Weird question: can unstable DRAM frequencies cause WiFi issues?

[Valantar]

So this is quite out there, but the correlation is too strong for me to ignore: about a week ago, I decided to have another go at getting my RAM to run closer to its rated speed. Of note: the Biostar X370GTN is not a good motherboard, 1st gen Ryzen doesn't like fast RAM, and my old TridentZs (Hynix MFR die) aren't particularly good either. I've never got 3200 to work, and even 3000 has had issues. After properly learning how to generate reports for 1usmus' dram calc from Taiphoon burner, I decided to give 3000 a go again, which was ... kind of stable. But not really, crashing ~once a day. At the same time I tried this, my WiFi started acting up (haven't had time to install Ethernet cabling in the new apartment yet), being entirely unusable at times, with pinging the router giving upwards of 30% packet loss, massive ping spikes (3000ms or more), and internet connectivity being very hit or miss. Using my phone or laptop in the same spot showed no connectivity issues, so it's not a coverage thing. Trying another WiFi adapter showed no change. Once the Windows networking stack seemed to crash entirely blocking me from accessing network settings, before cascading step by step into the system being entirely unresponsive, forcing a power button hard shutdown.

Needless to say, after that I first ran a few runs of MemTest86 (first run was error free until the 2nd to last test of the 4th run, subsequent runs crashed and rebooted), then backed off the memory speed a bit. Bumped the voltage by another ~10mV and went from 3000MT/s to 2933. Reran a few rounds of memtest, error-free, and ran prime95 Large FFTs for ~16h with no errors or warnings. And all of a sudden my WiFi is working perfectly - no packet loss, no ping spikes, perfect connection quality. Can there actually be a connection between the two?

 

[lexluthermiester]

Short answer, no. Not radiometricly anyway. The only possibility is that the instability in RAM might be at an address that happens to be where your WIFI driver and/or data stacks are located. Without seeing it first hand and running debug there's no way to know for sure.

 

[RyanOCallaghan01]

In my experience, yes. I can't reliably run my command rate at 1T on my 4x8 Micron Rev.D sticks, even though Karhu or TM5 couldn't find any errors. I'd get random Wi-Fi cutouts with increasing frequency if I lower the VCCSA voltage. 3200 MHz at it's rated 16-18-18-36 with a 2T command rate even with tweaked secondary and tertiary timings, and a fairly low 1.05v on the VCCSA, works fine.

 

[tabascosauz]

It doesn't seem like it should, because networking usually comes off the PCH (X370). But the AM4 chipsets are also fed by the CPU through a x4 uplink. It's not uncommon on Ryzen 3000 to run into sporadic USB issues/disappearing act when you're pushing the Infinity Fabric in a way it doesn't like (e.g. SOC too low for freq, SOC too high, freq just too high for the chip). Even if the specific USB ports are native to the PCH, not the CPU, so it doesn't seem too far fetched to think that Wifi could be affected.

I seem to remember that while the Ryzen 7s can hit 1600 IF (3200), the lower end Ryzen 1000s are a bit less fortunate. May be an indicator that your CPU would rather stay at 2933.

For strict memory testing, TM5 and HCI work well, but like all memory testing tools they don't test the stability of the memory controller and IF. There aren't really any ways to reliably test Infinity Fabric aside from the abrupt reboots/errors/peripherals throwing a fit that come naturally. Like on Intel, you generally just have to "know" when the mem controller or Infinity fabric isn't happy.

 

[Frick]

Seeing how memory issues can cause all sorts of weird stuff I'd say absolutely, for the reason @lexluthermiester gave. X is loaded into the memory, and if memory is unstable X will also be unstable. It would be interesting if you'd try to lower voltage and/or increase frequency and see if the problems came back.

 

[wavy]

I would get a spectrum analyzer to see if it is emitting any frequencies. I doubt it would but it would be fun to check it out anyways.

Wi-Fi and Non Wi-Fi Interference | MetaGeek

Learn about the shapes and behavior patterns of a few common devices that transmit in the 2.4 and 5 GHz bands, so you can identify Wi-Fi interference.

www.metageek.com

 

[TheLostSwede]

I would say you got interference from some USB 3.x 5Gbps device. Do you have long USB 3.x cables plugged in?
Some cables cause insane Wi-Fi interference, but obviously only on the 2.4GHz band and only close to the device that's having problems.
If it's on 5GHz, then it's something else entirely.

 

[lexluthermiester]

Seeing how memory issues can cause all sorts of weird stuff I'd say absolutely, for the reason @lexluthermiester gave. X is loaded into the memory, and if memory is unstable X will also be unstable. It would be interesting if you'd try to lower voltage and/or increase frequency and see if the problems came back.

This is correct.

Folks, no RAM ever made will cause interference to WIFI signals. Literally never! This is not opinion, it is science fact. RAM does not and can not emit radio frequencies in the spectrum and amplitudes that any WIFI does, has or will ever operate on. Whatever issues the OP is having are NOT related to EMF emissions from their system RAM. Not even CPU, GPU, mobo components or drives. The only possible source of EMF emissions from a PC that might cause WIFI problems is the PSU, but here's the thing, any PSU in poor enough condition to be causing that kind of interference is VERY unlikely to be providing stable power for the system to boot and operate off of and the OP would be having much more serious problems than just WIFI issues.

 

[Valantar]

Short answer, no. Not radiometricly anyway. The only possibility is that the instability in RAM might be at an address that happens to be where your WIFI driver and/or data stacks are located. Without seeing it first hand and running debug there's no way to know for sure.

Yeah, I never meant to imply that I thought this was RF interference (the RAM is inside of a grounded steel PC case with the WiFi adapter on the outside, so ...), just that I found it very odd that I got that specific issue and only that specific issue, persistently. Does Windows keep the same driver (or actually, the type of driver, given that it persisted across two WiFi adapters) in the same area of RAM at all times? It just baffles me that one specific thing can become that broken, with the rest of the system seemingly working as it should - yes, I had a couple of BSODs and system freezes (though I've also been playing No Man's Sky, which apparently isn't friends with my system), but no application crashes or other weird behaviour beyond specifically networking issues. Definitely interesting!

I would say you got interference from some USB 3.x 5Gbps device. Do you have long USB 3.x cables plugged in?
Some cables cause insane Wi-Fi interference, but obviously only on the 2.4GHz band and only close to the device that's having problems.
If it's on 5GHz, then it's something else entirely.

The router is set to auto select 2.4/5GHz, so I frankly don't know, but I would assume it's using 5GHz given the relatively short distance and single wall in between the two. But 2.4GHz interference doesn't really add up either - for that to suddenly appear something would need to change, and nothing has. No new devices connected, no driver updates, nothing. I have a USB 3.0 hub on my desk with my mouse receiver, xbox controller receiver and some other doodads connected, but I've never had issues with that (and if that was causing 2.4GHz interference, my mouse wouldn't work very well). I also tried both WiFi adapters connected to the front I/O, motherboard I/O and through the USB 2.0 hub in my monitor, with minimal difference. Also, the fact that no other devices have issues in the same area on the same network suggest that interference isn't the issue.

I would get a spectrum analyzer to see if it is emitting any frequencies. I doubt it would but it would be fun to check it out anyways.

Wi-Fi and Non Wi-Fi Interference | MetaGeek

Learn about the shapes and behavior patterns of a few common devices that transmit in the 2.4 and 5 GHz bands, so you can identify Wi-Fi interference.

www.metageek.com

That sounds very, very expensive for an issue that seems to have gone away entirely after I got my RAM stable And besides, as shown above it's highly unlike to solve anything

It doesn't seem like it should, because networking usually comes off the PCH (X370). But the AM4 chipsets are also fed by the CPU through a x4 uplink. It's not uncommon on Ryzen 3000 to run into sporadic USB issues/disappearing act when you're pushing the Infinity Fabric in a way it doesn't like (e.g. SOC too low for freq, SOC too high, freq just too high for the chip). Even if the specific USB ports are native to the PCH, not the CPU, so it doesn't seem too far fetched to think that Wifi could be affected.

I seem to remember that while the Ryzen 7s can hit 1600 IF (3200), the lower end Ryzen 1000s are a bit less fortunate. May be an indicator that your CPU would rather stay at 2933.

For strict memory testing, TM5 and HCI work well, but like all memory testing tools they don't test the stability of the memory controller and IF. There aren't really any ways to reliably test Infinity Fabric aside from the abrupt reboots/errors/peripherals throwing a fit that come naturally. Like on Intel, you generally just have to "know" when the mem controller or Infinity fabric isn't happy.

This is on my OG Ryzen system, so there's no IF involved. Both WiFi adapters are USB, btw.

 

[lexluthermiester]

Yeah, I never meant to imply that I thought this was RF interference (the RAM is inside of a grounded steel PC case with the WiFi adapter on the outside, so ...), just that I found it very odd that I got that specific issue and only that specific issue, persistently. Does Windows keep the same driver (or actually, the type of driver, given that it persisted across two WiFi adapters) in the same area of RAM at all times? It just baffles me that one specific thing can become that broken, with the rest of the system seemingly working as it should - yes, I had a couple of BSODs and system freezes (though I've also been playing No Man's Sky, which apparently isn't friends with my system), but no application crashes or other weird behaviour beyond specifically networking issues. Definitely interesting!

Just had a thought...

This is on my OG Ryzen system

...are you running your system with any of the "Spread Spectrum" features in the uefi enabled? I know it depends on the mobo model, but those have been known to cause confusing problems like the ones you're having. Go take a look and see what you find. If you do find them enabled, disable them and see how things turn out.

 

[Aquinus]

Both WiFi adapters are USB, btw.

I've had absolutely terrible experiences with USB WiFi adapters in both Linux and Windows. Sometimes it will work, sometimes you'll have weird issues, sometimes it just won't work. Also, USB adapters tend to have poor antennas, so I'll leave this pro tip here. Place the USB adapter somewhere that limits the physical objects that the signal has to go through. If you plug the USB adapter behind your tower and your WiFi AP is behind you, that signal has to travel through your chassis first. So from that perspective, make sure you put the adapter (antenna,) in a smart place.

Also, if memory was causing your WiFi to be unstable, I'd expect a litany of other issues with the stability of your system as well.

 

[tabascosauz]

This is on my OG Ryzen system, so there's no IF involved. Both WiFi adapters are USB, btw.

All Ryzen runs on Infinity Fabric......lol.

The only thing that changed after Ryzen 3000 is that dividers were implemented so that the UMC (now even better than Intel) could go further while leaving the shitty IF behind at the same clock. IIRC most Ryzen 3000 were guaranteed 3600 (1800MHz), most Ryzen 2000 able to hit 3200 (1600MHz), but most Ryzen 1000 only guaranteed 2933.

If they're external Wifi adapters over USB, that just reinforces my suspicions that you might want to stick to 2933 on that chip. IF ties all I/O together, think of it as the substrate or "mesh" that all the silicon lies on, so when you're running it on the ragged edge especially with less VSOC than it wants, weird shit happens.

That is, only if you're 100% sure that this only ever happens when you're at 3000MT/s or above.

 

[wavy]

But playing with a spectrum analyzer is so much fun! It was a tongue in cheek response.

 

[Valantar]

Just had a thought...

...are you running your system with any of the "Spread Spectrum" features in the uefi enabled? I know it depends on the mobo model, but those have been known to cause confusing problems like the ones you're having. Go take a look and see what you find. If you do find them enabled, disable them and see how things turn out.

That's a good thought, I'll take a look when I have the opportunity.

I've had absolutely terrible experiences with USB WiFi adapters in both Linux and Windows. Sometimes it will work, sometimes you'll have weird issues, sometimes it just won't work. Also, USB adapters tend to have poor antennas, so I'll leave this pro tip here. Place the USB adapter somewhere that limits the physical objects that the signal has to go through. If you plug the USB adapter behind your tower and your WiFi AP is behind you, that signal has to travel through your chassis first. So from that perspective, make sure you put the adapter (antenna,) in a smart place.

Also, if memory was causing your WiFi to be unstable, I'd expect a litany of other issues with the stability of your system as well.

That was what I was expecting as well! As I said, the system hasn't exactly been stable, but it's been unstable in the "~1 crash a day, only when stressed" way, not the "frequent and random BSODs" way that I would expect if things were as broken as the WiFi issue might suggest. A persistent and significant disruption of functton of one - and only one! - part of the system sounds very odd to my ears. Which is why I made this thread

As for the USB adapters, I'm well aware those can be pretty iffy. The one I've mainly been using is ancient, an old Dlink N300 adapter that's been my "what if I need WiFi in a pinch?" adapter for like a decade. The only issues I've had with it previously are that it tends to overheat if significantly stress, causing the connection to drop. Definitely not what's happening here. When the issues started showing I first thought it might be due to the new router (switched a while before the issues started, so no, it's not the router) not being friendly with that adapter, so I bought a new AC1200 adapter with a beefy high gain antenna (from the same brand as the router just for the illusion of added compatibility). It should also be noted that I've been using WiFi with the old adapter with only minor issues (as in: can't game over the connection, downloads are a bit slow, otherwise works fine) in this same location since we moved in here in August, so the issue is new. From what I can tell, the new adapter works perfectly fine, but only after I got my RAM stable.

 

[hat]

An unstable PC can cause all sorts of fuckery. My current PC can crash my router if it's unstable and I'm running a prime95 test or something. The router will refuse to function until the PC is turned off or restarted.