Is a single error an error too many? How do I fix it?

[leepox]

So I have a new Patriot Memory Viper Steel PVS416G440C9K and decided to tighten the timings a little bit at 3666 MHz to 15-15-15-15-30, ran memtest overnight and only a single error was found. Is this one error too many?

I have a 3900x and a B550 ITX MSI Gaming with AGESA 1.0.0.2. RAM voltage is set at 1.38.


Timings:



My VDDP and VDDG are reading 0.895 and 0.945 respectively. Should I bump this up by 0.1v to go over the recommended 0.9 and 0.95? Or is this a case of I need to loosen the timings a little bit... perhaps the tRCDRD to 16 (which is recommended based on the calculator) and tRAS to 31?

It's frustrating that there's one error and it seems so marginal!

[David (PassMark)]

In my opinion yes. One error is one too many.
Computers are already unreliable enough without dealing with a known faulty memory.

> My VDDP and VDDG are reading 0.895 and 0.945 respectively.
> Should I bump this up by 0.1v to go over the recommended 0.9 and 0.95?

Unless you have a masters degree in solid state memory I wouldn't touch the default settings. The manufacturers have experts that set these values. I would never presume to know more than they do about their own products. I don't know who "1usmus" is, but I wouldn't assume he knows more about every product than all the vendors. The performance gains from RAM overclocking (or latency reduction) are super minimal, but risk of instability fairly high.
If the sticks don't work at their default settings return them.