Memory errors running memtest86

[Bubblewrap]

This weekend I build a new system (pcpartpicker list). Bios is updated to the most recent version, as are all drivers and windows 11. 

 

CPU and GPU stresstests (Cinebench 23 and Heaven) were successful.

Running Passmark memtest86 gave 6 errors on test 8. MemTest86-Report-20231203-181232.html

I tried reseating the memory sticks and turning off XMP and running the test again. This gave no errors.

Turning on XMP again and running memtest86 gave 4 errors on test 8: MemTest86-Report-20231204-063729.html

 

As far as I understand, errors in Memtest86 should not happen at all. So far I've not had real issues other than Civ VI crashing twice, but I don't want it happen during work. So I'm wondering what to do. Since it doesn't give errors without xmp my conclusion is that there is no problem with the memory, but with the xmp. Could changing xmp settings solve the problem or should I just return these memory sticks and get new ones?

 

Edit:

PCPartPicker Part List

CPU: Intel Core i7-14700K 3.4 GHz 20-Core Processor  ($401.60 @ B&H) 
CPU Cooler: Alphacool Eisbaer Pro Aurora 69.98 CFM Liquid CPU Cooler  (Purchased For €185.00) 
Motherboard: MSI MEG Z690 UNIFY ATX LGA1700 Motherboard  (Purchased For €185.00) 
Memory: G.Skill Trident Z5 RGB 32 GB (2 x 16 GB) DDR5-6400 CL32 Memory  (Purchased For €130.00) 
Storage: Samsung 990 Pro 2 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive  (Purchased For €135.00) 
Video Card: Asus ProArt OC GeForce RTX 4070 12 GB Video Card  ($679.00 @ B&H) 
Case: Lian Li LANCOOL II-X ATX Mid Tower Case  (Purchased For €0.00) 
Power Supply: SeaSonic FOCUS GX-750 ATX 3.0 750 W 80+ Gold Certified Fully Modular ATX Power Supply  ($144.99 @ Amazon) 
Operating System: Microsoft Windows 11 Home OEM - DVD 64-bit  ($124.99 @ Amazon) 
Total: $1350.58
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2023-12-04 03:46 EST-0500

[RONOTHAN##]

10 minutes ago, Bubblewrap said:

This weekend I build a new system (pcpartpicker list).

The list is private. 

 

Without knowing what CPU/motherboard/RAM you're running, it's not really possible to say what the issue actually is. It could be you have an XMP that's unusably fast (I.E. DDR5 8400) or it could be a bad memory stick and you should test one stick at a time. Also, giving that info could point you in the direction a better memory stress test for your needs, as MemTest86 is not the most sensitive to memory errors out there and you might be able to get something that will error 6 times within a minute rather than 2-4 hours. 

[Bubblewrap]

Just now, RONOTHAN## said:

The list is private. 

 

Sorry, didn't know. I added the parts list in the starting post.

 

[RONOTHAN##]

18 minutes ago, Bubblewrap said:

Sorry, didn't know. I added the parts list in the starting post.

 

It's fine, this is more common than you'd think.

 

Alright, so it's not too aggressive of an XMP, that profile should work on every 14th gen CPU. There's 4 different things that I'd suspect can be the problem:

1. Bad memory stick
2. Bad motherboard
3. Bad CPU/memory mount
4. Bad BIOS revision. 

Ideally, it's the bad CPU/memory mount since those are the easiest to fix, though you seem to already have ruled out the memory mount and are left with the CPU (the less easy one to do). I'd start by trying to figure out if it's a bad stick or the motherboard that's the problem. Running each stick independently, check for stability in slots 2 and 4 to see if they're stable at XMP in them. If there's one memory stick that's consistently not working, it's a faulty stick and needs to be RMA'd. If it's one particular slot that consistently is having issues, it's either a bad motherboard or CPU mount and you should remount the CPU. If it's both sticks and both slots, it's probably a bad BIOS revision, in which case you have two options: manually try tuning voltages to get it to work or just running at JEDEC speeds for a month or so until MSI releases a better BIOS for 14th gen (if my Z690 Unify-X is similar to the standard Unify, the 14th gen BIOS revisions currently out are worse for memory support compared to some of the older BIOS revisions, though 6400 still should work without that much issue). 

 

Also, with 14th gen you will have some stress tests that are generally quicker to error than MemTest86. The two big ones are Y-Cruncher VT3 and TestMem5 with the 1usmus_v3 preset. VT3 more tests the memory controller than the memory sticks, though if the issue is related to the memory frequency at all this usually crashes within 5-10 seconds of turning it on. If the issue is more memory stick related, TestMem5 1usmus_v3 usually detects issues within 5-10 minutes, and this can detect some high-frequency related issues as well, though not as quickly as VT3 can. I'd start trying the 1usmus_v3 preset for 3 cycles, this will take 10 minutes and it should be a quick and dirty way to check for issues for the aforementioned individual RAM stick checks, though it might also be worth checking VT3. 

[DoctorNick]

1 hour ago, Bubblewrap said:

This weekend I build a new system (pcpartpicker list). Bios is updated to the most recent version, as are all drivers and windows 11. 

 

CPU and GPU stresstests (Cinebench 23 and Heaven) were successful.

Running Passmark memtest86 gave 6 errors on test 8. MemTest86-Report-20231203-181232.html

I tried reseating the memory sticks and turning off XMP and running the test again. This gave no errors.

Turning on XMP again and running memtest86 gave 4 errors on test 8: MemTest86-Report-20231204-063729.html

 

As far as I understand, errors in Memtest86 should not happen at all. So far I've not had real issues other than Civ VI crashing twice, but I don't want it happen during work. So I'm wondering what to do. Since it doesn't give errors without xmp my conclusion is that there is no problem with the memory, but with the xmp. Could changing xmp settings solve the problem or should I just return these memory sticks and get new ones?

 

Edit:

PCPartPicker Part List

CPU: Intel Core i7-14700K 3.4 GHz 20-Core Processor  ($401.60 @ B&H) 
CPU Cooler: Alphacool Eisbaer Pro Aurora 69.98 CFM Liquid CPU Cooler  (Purchased For €185.00) 
Motherboard: MSI MEG Z690 UNIFY ATX LGA1700 Motherboard  (Purchased For €185.00) 
Memory: G.Skill Trident Z5 RGB 32 GB (2 x 16 GB) DDR5-6400 CL32 Memory  (Purchased For €130.00) 
Storage: Samsung 990 Pro 2 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive  (Purchased For €135.00) 
Video Card: Asus ProArt OC GeForce RTX 4070 12 GB Video Card  ($679.00 @ B&H) 
Case: Lian Li LANCOOL II-X ATX Mid Tower Case  (Purchased For €0.00) 
Power Supply: SeaSonic FOCUS GX-750 ATX 3.0 750 W 80+ Gold Certified Fully Modular ATX Power Supply  ($144.99 @ Amazon) 
Operating System: Microsoft Windows 11 Home OEM - DVD 64-bit  ($124.99 @ Amazon) 
Total: $1350.58
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2023-12-04 03:46 EST-0500

Try disabling XMP and set memory frequency, dram voltage and primary timings manually in BIOS and test again. Could just be the profile, that's not working. 

[leclod]

Maybe a contact frame would improve the cpu contacts.

You could just loosen the frequency or a timing.

Edited December 4, 2023 by leclod

[Bubblewrap]

1 hour ago, leclod said:

Maybe a contact frame would improve the cpu contacts.

You could just loosen the frequency or a timing.

I've got a thermalright contact frame installed. Didn't yet put that on the list.

 

1 hour ago, DoctorNick said:

Try disabling XMP and set memory frequency, dram voltage and primary timings manually in BIOS and test again. Could just be the profile, that's not working. 

As in just using the same settings the XMP uses? (trying to make sure I understand correct)

 

2 hours ago, RONOTHAN## said:

Running each stick independently, check for stability in slots 2 and 4 to see if they're stable at XMP in them.

I tried both sticks independently in slots 2 and 4, with xmp enabled. Both sticks in both slots pass the memtest86 test 8. (due to lack of time I just let that one run for now, since that was the one giving errors). Trying both sticks again, but in reverse placement from original test again gives 24 errors. 

 

2 hours ago, RONOTHAN## said:

manually try tuning voltages to get it to work

With xmp DRAM voltage is set to 1.400V. Is it smart to go higher than that? VDDQ, is 1.400V as well, VPP is 1.800V

 

2 hours ago, RONOTHAN## said:

the 14th gen BIOS revisions currently out are worse for memory support compared to some of the older BIOS revisions, though 6400 still should work without that much issue

Would going back to an older bios solve the errors?

[Bubblewrap]

I tried increasing dram voltage to 1.410V. it seems to have worked since both test 8 from memtest86 and TestMem5 with the 1usmus_v3 preset dont give any errors. I'll do some more testing later when i have more time, but Memory sticks are probably not broken i guess?

[RONOTHAN##]

2 hours ago, Bubblewrap said:

I tried both sticks independently in slots 2 and 4, with xmp enabled. Both sticks in both slots pass the memtest86 test 8. (due to lack of time I just let that one run for now, since that was the one giving errors). Trying both sticks again, but in reverse placement from original test again gives 24 errors. 

Alright, that's just weird. 

 

2 hours ago, Bubblewrap said:

I've got a thermalright contact frame installed. Didn't yet put that on the list.

 

These can cause memory issues in some circumstances and certain boards. I've got one on my Unify-X without any issues, so I doubt it would cause issues, though just for troubleshooting I might put it back to stock and check.

 

2 hours ago, Bubblewrap said:

With xmp DRAM voltage is set to 1.400V. Is it smart to go higher than that? VDDQ, is 1.400V as well, VPP is 1.800V

 

The voltages I was referring to is more the IMC voltages rather than the DRAM voltages. You can set those higher (1.55V is fine for long term use if SK Hynix is to be believed), though IMO anything above 1.43V should have active cooling on those G.Skill kits due to their fairly inadequate heat spreaders. With the IMC voltages though, this isn't a problem, and those voltages are what actually help with stability on DDR5 setups. The big three are CPU VDDQ, CPU VDD2, and System Agent (I think it's called SA voltage, though I would need to head into the BIOS and check what MSI calls it). CPU VDDQ I'd want to keep below 1.45V, VDD2 I'd want to keep at or below 1.65V, and system agent I'd want to keep at or below 1.45V, though all of these sweet spot so it's unlikely you'd get up that high anyway. CPU VDDQ is usually best set around 1.35V, though from my testing of the more recent BIOS revisions having it a bit lower can be beneficial (1.25-1.3V was ideal on my setup). VDD2 is usually best a bit higher between 1.4 and 1.5V, though at just 6400 you really shouldn't need to have it at 1.5V+. System Agent is very much CPU dependent, though I'd have a hard time believing that 1.2V doesn't work at this frequency. 

 

Increasing the actual memory voltage can in theory help if it was a timing issue, though that's unlikely. If you want to give it a shot though feel free. 

 

2 hours ago, Bubblewrap said:

Would going back to an older bios solve the errors?

In theory yes, though those older BIOS revisions don't have support for the 14700K, so you're stuck on one of the most recent two updates. 

[Bubblewrap]

22 minutes ago, RONOTHAN## said:

Alright, that's just weird. 

I'm no expert, but that's what I was thinking. If both sticks in both slots individually work fine, then why is it a problem with both sticks together. 

 

23 minutes ago, RONOTHAN## said:

These can cause memory issues in some circumstances and certain boards. I've got one on my Unify-X without any issues, so I doubt it would cause issues, though just for troubleshooting I might put it back to stock and check.

I'm still looking for another case because cooling in the lancool II is far from optimal. If I'm moving everything to another case perhaps it's worth a try.

 

25 minutes ago, RONOTHAN## said:

The voltages I was referring to is more the IMC voltages rather than the DRAM voltages. You can set those higher (1.55V is fine for long term use if SK Hynix is to be believed), though IMO anything above 1.43V should have active cooling on those G.Skill kits due to their fairly inadequate heat spreaders. With the IMC voltages though, this isn't a problem, and those voltages are what actually help with stability on DDR5 setups. The big three are CPU VDDQ, CPU VDD2, and System Agent (I think it's called SA voltage, though I would need to head into the BIOS and check what MSI calls it). CPU VDDQ I'd want to keep below 1.45V, VDD2 I'd want to keep at or below 1.65V, and system agent I'd want to keep at or below 1.45V, though all of these sweet spot so it's unlikely you'd get up that high anyway. CPU VDDQ is usually best set around 1.35V, though from my testing of the more recent BIOS revisions having it a bit lower can be beneficial (1.25-1.3V was ideal on my setup). VDD2 is usually best a bit higher between 1.4 and 1.5V, though at just 6400 you really shouldn't need to have it at 1.5V+. System Agent is very much CPU dependent, though I'd have a hard time believing that 1.2V doesn't work at this frequency. 

 

Increasing the actual memory voltage can in theory help if it was a timing issue, though that's unlikely. If you want to give it a shot though feel free.  

I misunderstood then. I think found the voltages you mention. 

[RONOTHAN##]

10 minutes ago, Bubblewrap said:

I misunderstood then. I think found the voltages you mention. 

Yeah, those are the ones I referred to, CPU SA, CPU VDDQ, and CPU VDD2. Technically the PLL voltages can also help, but I've never seem them do anything at speeds below 7200 so I highly doubt they will do anything. 

 

12 minutes ago, Bubblewrap said:

I'm no expert, but that's what I was thinking. If both sticks in both slots individually work fine, then why is it a problem with both sticks together. 

 

Just realized I read that wrong. I thought you tried a single stick in slot 2 that works, the other stick individually in slot 4 that works, yet doing it vice versa doesn't work. The issue you're having is not nearly as weird (to be clear, it's still weird, but it's not unheard of), and it usually is the result of IMC instability. The weirdness of that comes in the fact that you're at such a low memory speed that really shouldn't be having IMC instability. 

 

If setting it to 1.41V for whatever reason fixes it, I'd just leave that and call it a day. I'd do a bit more stability testing just to make sure, probably an hour of VT3 and an hour of TestMem5. 

[Bubblewrap]

So far I've managed to get the ram working without errors at lower voltage. Learned a lot just following tutorials, but still a lot of the settings didn't really make sense to me yet. It seems to run quite ok now with:

 

DRAM Frequency at 6200MHZ (got errors when testing 6400, even at 1,5V)

DRAM voltage at 1.35V and DRAM VDDQ at 1.290V

CPU VDDQ at 1,250V and CPU VDD2 at 1,35V (but didn't try other voltages yet)

 

Primary timings are:

Command Rate: 2N

tCL: 30

tRCD: 38

tRCDW: 38

tRP: 38

tRAS:50

 

In the spoiler is a list of voltages and timings from the bios. I left out what I didn't change.

Spoiler

DRAM Setting
Extreme Memory Profile(XMP)		[Disabled]
DRAM Reference Clock	100	[100MHz]
CPU IMC : DRAM Clock	Gear2	[1/2 : 1 (Gear2)]
DRAM Frequency	31	[6200 G2 (62x1.00...)
Adjusted DRAM Frequency		6200MHz
Load Memory Presets		[Disabled]
Memory Try It!		[Disabled]
Memory Extension Mode		[Disabled]
DRAM Timing Mode		[Link]
Memory Fast Boot		[Disabled]
Memory Fast Boot When OC Fail		[Disabled]
Voltage Setting
CPU SA Voltage	1.260V	1.250
CPU VDDQ Voltage	1.260V	1.250
CPU VDD2 Voltage	1.356V	1.350
DRAM Voltage Mode		[ Unlink ]
DRAM DIMMA2 Voltage	1.350V	1.350
DRAM DIMMB2 Voltage	1.350V	1.350
DRAM DIMMA2 VDDQ Voltage	1.290V	1.290
DRAM DIMMB2 VDDQ Voltage	1.290V	1.290
ADVANCED DRAM CONFIG
Main timing configuration
Command Rate	2N	[ 2N ]
tCL	30	30
tRCD	38	38
tRCDW	38	38
tRP	38	38
tRAS	50	50
tRFC1	368	Auto
tRFC2	368	368
Sub timing Configuration
tRFCPB	240	240
tREFI	65024	65024
tWR	48	48
tWR_MR	48	48
tRRD	8	8
tRRD_L	12	12
tRTP	12	12
tRTP_MR	12	12
tFAW	32	32
tCWL	28	28
tCKE	10	10
Turn around timing config
Turn Around Timing Setting Mode		[ Fixed Mode ]
tRDRDSG	14	14
tRDRDDG	8	8
tWRWRSG	28	28
tWRWRDG	8	8
tRDWRSG	22	22
tRDWRDG	22	22
tWRRDSG	56	56
tWRRDDG	46	46
Advanced Timing Configuration
tXP	8	8
Misc Item
Safe boot retry		enabled
Enhanced interleave		enabled
Power down control
Power down mode		disabled
PDWN Idle counter	32	auto
APD	0	0
PPD	0	0
Global PD	0	0

Though I learned a lot from following tutorials, there are a lot of settings I just kinda copied. So if someone likes to take a look to see improvement, I'd love to hear them.

 

I still have to do some longer testing, but testmem5 with usmus_v3 for about 15 minutes doesn't give any errors while the ram stays below 50 degrees Celsius (something it did not do with 1,41V). Also still have to take a look at CPU settings, but for now I'm happy the ram is not giving any errors.