PC performance correlated with relative humidity?

[Nominal Animal]

What about if it was ECC memory. Would you still do the 48 hour tests ?

Definitely.  If obtainable, I'd check the ECC logs/statistics; if not, I'd just run a longer test.  I do not tolerate memory errors at all.

After passing that check, I'd test both storage and CPU (cooling).  For spinning drives, I mostly relied on SMART attributes; they're not "reliable" (in the sense of predicting failures), but they do catch Monday units (the really crappy ones that proper quality control should have caught).  For CPU, I run various CPU burning utilities for a few hours, measuring the core temperatures and the CPU frequencies to see continued steady state operation, and things like thermal throttling.  (However, I prefer low-TDP processors and huge heatsinks and silent machines, because I don't have a well ventilated but soundproof closet to keep a noisy one in.)  For the first year, I check kernel logs whenever I see anything that feels odd.

You could say I use my intuition and gut feeling as a reason to gather metrics and analyse the system and situation, because I do not trust my senses alone.

That is, I know that my senses can be affected by things like flickering fluorescent lights and so on, and cannot tell whether an issue is in the hardware or in my perception, unless I can find numbers that correlate with my observations.  For example, I vary the brightness of my display almost hourly through the day, to keep eye stress to a minimum.  If I do not, stuff like fluorescent lights in my kitchen can become irritating.  Because I am addicted to problem solving, finding the mechanism (or underlying reasons) for such oddities is kinda like another hobby to me.

[MK14]

What about if it was ECC memory. Would you still do the 48 hour tests ?

Definitely.  If obtainable, I'd check the ECC logs/statistics; if not, I'd just run a longer test.  I do not tolerate memory errors at all.

After passing that check, I'd test both storage and CPU (cooling).  For spinning drives, I mostly relied on SMART attributes; they're not "reliable" (in the sense of predicting failures), but they do catch Monday units (the really crappy ones that proper quality control should have caught).  For CPU, I run various CPU burning utilities for a few hours, measuring the core temperatures and the CPU frequencies to see continued steady state operation, and things like thermal throttling.  (However, I prefer low-TDP processors and huge heatsinks and silent machines, because I don't have a well ventilated but soundproof closet to keep a noisy one in.)  For the first year, I check kernel logs whenever I see anything that feels odd.

You could say I use my intuition and gut feeling as a reason to gather metrics and analyse the system and situation, because I do not trust my senses alone.

That is, I know that my senses can be affected by things like flickering fluorescent lights and so on, and cannot tell whether an issue is in the hardware or in my perception, unless I can find numbers that correlate with my observations.  For example, I vary the brightness of my display almost hourly through the day, to keep eye stress to a minimum.  If I do not, stuff like fluorescent lights in my kitchen can become irritating.  Because I am addicted to problem solving, finding the mechanism (or underlying reasons) for such oddities is kinda like another hobby to me.

Thanks for putting my mind at ease. If I was setting up a very long time (>= 48 hours) ECC memory checker running. I'd be tearing myself to pieces, thinking that it was silly to check (automatically self-checking) ECC memory. What you just said, makes a lot of sense. One wants to make sure that you haven't bought duff ECC modules, early on. While the possible guarantee return window is still open and/or weeding out faulty modules early on, should help with computer (Ram) reliability.

In the past, when CRT monitors were common place. I found their flicker could be (at lower refresh rates), rather noticeable, annoying and tiring/unsuitable for anything beyond, brief usage. For reasons that I don't really properly understand why. But since LCD screens, that seems to have largely removed that problem. I suspect/understand, that it is probably largely (or more) because of LCDs significantly slower update speed and/or because CRTs not only flash on and off at the frame rate, but the entire picture is drawn in brief scan lines, which rapidly fade, long before even a small part of the rest of that page is drawn. I.e. It is relying on the eyes persistence of vision, to even see the moving picture.
I didn't know that at the time, but there are extremely high speed photography pictures, of how CRT monitors, draw out the computer images, which demonstrates how it is working.
I.e. I thought it was flashing at perhaps 75 times a second (it is), but the primary flashing is at a considerably higher flash rate i.e. a couple or so of lines at a time. Let me show you (the first couple of minutes has it in various time places):



Whereas (also shown in the video), LCD is more of a show an entire page (even at slow-mo speeds). Which maybe is part/all of the reason, I have much less visual issues with LCDs and their (often unchangeable anyway) refresh rates.

I think part of the OP's issues, is that they are able to detect certain display image defects. But they can't necessarily point a finger as to what is precisely wrong with the images (game). In some extreme situations they can, like the image(s) of them firing the gun in one image, but what the server sees, is different, and doesn't have a firing gun image in it. But in others, the OP can sense issues, but not really be certain as to why it looks 'funny'/wrong.

[ultraknur]

I appreciate all your efforts guys and thank you but I think I'll just give up. It's basically unfixable however everything seems to be normal according to the tests. The last resort is to ask my neighbour if they have magnetron behind my wall. Maybe somebody else will find good solutions in this thread.

[Nominal Animal]

Thanks for putting my mind at ease. If I was setting up a very long time (>= 48 hours) ECC memory checker running. I'd be tearing myself to pieces, thinking that it was silly to check (automatically self-checking) ECC memory. What you just said, makes a lot of sense. One wants to make sure that you haven't bought duff ECC modules, early on. While the possible guarantee return window is still open and/or weeding out faulty modules early on, should help with computer (Ram) reliability.

It also catches problems like duff voltage regulators on the motherboard.  (I've had one on a Gigabyte GA-MA78GM-S2H motherboard explode: literally crater itself with a very loud POP!.)

Such "long" tests are especially useful when self-building a machine, because things like not seating a memory module properly are easy to miss since one does it so rarely.  As long as one uses the same BIOS/EFI memory timings, it also ensures that when one installs an OS on it, any immediate problems are just installation issues and not hardware issues.  Running an iobench or similar benchmarks for storage also gives you a metric to compare against when starting using the system.

(When I ran machines with local spinny-disk RAID clusters, I also experimented with the settings a bit before installing an OS to ensure I got "near-optimal" performance.  Once again, I wasn't actually looking for "the best settings", but the settings that didn't have any obvious drawbacks or weaknesses.  I also weigh scattered small (4k) I/O much more than normal storage benchmarks, because it is much closer to my real life usage on desktops.)

Also, the system burn-in process can easily take me a couple of days, because I then also experiment with the placement of the chassis fans and baffles, and the fan speeds.  (Remember, my target is not a high-wattage cooling system, but a silent one, using a 65W to 95W total-thermal-product CPU.)

In the past, when CRT monitors were common place. I found their flicker could be (at lower refresh rates), rather noticeable, annoying and tiring/unsuitable for anything beyond, brief usage. For reasons that I don't really properly understand why.

My first home installations of Linux were dual-booted to a command-line interface, with a custom 132x43 modeset on a Number Nine VESA graphics card with a 85+ Hz refresh rate on a NEC Multisync 17" monitor.  (I even did my own text reader program with soft (scanline-based) scrolling on it, first using Turbo Pascal in DOS.)
The flicker was a big problem for me, too.

But since LCD screens, that seems to have largely removed that problem.

The backlight generates the light, and the LCD only absorbs some of it.  The two are not at all synchronized, and the backlight are run at much higher frequencies.  (I don't know exactly what the frequencies are, and what the flicker amplitude in terms of light actually is, but the difference is so large that cats and dogs, who didn't really see the image from most ordinary CRTs due to the flicker, now see most LCD screens as humans do.)

I appreciate all your efforts guys and thank you but I think I'll just give up. It's basically unfixable however everything seems to be normal according to the tests. The last resort is to ask my neighbour if they have magnetron behind my wall. Maybe somebody else will find good solutions in this thread.

There is still one more experiment you can try: temporarily move your gaming computer to a friend's place, to see if the rig exhibits the same problems there or not.

If the rig exhibits the same problems there, you know it is the computer that is at fault.  (It will not tell what is wrong, only that it is something about the machine or Windows installation.)

If the rig does not exhibit the problems there, you know the problem occurs because of your home somehow, and the only real way to get rid of it is to move.

[MK14]

I appreciate all your efforts guys and thank you but I think I'll just give up. It's basically unfixable however everything seems to be normal according to the tests. The last resort is to ask my neighbour if they have magnetron behind my wall. Maybe somebody else will find good solutions in this thread.

Sometimes giving up, and moving on, is the best course of action. Not all problems can be realistically solved. Otherwise people could go through life, trying to solve a particular problem, and get nothing else sorted out, ever.
Not all purchases, turn out as expected or intended. Some turn out sour.

One last thing. If you use the gaming machine in a fast 240fps?, single player first person shooter game, N.B. while NOT gaming with other players over the internet. Does it still exhibit these problems ?
E.g. Some allow you to play single player and have bots to play against in matches.

Because if it doesn't (show any problems), that would tend to imply that your computer is basically fine, and whatever EMI/EMC issues (which probably are not the cause, anyway), is NOT affecting your computer.

But if it does still exhibit the problem, then that basically rules out the internet (connections). As you would be playing locally, against Game AI bots. Ideally don't have the internet connected (running) at all, if possible (game authorization might be tricky, but steam does allow off-line capabilities).
To rule out internet things glitching/affecting your computer, but anyway NOT connecting to the game server, would perform most of the test, even if the internet is still running/connected.

[ultraknur]

Yes, it still exhibits the problems. First of all it's not adequately smooth. Even G-Sync doesn't seem to help. On desktop when moving objects they slightly "vibrate" and cursor gets some kind of acceleration which makes it impossible to move cursor at same speed from point A to B (however mouse acceleration is 100% off in the settings). If I have to guess, I think that not only my PC is affected but also my ISP's equipment in the building.

[MK14]

Yes, it still exhibits the problems. First of all it's not adequately smooth. Even G-Sync doesn't seem to help. On desktop when moving objects they slightly "vibrate" and cursor gets some kind of acceleration which makes it impossible to move cursor at same speed from point A to B (however mouse acceleration is 100% off in the settings). If I have to guess, I think that not only my PC is affected but also my ISP's equipment in the building.

That's good, because that is eliminating your ISP/internet, as being the main cause of the problem. Which narrows it down, and gets us closer to solving your problem. It seems quite a lot of people on the internet, have problems similar to how you describe.

Does the following link, describe your problems and/or help ?

https://computerinfobits.com/micro-stuttering-in-games/

[SilverSolder]

Yes, it still exhibits the problems. First of all it's not adequately smooth. Even G-Sync doesn't seem to help. On desktop when moving objects they slightly "vibrate" and cursor gets some kind of acceleration which makes it impossible to move cursor at same speed from point A to B (however mouse acceleration is 100% off in the settings). If I have to guess, I think that not only my PC is affected but also my ISP's equipment in the building.

I have owned countless PCs since they first came out, and I have seen that kind of behaviour (desktop objects slightly "vibrate" when dragged with the mouse, the mouse cursor doesn't move smoothly / is slightly laggy).   In all cases, it was something to do with the PC itself, not external factors. 

The troubleshooting method that has worked best for me in the past has always been to strip the PC to the minimum configuration that will boot...   one stick of RAM, one CPU (if it is a multi socket mobo), simplest graphics, basic OS install.   Does that misbehave?  if no, start adding parts back in.

[dervu]

It might also be some faulty component which symptoms are exacerbated by interference?
Idk if it makes sense.
That would be only ruled out by having multiple PCs checked at same place and another place.
If one of PCs behaves different and others do not, then it is related to hardware.
If most/all of  PCs behave different, then it is probably related to interference. Unless all hardware is prone to interference. 

[MK14]

It might also be some faulty component which symptoms are exacerbated by interference?

I don't think I've heard of such a phenomenon. Basically if a PC component, such as a RAM module is faulty, then it is faulty. Interference doesn't really come into it. If the PC has faulty components, they really need to be replaced.
Assuming you have got a normal (all metal) PC case, without clear windowed areas, perhaps for flashing RGB lights. Even so, interference shouldn't mess up the PC, unless it was really, really extreme. In which case you would expect other household equipment to have issues as well.

Faulty PC components is a somewhat common problem with PCs, especially over a long period of time. But one doesn't seem to hear of PCs being interfered (EMI/EMC) with, usually.
Except for maybe any wireless connections (WiFi, Mouse/Keyboard), which can be interfered with, which might make them not perform too well.

But I have heard of stories (but reasonably rarely), where the incoming mains supply is very bad (especially as regards its voltage and Earthing), which can make the PC problematic, such as regular reboots and things. But then you probably would notice your room lights going funny at times (going dim, flickering or not lasting very long). In such cases, the solution is to get a proper qualified electrician to diagnose and fix it.

[Nominal Animal]

Here are the key components of ultraknur's system:

• Processor: AMD Ryzen 7 5800X.
  Overclockable processor with default TDP of 105W.
• Memory: G.SKILL 32GB 3600MHz CL16 TridentZ RGB Neo.
  In the Asus Qualified Memory Vendor List, the only G.Skill 16MB 3600MHz CL16 modules are those using SK Hynix chips; the otherwise similar CL17 modules using Samsung chips are also listed.
• Motherboard: Asus TUF Gaming X570-PLUS.
  AMD Ryzen 7 5800X CPU is supported by this motherboard in BIOS 2607 and later.
• Graphics card: Gigabyte GeForce RTX 3070 Aorus Master 8GB.
  According to TechPowerup, the TDP of this card is about 220W.
• Power supply: BE QUIET! Straight Power 11 750W 80 Plus Gold.
  Should fulfill the needs of this system.

I do not believe the memory modules are an issue, because memory issues show up as instability and crashes, not performance glitches.
Every component in isolation seems fine for this setup, and assuming the motherboard BIOS version is 2607 or later, they should work fine together.

ultraknur mentioned they use AiO cooling solution (closed-loop liquid cooling).  This got me thinking.

TDP is Thermal Design Power, and is basically a measure of how much waste heat the part produces that must be, uh, excreted outside the enclosure.  105W from the CPU and 220W from the GPU is not that much to get rid of using a closed-loop cooling system, but what about the rest of the motherboard?
There are all kinds of additional chips from voltage regulators (actually DC-DC converter banks) to bus logic that do heat up.

ultraknur, do you have adequate intake and exhaust fans in the enclosure to keep the air within the enclosure at a reasonable level, say < 40°C?
Double AiO for the GPU and the CPU without any enclosure fans is definitely not sufficient.
And have you verified your airflow pattern inside the enclosure to fully mix the air inside the enclosure, without "pockets" of non-mixing, hot air?

The best device to use for this is a normal indoor-outdoor thermometer with an external sensor, since their sensors are designed to measure air temperature, and they're quite cheap and ubiquitous.  A few degrees off is not that important either, as you really are just comparing the air temperature outside the enclosure, to air temperature inside the enclosure at a specific spot.  Of course, you do need to close the enclosure fully, to get relevant measurements.  Using one of the holes at the back of the enclosure to slip the sensor through, works.

Thermal throttling could well explain the slowdowns and lagging, but I don't know the X570 chipset well enough to say for sure if the chipset can cause thermal throttling, and whether say the chipset thermal throttling due to too high ambient temperature within the enclosure could explain the symptoms.  According to Wikipedia, the X570 chip itself has about 15W TDP; not that much, but not completely insignificant either.

(I am aware that ultraknur does not see any excessive GPU or CPU temperatures, and am not talking about those components causing thermal throttling.  I am talking about other components on the motherboard, mainly the X570 chipset, or perhaps the CPU voltage regulators, getting too hot because of inadequate enclosure airflow, that the motherboard compensates for by automatically dropping the CPU frequency or causing some bus stalls or some such, that present as sudden "lagging" or "desyncing", because the computer basically freezes or at least slows down a lot whenever that happens.)

When heatpipe CPU coolers first came on the market, many motherboards started suffering from e.g. the CPU voltage regulators overheating because of insufficient airflow.  Then, the end result was usually an otherwise unexplicable freeze; the systems just froze, without even crashing.  Currently, that problem is usually avoided by the motherboard manufacturer using a built-in heatsink, often using heatpipe(s), connecting the chipset and the most heat producing components like the voltage regulators, with taller cooling fins near the connectors at the backside of the motherboard, tall enough to catch some of the airflow from the processor heatsink and fan.  I am not up to date on the niceties of the various chipsets like the X570, but them having some capability of throttling the machine to avoid overheating, even if they do not expose a software-accessible thermal sensor, would not surprise me at all.  (It would not be something expected to trigger in normal use, it would be something designed to stop component damage, you see.)

[MK14]

Can you get it to work well enough in Linux (mentioned earlier in the thread), to play a reasonable game for long enough to exhibit your problem(s) ?
Ideally, just boot into Linux (I don't know how you did it previously), and install enough to play representative games (not all will work without windows). Nothing else.
If it continues to show your problems, then that would tend to point to hardware. But if it doesn't, and the game(s) play nice and smoothly, then that would tend to indicate that windows/drivers/software, other stuff you installed on the windows disk(s), may be causing your issues.
tl;dr
It would be nice to know if it is hardware or software issues, you might be suffering from.

[dervu]

What do you guys think about graphs below from rf spectrum analyzer?
First one is my home, where either disabling old radio triggered improvement or some external factor did it. Still I see difference when switching on/off highend EMI/RFI filter into socket <-> PC circuit.
Which makes me think that it can come back at some point with full strength, as it always does after some time passes after some change.
Second one is office.
Both places have elevators.
When elevators are moving, there is none interference present at the office.
At my place, at every time of any day there is some steady noise at that range, noise is much higher when elevator is moving (as you can see example in screenshot)

I am still looking into recordings to see any differences.

[dervu]

I have 1000fps camera now. I will share my findings when I setup my tests good enough.

[MK14]

I have 1000fps camera now. I will share my findings when I setup my tests good enough.

That would provide much more down to earth/definitive results, than just playing games for a few days. Then saying, it felt a tiny bit odd, maybe.

[dervu]

I am not sure if it will be the same for original poster. As many know, EMI (if it even is EMI or not some other issue related to electricity) issues might not always be the same.
In my case after years of looking I found the source.
There was loose electrical connection for kitchen hood in my apartment.
Immediate difference after removing. Then replaced with new one and still no issues.
It was the last place where I would look, but I had to replace it to comply with new rules regarding ventilation. 

Either it was:
- Just loose electrical connection
- Loose connection combined with little shakes from elevator that is just behind the wall this kitchen hood is placed at
Maybe this connection was loose where the fan was replaced many years ago or it happened gradually from shakes.

When I will have more free time, I will try to reproduce it with connecting this kitchen hood closer to the PC and measure what I can with RFI spectrum analyzer.
For now I can record another 1000fps videos with exactly the same positioning and same emulated inputs to see if there are any differences to my older recordings.

[radiolistener]

Some air conditioners from neighbors can leads to a very high SMPS EMI noise in a range from 50 kHz up to 20-30 MHz. Probably due to bad quality Chinese power supply

[Nominal Animal]

In my case after years of looking I found the source.
There was loose electrical connection for kitchen hood in my apartment.
Immediate difference after removing. Then replaced with new one and still no issues.

I admit, this is the first time I've heard that EMI and/or mains power issues has affected computer performance.
I have only seen and heard of crashes and instability thus far (in roughly 35 years of computing), due to power supply issues, mains power issues, and extreme EMI (most often due to large brushed motors and internal combustion engines with harsh unshielded spark plugs).
(In my observation, sparking seems to be the worst offender in all cases.)

It has to be current graphics cards with their rather large current consumption, being somehow performance-sensitive but not crash-prone to these kinds of issues.  Live and learn!

Thanks for the report, too.  It's good to hear you got it solved, even if by accident.