I ran V11.7 with 2 passes and got 1 error in Test 11. I ran V11.7 with 3 passes and got 1 error in Test 8. With all previous version of MemTest86 I have never got an error with this system. There have been no changes in system hardware. If I just run 1 pass with V11.7, I get no error (I believe the first pass is a quicker test than the passes after pass 1). If I just run Test 11 or 8 by themselves, I get no error. I went back and ran MemTest86 V11.6 2 passes and get no errors. I have attached 2 zip files, 11_6 Good.zip with log files for the pass with V11.6 and 11_7 Fail.zip with log files for the failure with V11.7. Thanks
-
-
Correct. This was done to allow for a quicker initial test if your time was limited.I believe the first pass is a quicker test than the passes after pass 1
We didn't make any change to the actual tests between V11.7 and V11.6 that would cause this.
In a 4 hour test you got 1 error. This suggests it is pretty random. If that was the error rate (1 per 4 hours) then it is easy to imagine doing a 2 or 3 hour test with no errors.
This was the error:
2026-05-04 12:45:41 - [MEM ERROR - Data] Test: 8, CPU: 21, Address: AF06744EC/B2, Expected: F7A5AFAF, Actual: F725AFAF
So if you did a few more tests and the error stayed in the B2 slot, then that would be the bad stick.
-
After some quick testing it seems it may be due to setting the memory profile to XMP Tweaked (ASUS motherboard). I set it to XMP II and don't seem to be getting any errors now. Will do some further testing. Just that it takes a while since I need to run MemTest86 several hours for each test. ThanksComment
-
Yea. Maybe they should have called this XMP Unstable.XMP Tweaked
In their marketing departments desperate search for motherboard product differentiation they have made this worse not better.
As far as we know none of RAM manufacturers are testing with these settings (that is to say settings that are more aggressive than the advertised specs). So no surprise it doesn't work a lot of the time.
Comment
-
Could be a bit flip caused by a cosmic ray.
They are far more common than people imagine.Comment
-
Yeah, I think the motherboard manufacturer is supposed to test the memory with XMP Tweaked.
I ran V11.6 with 3 Passes and Memory set to XMP Tweaked and no errors all passed. I ran V11.7 with 3 Passes and Memory set to XMP Tweaked and no errors all passed. I ran V11.7 with 3 Passes and Memory set to XMP II and no errors all passed. So, it could just be a random thing with setting Memory to XMP Tweaked (sometimes stable sometimes not). Maybe if I went to all the trouble to up the voltages and etc... maybe XPM Tweaked would be more stable.
One thing I did noticed was that on the Main Screen of V11.7 with Memory set to XMP Tweaked, the Ram Configuration shows the Memory as DDR5 6798MT/s. Should be 6800. V11.7 with memory set to XMP II shows 6800 on Main Screen. V11.6 showed DDR5 6800 even when set to XPM Tweaked. Funny thing was, on V11.7 with XMP Tweaked after running MemTest86 and the going back to the Main Screen it showed DDR5 6800 not DDR5 6798 as it did before running the test. But maybe just has to do with settings on XMP Tweaked. Have attached screenshots. ThanksComment
-
If the memory vendors thought the RAM could be run faster and still be stable, they would have advertised and sold the RAM like that, for a higher margin. So it doesn't really make sense to expect all RAM to run stable beyond the manufacturers specifications. In short, you are asking for problems running it out of spec. For what might be a 1% or 2% overall performance gain if only the sub-timings are modified.
Yes a bit strange. But might be correct. The difference in the clock speed is very small.it showed DDR5 6800 not DDR5 6798
Maybe very small clock change to during the training sequence to improve signal quality.
Or maybe a Spread Spectrum Clocking (SSC), where small changes are made to the clock to spread the EMI (electromagnetic interference). So this would be normal and expected in this case.
Comment
-
I agree. But it's just like trying to overclock a CPU. The manufactures don't advertise it to go that fast, but people overclock them and if they can supply enough voltage, keep it cool and other things, sometimes it stays stable. ASUS was doing the same thing sort of. They tested certain memory with certain CPUs to see how fast it would go and stay stable. But I never could find a list of the memory they tested with XMP Tweaked, just XMP. And even if it's the same memory, it may not be the same if made at different times. Kind of like CPUs, just because one will overclock, doesn't mean the next one will. Even if they are the same.
Anyway, I think I will just keep it at XMP II. Thanks for the info and help.Comment
Comment