K with 2 sticks, and does 1900 with 4 sticks. Anything over 1900 is unstable with 2 or 4 sticks on that CPU. My 3600XT wouldn't do over 1900 with 2 or 4 sticks neither. Looks like there is a shitload of variance with AMD CPU's.. but my 5900X acts a little like my 3600XT.. so looks like they share some components as reported..

 

[_UV_]

Hi,
Four hours is not enough time to test all 16gb
It should be allowed to run at least 6 passes with no errors at what ever speed you're having issues with.

My experience tells that even 1 pass is enough and in some weird cases you may run memtest for whole 24h (about 10 passes on average if you have max RAM capacity for platform) and not get any kind of errors, but once you try to use such system you get BSOD or instability.

 

[eidairaman1]

My own kit of Vengeance LPX at home has been running flawlessly for years now. I do avoid their coolers, fans and peripherals like the plague though

Yeah you must of gotten a true Ryzen Kit that just works with DOCP, but most who come here dont know about the Ryzen Certified ram

The value ram may not be ryzen certed, plus you must have a 4 module kit. Go to Asus Website and get the highest bios possible, reseat all ram or check each module individually in each slot (16 different combos) also read the motherboard manual for the highest you can go with 4 modules present

You could have a bad motherboard too

 

[Shrek]

I'd try a BIOS update

PRIME B450M-A｜Motherboards｜ASUS USA

"Improve DRAM stability"
version 2202 and above

 

[Palindrome]

Yes and no, depends on the CPU I think. My 5600X with 4 sticks can only do 1833 stable. With 2 sticks it does 2K 1:1 stable. My 5900X is not stable at 2K with 2 sticks, and does 1900 with 4 sticks. Anything over 1900 is unstable with 2 or 4 sticks on that CPU. My 3600XT wouldn't do over 1900 with 2 or 4 sticks neither. Looks like there is a shitload of variance with AMD CPU's.. but my 5900X acts a little like my 3600XT.. so looks like they share some components as reported..

Memory controller load was a thought that crossed my mind. I find it odd though that the "unstable" 4x4 config would pass MemTest (at least once that is, did not test multiple runs), but would absolutely sh*t the bed once booted into Windows.

My experience tells that even 1 pass is enough and in some weird cases you may run memtest for whole 24h (about 10 passes on average if you have max RAM capacity for platform) and not get any kind of errors, but once you try to use such system you get BSOD or instability.

This was basically my issue. Was fine in MemTest (again only 1 pass, but still), but completely unstable once booted to Windows.

Yeah you must of gotten a true Ryzen Kit that just works with DOCP, but most who come here dont know about the Ryzen Certified ram

The value ram may not be ryzen certed, plus you must have a 4 module kit. Go to Asus Website and get the highest bios possible, reseat all ram or check each module individually in each slot (16 different combos) also read the motherboard manual for the highest you can go with 4 modules present

You could have a bad motherboard too

For my home PC? Yeah, it's possible I got a true Ryzen Kit, even though the box and product info made no mention of it; in fact, it only said "Intel XMP certified" or something along those lines. Regardless, it's been working flawlessly for years so I won't complain

As for the PC at work, I simply did not have time to test all those things you suggested. I was able to source a different kit of memory which has been working fine ever since (at least to my knowledge, I've not heard any feedback from them indicating otherwise).

I'd try a BIOS update

PRIME B450M-A｜Motherboards｜ASUS USA

"Improve DRAM stability"
version 2202 and above

Good point, I can't recall whether the BIOS was fully up to date. I've been updating loads of boards, hard to remember whether I also updated this particular one Regardless, the problem resolved itself since I was able to find a different kit that has been working well since.

Thanks everyone again for your help. Now I'm not sure whether I can manually close the thread or if I just let it sit as is. The forums I've been on usually have a mod or the OP close the thread, not sure if that's also the case here.

 

[TheoneandonlyMrK]

Hi all, new user here. Been lurking for a few months though.

I recently assembled a PC at work, and it's been incredibly unhappy when I fill all 4 DIMM slots with memory modules. Populating all 4 slots causes the system to stutter, freeze, video output to cut in and out intermittently, and usually ends in the PC just restarting itself (I'm assuming it BSODs, however I cant verify this, since it happens when the video cuts out, and the next time I get video out, it shows the system at the motherboard splash screen). The CPU works completely fine and functions as expected when stresstested (Cinebench R20 on loop), same goes for the GPU (Unigine Heaven on loop). My main suspicion therefore is the memory, since it is not a 4x4GB kit, but rather 4 sticks of the same memory modules (essentially 4 "kits" of 1 stick each). They are the exact same model with the same frequencies, latencies and voltages across the board. Could it be that one stick is "bad"? Or that this 4x4 mix and match configuration doesnt work, despite the fact that the modules are, for all intents and purposes, identical to each other?

What I also found suspicious is that when booting into Windows and getting into task manager with 4 modules, it only shows 1.5GB of RAM in use after startup, which is suspiciously low imo. Booting into windows and checking task manager with 2 modules yields roughly the same amount of memory used as well. Open to any feedback and suggestions on how I might solve this issue, thank you in advance. Specs of the PC in question are as follows:

Ryzen 7 2700 non X, PBO Enabled, no other BIOS tweaks done
4x4GB Corsair ValueRAM 2400MHz (DOCP/XMP disabled)
Asus Prime B450M-A
MSI GTX1650 LP OC
Samsung 970 Evo Plus 250GB
SeaSonic 500W Bronze PSU (don't have the exact model on hand)

The timings of the memory are setup for single not quad use, set timings to auto or manually set looser timings, plus possibly more volts.

 

[HD64G]

Have you connected anything to the front usb hub of the case? Why don't you try to disconnect that from the motherboard? Someone here in the forum recently solved his pc's very weird problems by doing that.

 

[eidairaman1]

The timings of the memory are setup for single not quad use, set timings to auto or manually set looser timings, plus possibly more volts.

Yup cas 9 on my kit makes it drop capacity

 

[P4-630]

My own kit of Vengeance LPX at home has been running flawlessly for years now.

Same here, no issues at all running 4 sticks, intel system here though...

 

[ThrashZone]

Hi,
Not sure I've seen many cases that a user is using four sticks and none were an actual kit not even 2 stick kits lol just four separate sticks bought and used :/