howiec wrote: ↑02 Oct 2020, 22:14
As for power going into your PC, it's always worth buying a good power supply. It's job is to ensure "clean" DC power output and protect/mitigate various failure modes. If you have a bad one, usually the failure isn't simply input lag, it's usually freezing/crashing, inability to power on, or catastrophic failure such as shorting and overheating.
Part 2 of 2
Also, read the
Milliseconds Matter Thread: The Amazing Human Visible Feats Of The Millisecond, where some milliseconds are invisible, and other milliseconds create human-visible effects. I don't own any appliance that's a defacto EMP bomb, but
even a single analog television set took out an entire town's DSL infrastructure (it really happened). Now, imagine a super-EMI appliance (Imagine; The 0.001% of the worst EMI-emitters, that's still a few out of a million) being right next to a computer, injecting EMI through a cheap computer power supply, and then really dirtying all the voltages inside the computer, creating those massive double-digit-percentage-loss errorcorrection/ECC bus floods that creates multithousand human visible millisecond-scale microfreezes (aka stutter).
A 1990s computer crashes, but a 2020s computer is full of error correcting buses & behaviors. So, thusly, the said computer is no longer guaranteed to crash;
Even today's PCI Express bus isn't like a non-errorcorrected 2400 baud modem that spews garbage when somebody else picks up the phone line (qF2{TfRJZ^s-y,a....NO CARRIER) while you were calling a BBS. Today, a PCI Express bus (in conjunctsion with PCI Express peripherals) is a packetized "error detection and retransmit mechanism". Yep. LAG, if there's a million-retransmit flood. Those bus lines are almost literally getting more like a DSL phone line now (just a different order of magnitude; the multi gigabit scale). While yesterday's PCI had error detect and retransmit, tomorrow, PCI 6.0 will have FEC too (forward error correction) so error corrections out of the wazoo because buses are getting faster and more sensitive to EMI! PCI Express is capable of lagging noticeably if tons of of packets are lost, for example. Same for NVME, SSDs, USB buses, etc. Any EMI could go anywhere, like a chip pin thanks to a cheap chinese design oversight, and whatnot that was not detected by automatic circuit board EMI-checking software, etc. BOOM. Lag instead of crash. There's billions of potential weak links in a modern computer (counting on-chip stuff -- even a single transistor can sometimes be a Chaos Theory Butterfly) and you only need enough weaknesses here n there. See, engineers? (Like the way I school disbelieving engineers about 240Hz benefits).
Heck, today DisplayPort and HDMI 2.0 is also an error-correct packet system now. Yep. Packets with error correction. On your cable between computer and monitor. Also, recently, very slightly variable latency sometimes occurs on DisplayPort and HDMI, though usually in the sub-millisecond timesales (<0.1ms), and most monitors just glitch (brief blackouts) though some freezes the last refresh cycle on-screen while it's trying to get the next refresh cycle. It might even happen more often as video cables try to go faster (HDMI can now go up to 77 gigabits per second over copper. That's almost as fast as 100 gigabit Ethernet), as electronics in display decides to lag (freezeframe the last refresh cycle) instead of doing blackouts, etc. LAG, baby. It ain't 1990 Toto, we're not in Kansas anymore. Many buses in a 2020 computer have "error-detect-and-retransmit" mechanisms -- probably over one hundred of them (from USB layers, PCI-X layers, SATA layers, NVMe layers, etc, etc, etc, etc). And y'know, engineers, do you know what "error-detect-and-retransmit" means, buddies? LAG. Yes, nanoseconds, but there is such a thing as error floods and deaths by a million nanoseconds to human feelable milliseconds and microfreezes galore.
Yesterday 1985 IDE bus .... CRASH upon EMP
Today's 2020 nVME .... LAG upon EMP (error retransmit)
Yesterday's 1985 ISA bus ... CRASH upon EMP
Today's 2020 PCI-Express ... LAG upon EMP (error retransmit)
Yesterday's 1985 RS232 serial bus ... CORRUPTION
Today's 2020 USB ... LAG upon EMP (error retransmit)
etc, etc.
See...We're Blur Busters. We're truly in the temporal technology business (Hz/lag/GtG/MPRT/stutter/sync, all a matter of temporals), and thusly, we keeps an open mind on whac-a-mole of the hundreds of weak links in a system, including longshots. We're science and evidence-minded here.
Milliseconds (
many scientifically proven human visible) pays the bills that put the food on my table and pays my mortgage. We do realize that we attract a lot of "out there" people, but we also pull the Wizard of Oz curtain to reveal amazing temporal discoveries in the midst of FUD/disinformation.
I frequently have to give the Obviousman Duh walls-of-texts to many people throughout the industry, to mythbust all those tiny descendants of "Human Can't Tell 30fps vs 60fps" myths. Sometimes via great show-and-tells like a simple TestUFO animation. Whether be 360Hz monitors, 8000Hz mice, or other formerly-assumed imperceptible things. Assume is a no-no for researchers; so we actually dig and and find surprises. So, please allow me to microp that EMI is becoming an increasingly more frequent problem today in computer latency. (Excuse me.)
Historically, I'd find it nuts to put a Faraday Cage around a computer monitor, but I've seen everything -- the
LG 5K monitor EMI-sensitivity problem isn't the only culprit! For every reliably EMI-sensitive device (e.g. the LG 5K recall), there's thousands of EMI-less-sensitive situations where 1000 units out of 1000000 actually had a definite case (e.g. more-sensitive-than-usual to some form of EMI creating lagging or glitching or other non-crash effects from EMI) -- whether manufacturing tolerances varying or the users being closer to EMI superemitters. Those don't make media, those don't make the news, those often get swept under the rug as unreproducible/unexplainable. So "designed-correctly" won't work if you've got an unavoidable defacto nuclear bombesque EMI source nearby.
Also EMI is a massive universe (radiofrequency through the air versus through the wires, of all kinds of strengths, all kinds of frequencies, of all kinds of looks on an oscilloscope ranging from tuning-fork looks to 1920s sparkgap transmitter looks), so EMI isn't just "one thing", and sometimes one defective appliance EMI super-emitter is >10,000,000 times stronger at a specific frequency for a specific instant than the EMI of an iPad. Like a failing
CRT flyback transformer unknowingly shorting/arcing to the house's wiring (mains and/or phone) and turning the whole town's grid into a DSL-killing radio antenna
Or more mudane stuff such as:
- Neighbour's defective motor in a very old dryer.
- Roommate's 1970s music equipment with defective power supplies.
- Grandpa's 1940s electric fan running in workshop.
- Upstairs apartment tenant's failing fridge compressor.
- Nearby nerd's tesla coil build down the street generating intense spark-transmitter style background noise.
- Undiagnosed arcing between wires in your attic.
- The next door hospital's old MRI machine on its original power supply.
- That sketchy-looking huge rusting old 11kV power company transformer on the pole just outside your 2nd floor window.
- Large flash tubes (large xenon photo flashes, old dental analog X-ray, etc) and their associated capacitor/inductor/transformer circuit.
- Unshielded sparkplugs in a nearby old gas generator on the opposite side of the wall (exerior wall of your office).
- Or other super-EMI-emitters emitting blindingly hundreds or thousands or even millions times beyond FCC limits.
Most won't be problematic, but sometimes it is a problem (I've seen an Arduino crash from a single open-air sparkplug 2 meters away, and a RIM 950 BlackBerry pager irreparably die from a nearby Xenon flash tube/circuit two inches away! And other times, the problematic EMI sphere from the superemitter is big enough to envelope the whole residence, with no way to move your computer far enough outside the EMI sphere, or the EMI is in the wiring instead of the air. Like the example of an appliance weirdly creating unexplained microfreezes in window-dragging on a nearby computer.
Many engineers just check their circuit design through vetting software, send to fabricators such as PCBway or whatever, do a small beta, and then ship out the retail, call it a day. Don't cry to me if 1000 of the 1,000,000 retail users has a next-door neighbour having an EMI-superemiitter appliance that is doing wonky things to their setup....just because one circuit trace or one chip wasn't properly shielded, and/or the user was using a tempered-glass-sided computer tower that wasn't a complete Faraday cage (EMI via air)...and/or the power supply was defective in shielding specific power frequencies (EMI via wire), and/or multiple concurrent factors. Doh?
FCC has no God Mode that can wave a wand and make all those EMI superviolators dissappear. Unfortunately no better-than-FCC/UT/ETL/ISO/KitchenSink Approved Shielding, will be 100% completely opaque to 100% all of EMI frequencies of 100% of all EMI superemitters lurking worldwide. Lot of the superemitters are often only short-distance (inverse square law FTW!), killing only a couple computers with lag/crashes, but doing zilch for the rest of the town. FCC ain't going to hunt this particular superemitter down, since it's not causing thousands of complaints. Anyway, we can only do our best.
At some EMI-strong point (e.g. a nuclear bomb), no shielding of standard retail computer builds is immune to such strong EMI of the EMP pulse of the bomb. Shielding isn't completely opaque to all radiofrequencies; and sufficiently strong EMI of various frequencies will tend to punch through garden-variety computer shielding. But plenty of lesser EMI-superemitters (whether by air or via wire) capable of doing to a computer, lots of hard-to-troubleshoot stuff that feels voodoo.
Turn the dial somewhere between "EMI successfully blocked by shielding" and "Nuclear bomb league EMP pulse", and some computers will find the sweet spot (lagging computer) at EMI signal strengths just barely below causing computer crash. It may only be a specific EMI frequency/pattern, and immune to other EMI frequencies/pattern of the same signal strength. And yes, many systems just crash, but more often, computers now begin lagging first when the EMI is a smidge below computer-crashing-strength.
So.... "Properly Designed" doesn't five-sigma-out the FCC-violating EMI super emitters that probably exists by the millions throughout North America -- who knows how many there are -- but it's somewhere "Rare enough that it doesn't affect a single beta tester" but also "Common enough to affect a few retail users". Mathematics 101. See?
It does sounds very tinfoilhattery, even to many University Ph.D engineer (especially those who don't know how error-corrected-out-of-the-wazoo a 2020 computer is), but, alas, EMI super-emitters exist here and there, that can indeed visibly lag a specific computer, glitch a specific monitor, or create weird non-crash effects on specific devices. Super-rare, and it's probably certainly not the problem of most of the forum members troubleshooting unexplained lag (found even in window dragging and offline gaming), but it's certainly extant.
But yes. Obligatory troubleshooter wasted-time warning. This is a wild goose chase to red herrings more than 99% of the time. It's much rarer than two-sigma, but more common than five-sigma, but it's now rapidly becoming more common thanks to the more-EMI-sensitive faster speeds. And yes, there's placebos & wrong claims in the mix too, which muddy the signal-noise ratio finding trustedly actual EMI-lagged computers. And, definitely, focus on more common causes of lag first.
</MicDrop>