Memtest giving false positives?

[Roxxors]

I have 128Gig of mem and run memtest. over 4 passes. I get 1,2, maybe 3 errors. Never the same address, never the same core, never the same data, and never the same test. Each different error passes the other 3 runs. It's more like a spurious inconsistent error. The system works fine and passes all other tests and benches. I know that any software can give false readings no matter how good it is. Is this a memtest issue and can I ignore these errors as false? (I tried to edit the last post to add this but it wouldn't let me so I posted this one and you can delete that older one)

[David (PassMark)]

See this page
https://www.memtest86.com/troublesho...htm#consistent

Random errors can be real errors.

A simple example is low voltage to the entire RAM stick. The whole stick gets impacted. So the error location is random.

[Roxxors]

I tried to simplify that post. I have run like 6-4 pass runs, that's 24 in total and wrote the errors. None are the same, so that means that each fail has passed 23 other runs... This can't be a hardware issue, at least that is not what it looks like to me. It's possible that it's the MB ME version (I have the latest), but it could be a new version will fix it. I am just trying to narrow it down.

[David (PassMark)]

You haven't posted any details about your hardware, so don't know the setup.

128GB is a lot of RAM, at least for a home system. There have been various stability issues over the last couple of years with motherboards that have fully populated RAM slots. Especially the AMD AM5 platform.

There is a good video on the issue here
https://www.youtube.com/watch?time_continue=3&v=P58VqVvDjxo&embeds_referri ng_euri=https%3A%2F%2Flinustechtips.com%2F

To be honest it is a bit of a mess. As it should either fully work at full speed, or gracefully fail. Having is half work, some of the time, creates a lot of grief.

[Roxxors]

Build:

Asus ROG Maximus Z890 Extreme MB, Intel Ultra 285K, Corsair Titan 360 RX Liquid CPU cooler, 128GB Corsair Vengeance 6000 CL30 (4 sticks), 2-Corsair 700 4TB PCIE NVME drives in a Raid 0 config. RTX 4070 (waiting for a 5090 when I can find one), SB AE-9 Sound card, Corsair HX1500 watt PS. Corsair 7000D Airflow case. Win 11 Latest build.

In normal operation the system uses about 40Gig of memory and the CPU runs at about 12% load.

Little more info if it helps, The P-Cores are 5.7G, E-Cores 5.0G, and the CPU Cache ratio is 40, and memory speed is 5600.
(At 5800 the system randomly locks, and some programs won't run, but it is stable at 5600 and runs fine with everything I can throw at it).

One other bit of info. As above the Ps run at 5.7 and the Es at 5.0 but Memtest reports on the CLK/Temp line at the top the speed is 3.752. What exactly is that reporting? (Ver 10.7 reported 5803, when I upgraded to 11.2 it changed).

[David (PassMark)]

At 5800 the system randomly locks, and some programs won't run, but it is stable at 5600 and runs fine with everything I can throw at it

Except that you threw Memtest86 at it and it wasn't stable
Try 5400.

the top the speed is 3.752.

CPU clock speed at the time of measurement. It is normal that it varies. But it should got higher under load, and higher again under single core load.
MemTest86 does try and load the CPU to measure the max clock. But clearly that didn't happen here.

[Roxxors]

(Ver 10.7 reported 5803, when I upgraded to 11.2 it changed).

I was adding this to my post at 8:02 when you replied at 7:57. What changed between versions 10.7 and 11.2.?

And one other thing. Memtest seems to be running in "minimal mode" . There used to be a splash screen with a menu and after a 10 sec countdown the test would run and when finished you could scroll through the results using the mouse. That is not happening anymore after updating to 11.2. It boots off USB, loads and runs and puts up the Pass/Fail window and you have to restart... No mouse or any results. Is there a way to force it to boot to the graphical interface?

[Roxxors]

More interesting stuff. I got the Interface back but had to unplug monitor 5 and force it to run on monitor 1 off the video card. Monitor 5 was on the internal graphics. I ran again, got 5 errors all on different sticks and addresses yet the system still runs fine and passes all stress tests and benches. It's just so odd... But as long as it's stable and runs I am good, but I would really to not see errors (real or not) so I will keep digging.

I really want to see what a new ME FW update does since this MB is pretty new.

[David (PassMark)]

MemTest86 change list can be found here. I don't think there were any changes that would effect the measurement of the CPU clock speed. At lot of code was added to display the currently active memory clock speed however.

[Roxxors]

Ver 10.7 reported 5803, when I upgraded to 11.2 it changed to 3752 and V10.7 didn't display the CPU temp, 11.2 does, so they changed something. It also broke the graphics interface when using the built-in graphics, to get it to show up you have to use the video card.

I can only tell you WHAT it is doing, not WHY it's doing it... The WHY is what I am troubleshooting to try to find out. Don't worry. I also have tickets in with Intel and Asus and Corsair to try and figure this out. As soon as I hear back, I can update info here.

[David (PassMark)]

I also have tickets in with Intel and Asus and Corsair

If you get any reply at all, it will just be general canned reply.

If the graphics aren't working it is likely your UEFI BIOS. Maybe it doesn't deal with 5 monitors very well. Likely ASUS never tested this.

You might be shooting the messenger here.

[Roxxors]

Possible, but as you should know, all possibilities have to be checked and eliminated one by one.

I am actually leaning toward the ME FW atm... But we shall see where all this troubleshooting and questions lead till, I get it pinpointed.

[Roxxors]

Welp. Just so you and anyone else that reads this knows. This MB (Asus ROG Maximus Z890 Extreme ) doesn't seem to like having all 4 DIMM slots populated. I had 4-32Gs in, pulled 2 out (just left the 2 default dimm slots in) and things are much better. The 6000 cl30s run at 6333 just fine. Still not sure why Memtest won't report the correct CPU speed, but that can wait.

And before you ask, I made sure the 2 sticks were good, replaced them with the other 2 sticks, still good, put all 4 in and broke again. So it's not the amount of ram, it's the number of dimm sockets filled. Who would have thunk that...

Still eyeing the ME FW as the main issue... *Glares at the ME FW*

[David (PassMark)]

Thanks for the update.

Likely it is just some permutation that doesn't work. (Voltage levels, timings, power usage of the sticks, RAM clock speeds, etc..). But the motherboard manufacturer should be testing all this for you, so you don't have to.

I just had a look at the ASUS web site QVL. Your RAM isn't on the support list. Looks like they only tested 4x24GB and 4x16GB for Corsair 6000
https://rog.asus.com/motherboards/ro...sk_qvl_memory/

And even if they did test it, they might have missed the fault, as it seems to be almost working.

[Roxxors]

At this point I think I can rule out Memtest as an issue, but you still should fix the CPU speed reporting at some point... Even with the QVL they only tested it at stock speeds. I never run at stock speeds... If you are not bleeding, you are not on the cutting edge...