Solve one ESD problem discover another

[Grogan]

While the new video card solved the problem of the PC needing a hard, cold boot when touching the case (or any ill effects), I discovered that sometimes it turns on when the case is touched. Bleary eyed in the morning, I shut down and went to my big sleepy chair and out of the corner of my eye I saw it was turned on again. Huh, maybe I said "-r" instead of "-h" in the shutdown command. Whatever. Then it happened again a few nights later. This morning, I came in, discharged my static on the metal table frame like a good boy and the computer powered up by itself.

Know what that is, I think it's a buggy BIOS. It doesn't know its wake settings from its ass. While there shouldn't be "wake on" anything when the PC is shut down (as opposed to sleep or hibernation states), we know from tales of old that's not always the case.

Standby power, USB power. I had it set to Wake Events controlled by BIOS, and Wake on USB enabled (not LAN or PCI-E etc.). The only reason I had wake on USB enabled was that before I had an OS, if I left it sitting on the BIOS screen it would go to sleep and nothing would wake it and I'd have to hard shut down. I actually don't need it, because I don't sleep or hibernate. Not ever.

This problem got more pronounced when I got this new keyboard. I used to sometimes notice the keyboard status lights blink in unison for a split second when powered off, when discharging on the table frame (especially if I was right underneath it) but it had never triggered a wake interrupt/event. This is a symptom, not cause of the problem.

Anyway, send interrupt (or what's interpreted as one by the bios, a signal from a device), PC powers up. Shouldn't, but does.

Boards now have "ErP" mode to conform with priggish regulations in some places. It specifies that the PC uses less than 1 Watt for any kind of standby power when powered down. When that's enabled, it can cut all power except the standby power for the "power good" signal, when the PC is shut down, if you also disable USB standby power in S4/S5 states (which I have also done). At least I think that's how it should work. There should be no more standby power for wake on anything in S4 or S5 (which is what counts here). No power for LED's... nothing.

So far, I can't reproduce it. Off means Off. We'll see what happens. I consider it a workaround that has no choice but to work

I wonder though, if this is going to kill CMOS batteries sooner. With all the intricate settings, some of them detrimental, even dangerous, that you have to (remember to) change in these UEFI bios setups, you don't want to have to replace a battery any time soon. They are also usually located such that the video card has to come out to get at them too these days.

[Zema Bus]

I have had an occasional machine that would immediately start back up after shutting it down, it was very annoying. I had to reach back and use the power supply switch to keep it off.

[Grogan]

... and let me guess, did they have MSI boards? Why the Hell would there even be that kind of bus power in S5 (powered off). That my bios doesn't distinguish between S4 and S5 (as indicated by the setting, even) is the problem. They are really the same thing except the system can resume from a hibernation file, but this shouldn't be happening. The system shouldn't be getting a wake signal from a bit of ESD either. At first I thought maybe the case connector pin block was shorting (or being "shorted" by EMI) but it's this.

Today my kernel didn't boot up. Grub loaded it but it froze while initializing. Fine on the next boot (and it could be just the kernel that goofed), but that's just not something that happens to me. I think this board's days are numbered. One part of me says just live with it and replace it when I get tired (I'm really not in any kind of mood to do that, or spend another $400 on something better at this time). Another part of me says "you'll be sooorrryyyy" if the board kills a $700 video card in 31 days. (I don't think so... that other card evidently had an internal short itself by the way it was behaving. It should not have been getting anything from ground)

[Zema Bus]

It's been a while but most likely some of them were MSI. One of them was a Supermicro board which only started doing that after I flashed the BIOS in an attempt to solve another issue.

My Gigabyte board was $170 US, so about $235 CA. It solved all the issues I had with the MSI board and gave me more NVMe and sata ports, and gave me a PS/2 port

[Grogan]

OK, at least one scenario you are describing corroborates a buggy BIOS (the supermicro) causing that behaviour.

I suppose I don't have to buy a $400 Gigabyte board, it's just what I'm used to spending on a good board.

I'm still kind of fixed on just living with this board the way it is. I'm not having any problems at all now that I've sussed the source of the thing powering up. The ErP mode is a good workaround for that. Using the system is always good, I don't have OS reliability problems or any undesirable behaviour compiling or playing games.

The kernel boot hang could have been a one-off. It was early in boot (no hand off to frame buffer yet). On my old rig, I've very rarely seen the kernel stall for a second or so while probing and sorting hardware resources, but never freeze. Whenever it happened, it creeped me out and I ctrl-alt-del'd and it never did it any time again soon. So maybe it's something similar to that, only with all those cores now and how fast the kernel loads and explodes on here now, it deadlocks or something.

[Zema Bus]

Give it to your sister for her machine and get yourself a Gigabyte board In actuality though it would probably be cheaper to get her an AMD board and processor vs a Gigabyte board for you and an Intel processor for her machine, unless you ultimately end up deciding to get a Gigabyte board down the road.

[Grogan]

Yes, we're going to go with an AMD build for her. As I said her requirements aren't special. Put it together and stock settings. Onboard graphics. 32 Gb RAM. One NVME etc.

My nephew is keen and he's looking at processors. I'm sure that whatever we choose, it will run Microsoft Word like a champ