Can electrical wiring adjacent to operating ram cause errors?

Hardware Motherboards
1

Current passing through a wire generates a magnetic field, and magnetic fields can disturb/flip bits on ram modules. There’s a fan header on my motherboard that is situated right next to my ram. I’m noticing that if I plug in about > .05 amps worth of current into that header, hcimemtest will throw errors in the first 25%. When I have < .05 amps, I can run hcimemtest to 1000%.

Is it possible that too much current near my ram can generate a magnetic field strong enough to disturb it during memtest?

2

You’d need a changing magnetic field. A magnetic field on its own doesnt induce a current. It needs to change polarity to induce current flow in something else. This is known as induction.

Id say you’ve got something funky going on with the board, not some kind of magnetic interference personally. Perhaps some kind of signal noise.

1 Like
3

It is a pwm signal passing through. Could the rapid pulsing generate that current?

I’m not ruling out signal noise either. The fact that the header is right next to the ram modules could be causing signal interference in the motherboard traces somewhere.

4

its not changing polarity, so no. induction works best with a coil of wire as well. you arent likely getting much magnetic flux from a straight piece of wire.

1 Like
5

Okay, so something with the mobo likely. I’ll just remove the Y splitter I was using on that header, and use it only for pwm signal to a separately powered fan hub.

6

completely besides the fact, that it is most certainly not the cause of this. I must say changing polarity is not required for emitting a electromagnetic field. Just change in current is needed and yes a coil is better for emitting, but there is a reason a mainboard consists of multiple gnd-layers to absorb even this field.

I suggest you try your test with no fans attached on this header if you really want to test your hypothesis

7

I am testing now. Do you guys know of any memory test software that is faster than hcimemtest? Right now it takes 48 hours to reach 1000% coverage which is just stupid long.

I’ve read karhu is much faster, but their support page says it only supports 64 threads. Can I run 2 instances of karhu to overcome this?

Or maybe I could disable SMT, but I’m worried that may skew my results.

8

Curious what mobo you have that is causing this issue

9

I can’t remember for the life of me where I saw it. But one of the tech tubers I follow whether it was Wendell, Kyle, Paul or Linus, they had this exact kind of issue with a fan cable causing interference with their ram. And as soon as they rerouted the cable away from the ram, the problem went away.

2 Likes
10

I have the asus zenith 2 extreme alpha. I’m referring to the cha 1 fan header right next to the leftmost dimm slot.

I pretty much ran hci memtest every night to make ensure my system was stable since I had major issues with my prior board. Everything was rock solid even over 1000% coverage until I added in a second fan to the chassis 1 header and a UPS to my build. Those are the only changes and then I started getting errors within the first 100% of memtest.

11

@whizdumb I’d love to find that video too!

12

I also have that same vague memory as well, and have kept wires away from my ram since. No idea where the source of it now though.

13

@Whizdumb it was Wendell, I also remember it but cannot find it as it was more of a passing comment and not the subject of the videos itself I think.

1 Like
14

I’m pretty much able to confirm more or less now that it was the fan header and passing too much amperage through it causing instability in hcimemtest.

I’ve run it 3 times now with my Y splitter and it usually gets 8 or 16 errors at <50% coverage.

Now passing 400% coverage with no errors when I switched to a fan hub, with the header only being used to supply PWM and Tach signal.

So to recap, dont push too much on your motherboard fan headers, especially if it’s near your ram. Use separately powered fan hubs if possible to keep the electrical load off your mobo.

3990x, 256gb G skill 3600mhz, Asus Zenith 2 Extreme Alpha

15

This is really interesting. I know it was mentioned earlier, but a constant current should not cause interference. It would be interesting to know if the fan was given just a DC supply instead of a PWM supply if there would be interference.

16

It could. On some mainboards on higher speeds, running the CPU power over the RAM will cause problems aswell.

17

Here’s the setup/fans used if anybody or perhaps Wendell wants to try to reproduce:

Setup is 3990x, 256gb g skill 3600mhz, asus zenith 2 extreme alpha. Fan header is chassis 1 near ram. Memory test program is Hci memtest.

Memory errors:
Noctua nf f14 ippc 3000 and noctua nf a12 slim plugged into header near ram with Y splitter.

No memory errors:
Noctua nf f14 ippc 3000 pwm by itself plugged into the header.
OR
Sata powered fan hub plugged into header, and running noctua nf f14 ippc 3000 and noctua nf a12

18

I remember that as well but don’t know what the video was.

1 Like
19

Thanks for the sanity check Zibob. Maybe an @wendell might help bring some context to the question at hand.

1 Like
20

If you want to make sure there’s something wrong with the cables or maybe it’s something else wrap the cable in tinfoil, punch a hole in the tinfoil and couple it as well as you can with a cable and than back out a screw you’re using to mount the motherboard and wrap on the thread the cable you just connected to the tinfoil.

This way every potential current that’s being generated by the PWM signal will go traight to ground. If things don’t change this way there might be an issue with the motherboard I guess. Maybe the ground plane of the PCB is not good enough.