Conflicting memory test results

[Hopper64]

Flashed 1505 on my MZ690E a few days ago and noticed I needed a little more SAV for stability with the memory. Hardware in my sig. Increased SAV to 1.12V from 1.10V that I was using in 1403. Ramtest and TM5 both finding no errors after some tuning, but Memtest86 is another story. Had to increase VDD/VDDQ to 1.375V for memtest86 to pass all tests. Running both TM5 and Ramtest with VDD/VDDQ at 1.375V produces errors within seconds of starting each test. Falling back to VDD/VDDQ with 1.370V and TM5 and Ramtest pass with 0 errors. Running memtest86 at VDD/VDDQ 1.370V produces errors in the first pass.

Quite odd. Never had seen this before. Any ideas? Thanks.*

[Silent_Scone]

Hi,

If the system is stable in the OS with preferred settings I wouldn't read too much into it unless you experience instability. 5mV on VDD/VDDQ is a small increment, so it depends on where the point of failure is. I only tend to use Memtest86 to test a baseline and not much else.

You can try returning both the CPU and Uncore to defaults and see if the behaviour repeats itself.

[Hopper64]

At stock settings and no OC, it passes memtest86 with no errors. It’s clearly the OC causing the issue. I will just watch it and see what happens. Thanks.

[Silent_Scone]

Hopper64 wrote:
At stock settings and no OC, it passes memtest86 with no errors. It’s clearly the OC causing the issue. I will just watch it and see what happens. Thanks.

Of course, I'm not sure why you'd think it was anything else. You can try TM5 in the OS to see if this finds the same instability, however, if you're able to pass Karhu or similar and use the machine without instability (and data integrity isn't critical), simply soak test the machine with the settings within the OS that work best. Memtest DOS and OS stability are mutually exclusive.

[Hopper64]

Silent Scone@ROG wrote:
Of course, I'm not sure why you'd think it was anything else. You can try TM5 in the OS to see if this finds the same instability, however, if you're able to pass Karhu or similar and use the machine without instability (and data integrity isn't critical), simply soak test the machine with the settings within the OS that work best. Memtest DOS and OS stability are mutually exclusive.

I know, I was curious as to whether it would pass using my OC. It has in the past. 🙂

[Silent_Scone]

Hopper64 wrote:
I know, I was curious as to whether it would pass using my OC. It has in the past. 🙂

The other thing to note is that different data patterns expose different instability, because the swing in current isn't the same. The same reason some OS stress tests expose instability that alternatives do not

[xarot]

I think it is a board issue with the Z690 Extreme. Fine-tune RAM and voltages and pass Karhu 10000%. Restart the PC, fails in few %. Change some voltages, you might be stable again. Basically very hard to find what is causing this without any consistent results. I have seen this on both Hynix and Samsung 2x16GB 6000 MT/s kits.

5600 MT/s provides consistent results with my board.

We are not alone with this issue.