The issue may present itself as electrical noise from the 5KHz-20MHz range. The higher frequency noise in the MHz range likely isn't the cause of our problems, but it probably exacerbates it. I believe the main issue lies in the Hz and KHz range, and is caused by interharmonics and supraharmonics. It's normal to get some noise in those frequencies, but there's a threshold where they interfere in the power distribution and cause problems with electrical components (namely saturation of things like capacitors and transformer cores, which creates a snow-ball effect of larger and larger problems). Distortion of the fundamental frequency waveform can cause issues such as clock drift and ripple current inside your PC. Despite all the protective mechanisms inside a PSU, there are vulnerabilities to power quality issues, e.g. inductors and capacitors. These components are rated for a certain amount of current, and with harmonic distortion and subsequent low power factor, you can end up with saturation, and the components no longer do their 'job' of ensuring a stable flow of current. The same goes for things like online UPS's and power filters. Exceed certain thresholds and you'll end up with more distortion from those devices than you would directly plugging it into the wall (interharmonics are renouned for their difficult nature, they can easily overwhelm filters, AND incomplete filtering can result in the frequencies and effects being AMPLIFIED).
You can either read the journal article yourself (Interharmonics: What They Are, Where They Come From and What They Do) https://ccaps.umn.edu/documents/CPE-Con ... monics.pdf, or you can read my posts on page 4 where I post some of the TL;DR's and refer to other articles too. .
Shoutout to nug, unixko, numberfive and a bunch of the others in the discord, we share journal articles and bounce ideas off each other to sorta educate ourselves and go down investigative pathways.
THE TL;DR is that interharmonics can mess with ripple control methods, meaning it'll INDUCE RIPPLE CURRENT inside household electronics, and can present itself alongside other issues like differential-mode noise, meaning it can't be filtered by conventional means, including isolation transformers and double conversion UPS's (unless the UPS's are medical grade and are engineered to keep the ripple controllers from being affected, though I'm not sure even those can stop it, options are quite limited, it seems harmonic filters are pretty much your only option, but they need to be multistage filters with dampening if they are passive harmonic filters, or you go the active harmonic filter route you need to be sure they are rated to deal with interharmonics and subharmonics).
Anyway, ripple current inside your PC means there's harmonic distortion inside your PC. Once the capacitors in the PSU and then the mobo are saturated, components inside the PC affected and saturated too, they'll emit much more EMI than normal, they'll emit more coil whine than usual, they'll overheat, they'll experience voltage overload events if there's harmonic resonance or slightly below resonance, depending on the mobo you'll either have damaged components or the mobo will throttle those components to protect them. It doesn't matter how good your cooling is, if you CPU gets a split-second of 1.8V the mobo will throttle.
(this is the original post)
Here we go folks, did some input lag testing with MSI afterburner and RTSS for your viewing pleasure.
As you can see, the lag/stutter/input lag issue doesn't seem linked to frametimes, going from 0.5 to 7.5ms isn't that big a deal immediately after killing a bot. The input lag also isn't 100% linked to mouse clicks, but pretty frequently I get unlucky and it corresponds perfectly with a lag spike, and I get values ranging from 100-400+ms. (when the monitor just shows x's, it means it was over 400ms input lag).
The issue also seems to vary in severity on different game engines. On games like aimlab and CS:GO (at least when offline) it's not too bad, but on games like Kovaak 2.0, PUBG, Battlefield etc. it's pretty terrible. When playing online on CS:GO, it'll start off relatively smoothly, but after about 40 minutes of playing, things get a lot more stuttery, and it ends up feeling like I'm dragging my mouse through mud and people are just warping around the place e.g. youtube com/watch?v=PcgOpLRh0do
Additionally, I know things are not right, because when I brought my mouse and mousepad to a friends house, with him using almost a carbon copy of my rig, I couldn't use my normal sensitivity - 2200eDPI, I had to use 1400eDPI, and the tracking and general gameplay felt so much better, much more crisp, and I was just bullying people with 1 taps.
EDIT: I have a potential method to help reduce these mouse tracking/input lag/stutter issues: Long story short, when you notice the issue is at it's peak crappiness, try underclocking, using a frequency about 300-400MHz below the 'intel turbo boost' default (e.g. i7 8700k is 4.7GHz, so I'm using 4.3GHz), with a medium to high LLC (e.g. level 5-6 ASUS, or 'high' to 'extreme' with gigabyte, but using your CPU VID as a point of reference for your vcore voltage instead of normal-style overclocking.
People usually ignore CPU VID because it's the value your mobo thinks the CPU needs to function, which is usually too high, but I'm thinking we can use the VID to get an idea of how the issue is affecting our CPU power draw, and approximate an underclock that allows 'additional capacity' for voltage for when we're being affected by this power issue - BUT with a twist, using CPU VID to figure out what vcore your mobo thinks the CPU needs, and aiming to keep the CPU VID lower than 85% of your CPU's 'safe upper limit' vcore voltage, by using low enough frequency, LLCs and vcore settings to facilitate this. (e.g. my 8700k's safe upper limit is 1.4V, 85% of that is 1.19V, I want my CPU VID to be lower than 1.19V, so I used a high LLC and lowered my frequency bit by bit until I saw the CPU VID was under 1.19V, I then set my 'maximum allowed vcore' to 1.19V)
If you want to try this yourself here's how:
- 1. Set your CPU frequency/clock speed 300-400MHz lower, set a high-ish LLC, KEEP your vcore on auto, and then load up windows. Using a monitoring program like hwinfo, look at your CPU VID, spam a few clicks on desktop, play a game, then do like 5 minutes of stress testing with prime95.
- 2. Look at the maximum CPU VID your CPU reached, and also the vcore, the highest value you see is your new baseline vcore voltage. Is it less than 85%? If no, use a lower CPU frequency/clock speed.
- 3. Now that you're at a CPU frequency that's low enough for your VID to be less than 85% the safe upper limit, set your vcore to that limit. Enjoy smooth gameplay. You can try to increase your CPU frequency a little bit and get more performance out of your PC, but when the interharmonics are at 'peak awful' territory, you may not have left yourself enough 'overhead' to avoid having issues. Also, if you want to get rid of the stutters, you'll need to read the stuff on page 4 (control+f, search for ripple current).
Here's an early attempt at figuring out my ideal underclock. I wasn't doing p95 tests and looking at VID then, and the underclock wasn't quite right, but the difference is still very noticable. Look at the difference in mouse control from 1:40 and 2:17).