Subj : Re: USB locking up To : Barry Martin From : Ky Moffet Date : Thu Sep 17 2020 13:05:00 BARRY MARTIN wrote: > Hi Ky! > > > > > > the Round TuIts, But Firsts, etc. You had given a link which > > > > > suggested going in to the BIOS to disable the Marvell SATA ports > > > > KM> I did? :) > > > > Sure; remember when you had amnesia?! > > > KM> Is that what I forgot?? > > > We don't recall now... > > KM> Wait, now we both have amnesia? > > Maybe - what do we compare it to? > KM> Why are you asking *me*??! > You were handy. :) But I'm way over here! [BIOS report] > KM> Did you watch if it changed significantly under load? > > Uh, no. Never really occured to me! ...So far just verified without a > change in load the voltage didn't alter significantly, which correlates > 100% with the system doesn't lock up as long as no new USB device is > plugged in! D'oh!! > KM> Oh, no, mine that has the bad Southbridge is an Asus P5B Deluxe > KM> (2008). So far Silver's new board (2014) has, knock wood, no > KM> issues. But bad southbridge does seem to be kind of an Asus > KM> thing... hopefully yours doesn't predict mine's future. > > Wonder if the Asus engineers used basically the same circuit design over > the years -- it works, why change? Or went with cheaper components -- a > half cent adds up after not too long! Sounds like it, actually. Similar problems reported across about a ten year span (went I went looking, found it was a broadly-distributed complaint). First Asus-vs-USB I've any experience of was in Double Vision, an Asus A8N-SLI (2006; AMD socket939). Supposedly USB2, but during boot it can only do USB1. Eventually USB stopped working entirely. That's the one with the Athlon 64 3200+ 2.2GHz ... supposedly a 64bit CPU, but has AMD's not-true-64bit bug. (At least they're consistent; I remember when the K6-2 had a not-true-32bit bug. Known issue and they shipped 'em anyway.) Having got the attention of the retro-gaming crowd, Socket939 CPUs are still quite expensive to upgrade, so since this system was rapidly superceded by newer and better, I never did... and just as well, because guess what -- it's recently blown several capacitors. Anyone care to guess whether they're associated with the southbridge chip?? Ya know, there might be a reason why I kinda prefer MSI boards... beyond their feature set. But weren't any in the price range at the time... and Silver's new board at least was from the higher end of things... well, hopefully it'll be lucky, and outlast its poor relations. > KM> And I think the other P5B is starting to go, as I'm getting that > KM> popping "device plugged/unplugged" sound once in a while... > KM> another early symptom. :( > > That's one thing I haven't heard: the popping noise. Something from the > board/BIOS Beep Speaker (piezo) or audio speakers or ...?? I was > initially thinking the capacitor was leaking and popped. No, this is the Windows device found/not-found sound. the little POP sound it does when a device fails to register properly, or unplugs itself. > > KM> Yours: > > KM> https://www.asus.com/us/Motherboards/M5A97_R20/overview/ > > KM> Mine: > > KM> https://www.asus.com/us/Motherboards/P9X79_LE/overview/ > > Should have read ahead (on the Intel guess). > KM> We do NOT *buy* AMD. > > I should not have bought AMD! Was an old rule of mine, don't recall > when I decided to try. Ooops. :) I got cured of paying for AMD over 20 years ago (see the aforementioned K6-2 debacle; this was the socket7 era!) and have seen no reason to regret that decision. Quite the contrary... Have been seeing all manner of complaints about Ryzen CPUs, nothing consistent but kinda looks like rushed-to-market. Wait a year and buy the equivalent Intel CPU (after the price comes down) and be happy for years instead. > > > Sort of got excited when scanned through the overview and came across > > "check your PC in real time". I'm in "monitor the voltages" mode so > > interpreting that as seeing the voltages, temperatures, etc.; probably > > is more like VNC or Remote Desktop. Wait, tell me about VNC? > KM> I don't know what can monitor voltage from within linux. > > Installing 'lm-sensors' then running 'sensors' or 'watch sensors' at > Terminal. "watch sensors" wil update every two seconds. Powewr seems > rock-stable: > +12v +12.05v > + 5v + 5.01v (though as I typed that saw it dip to 4.99 for > one display cycle) What you don't want to see is dips and spikes for no reason... > Then I noticed something: Vcore: +1.27v (min = +2.98 v, max = +2.23 > v). Waitaminute: the minimum is more than the maximum! (No, that > wasn't a typo.) Chart's got a couple of other references backwards for > the GPU and CPU. Whoops... > KM> Haha... having twin boards has been handy more than once. One > KM> dies, just plug everything into the other and life goes on as > KM> before. > > Yes, does tend to make repairs easier! Of course the problem is after a > while run out of the duplicate parts. Cannibalizing does have > advantages, though upgrading isn't one. This is a minor difficulty. > KM> Wasn't when mine was going... you wouldn't think a mouse dongle > KM> would overtax it, but apparently it did. Didn't lock up but would > KM> simply Not Work. > > I'm thinking along those lines too: it seems more the act of plugging in > is the trigger as opposed to the electrical draw. LIS the other day, I Same as one too many HDs would do to a marginal PSU -- it wasn't the running draw, it was the startup draw that would make one HD play dead. Or why staggered startup evolved in the first place. > plugged in the same card reader (and SD card) numerous times and only > once did it cause a lock up. Doesn't help either arguement: same USB > thing being plugged in, so seems should always/usually work or not work. Too many variables in a complete system. > Right. And to extend I have an external USB 3.0 hub, so external power > source, and it does the same maybe-I'll-lock-up-maybe-I-won't thing when > a device is plugged in, which is why I'm leaning towards data rather > than voltage as the trigger. And data is what the Southbridge handles... yeah, I saw the same thing with my misbehaving beast. Seemed to be data was the trigger, but data requires voltage to move... > KM> Short usually doesn't lock up, rather it causes an instant > KM> reboot. > > Guess that pretty much excludes the problem being a short, or short > duration high current draw episode! (Isn't a true short but like a > motor which dims the lights when it starts.) This was one reason I decided to cannibalize the server rather than try to use it... every time I'd power up, it brought to mind Frank Hayes... And when it's all assembled there's computer to your collar It's nice to have a micro but a mainframe would be smaller And when they turn the power on, it's sure to dim the lamps At plus and minus sixteen volts and fourteen hundred amps! Ah! Found it! https://www.youtube.com/watch?v=ow78cUDdTOg > Reminds me of the Heathkit GR-300 TV I built: just sitting there, SNAP! > and shuts off. HVPS board would arc to the chassis. First solution to > mind: round off solder pad points. Helped, but not completely. > Eventually glued a rubber pad on the chassis under the board -- solved! > Well, not a true solution but stopped the problem. Sounds like a solution to me! :D With mine originally I put some tape over the spot, but when a better case came along -- transplanted! > > (BTW, the problem wasn't with the hardware but the software. Issue was > > the software is supposed to format the SD card. Yup: I was using a 64 > > GB card and so over that limit. Switched to a 32 GB card -- no > > problems.) > KM> Huh. What file format were you trying to use? > > Default/FAT32/something else. In this specific case I don't think it's > the format but the software/utility: MotionEye. Seems to have problems > if the card is larger than 32 GB (so 64), current and prior versions. Um, no. With FAT32 the SAFE limit is 32GB, which is why original FDISK was limited to 32GB for FAT32 disks. Yeah, later versions of FDISK could do 64GB, but this was a Bad Idea from someone evidently not fully in the know, because there's a known bug (or if you prefer, limitation) in FAT32: if the partition is larger than 32GB, as soon as data crosses that 32GB barrier (either in total, or just getting written that far out on the disk) FAT32 starts eating files, in a manner that looks a lot like a disk failure. I personally experienced this, so went looking, and found Microsoft's original documentation on the problem. (Which vanished when they nuked all the old support files.) So if you are trying to make a FAT32 partition larger than 32GB, and your software refuses -- it's just trying to save your sorry butt. exFAT (Extended FAT, commonly used on flash drives) is a different filesystem entirely, more like NTFS without the journaling overhead (and without the allocation table redundancy, so if it fails, there's no recovery). Older Windows needs a patch to see exFAT. It can be used on Really Big Disks. I usually reformat 'em NTFS so everything can see it without any hoop jumping. Of course FAT32 also has a filesize limit of about 4GB. NTFS and exFAT have no such limit. > I'm not the only one but of course now can't find anything to support. > Current creation: the command line to copy the image, auto-install the > static address, WiFi, etc., is rather long so I have a template on file > and copy that to Terminal. Several attempts with the 64 GB card, one > with the 32. See above.... > KM> Ya think? this is the same board with the USB low voltage issue!! > > I'm starting to see a pattern! ...Brilliant thought: PSU issue?? > What brand is the PSU? Have you put it on a voltage tester? (one that shows voltage for each line, not just a good/bad LED.) I've become an Enermax bigot because those are the only ones I've tested that have consistent voltage without sags or spikes. And about half or 2/3rds of my stash that passed the good/bad tester... got nixed by the one that shows actual voltage. Best $15 I ever spent. Especially since a couple of the supposedly-good were spiking bad enough to kill components. The other reason I've become an Enermax bigot (tho I buy the old ones from when they were still 100% vertical) is that I've only had one die, and it had somewhere high of 20 years 24/7/365 under its belt. (And lordy, the size of the capacitors and heatsinks in there... capacitors the size of your thumb. No wonder they don't sag.) > KM> Which Lenovo board is that? > > M51-8141 KNB. ...Well that's weird: checked another file and it says it > only takes 4 GB and has two slots. Definately has four slots. Maybe I > have the wrong card in it? (It and a couple others are currently being > stored.) From what I can find, an M51 Thinkcentre (made in 2005 for WinXP) only supports a 32bit CPU, or at least that's what it shipped with, so would max out at 4GB RAM regardless of however many slots. But it's socket775, so you'd think would support at least early x64 CPUs?? What exactly CPU is in it? https://www.cnet.com/products/thinkcentre-m51/ https://www.cpu-upgrade.com/mb-Intel_%28chipsets%29/915G_Express.html Hmm, nope, all the CPUs that chipset supports are 32bit. So 4GB is the max RAM. But being kinda on the cusp of the Next Great Leap Forward, it might have some rudimentary ability to SEE, if not properly USE, more RAM. But the CPU can't address it anyway, unless you LIKE data corruption. Four slots probably means at the time 4x1GB sticks were significantly cheaper than 2x2GB sticks. (They probably all have the four slots in the PCB, just not the external part soldered on.) Lenovo seems to have some funny ideas about its own products... frex that RMA'd dual CPU board I got back as "refused"?? Lenovo support guy swears up and down it's not theirs, and he worked on developing the D20 line so he knows! BUT when I plug the model number into Lenovo search, guess what, up comes the D20. Which it doesn't look like... but BIOS (when it had one CPU, thus before it PFFZT'd) said it IS a Lenovo D20. Must be some hands that never shook other hands, is all I can think. I can't find the M51 on their site at all. Closest is the M53. > > KM> Ah, well... when I get completely moved into New Silver, Cash can > > KM> have its old job back (secondary PCLOS streaming box) and Tarnish > > KM> can run ReactOS (which runs well on Tarnish, and doesn't like it > > KM> when there's more than 4GB RAM anyway). > > Sounds like winter projects! > KM> At least, after the baseball season. > > But now they're showing reruns! Not yet! > > It almost seems like CenturyLink is setting themselves up to fail. When > KM> They don't seem to want customers anymore, that's for sure... > > That had been pretty much my thought for several years. It seemed like > they weren't making much of an effort to combat Mediacom (cable) around > here. Sure, CenturyLink is more telephone service with Internet and > television as additional services and Mediacom more TV with Internet and > telephone as additional but seems the average consumer looks at the > three options as one, or at least best bang for the buck. Or so they've been taught to do by the cable/media/phone companies. > And there have been fiber optic 'noises' for years. Some time ago the I have fiber right across the road, laid when Montana Power got sucked into Enron's scam (I won't say suckered, cuz MP execs were in on it as their exit strategy). 200 feet away and might as well be on Mars. You can actually see this stretch (well, a couple miles down the road) being laid in the Montana Power Debacle documentary on PBS. So they have fiber down at the casino at the next wide spot in the road, but not here. > City of Bettendorf was considering offering free WiFi (so Internet) and > then that faded and a bit later Metronet (fiber optic connection) made > its bid to provide service to Davenport and Bettendorf. To me it would > seem that should have shaken-awake CenturyLink: "hey, we're going to > loose customers! That's income!" but AFAICT they just kept snoozing. > Mediacom (cable) did (and continues to) actively upgrade their services > and offer some rather attractive come-on pricings. I think a lot of 'em are in exit strategy mode and don't want to increase stock value beyond what some other corp might be willing to pay to snatch 'em up. > KM> Egads... I did find the phone number for the local tech, assuming > KM> the number still goes anywhere useful.. need to get around to > KM> calling him. > > May or may not be useful. Last Spring called CenturyLink because of a He's good... knows his stuff and is thorough.. but he's also not in some damn call center. Have a friend who worked in an EDS call center for about a year. Thought he'd died and gone to hell. It was all about being useless enough to keep people on the line longer so EDS could charge overtime to HP. For this precise purpose they'd actually designed the support setup so it was impossible to get any support call done in the mandated time. You gotta wonder who at HP got kickbacks to look the other way over what was flamingly obvious to anyone with half a brain. Or, why call centers prefer to hire dumb people. > very noisy and then loss of voice telephone (oddly DSL seemed > reasonable). Two or three loose wires were found by the technician, > whom I considered excellent: spent the time to track down and fix, plus > was pleasant. Unfortunately he travelled the country -- had just come > from Hawaii (!) and was going somewhere towards the East Coast next. Didn't know they had traveling salesmen! > > So yes, might be good for you to consider switching to the wireless > > company. I'd be checking the connectivity: decent during bad weather > KM> Had fixed wireless in SoCal and it was really spotty in anything > KM> that looked like weather. Had fixed wireless in Clarkston MT and > KM> it was solid even during a blizzard. One could hope the equipment > KM> is getting better, but it's still strictly line of sight. > > And of course you don't know how good (or bad) it will be until after the > installation. Any correlation to your cell phone service? Nope, tho their radio may live on the same tower. Hey, put a tower on my hill, pretty please!! þ RNET 2.10U: ILink: Techware BBS þ Hollywood, Ca þ www.techware2k.com --- QScan/PCB v1.20a / 01-0462 * Origin: ILink: CFBBS | cfbbs.no-ip.com | 856-933-7096 (454:1/1) .