| systemdlete | here is a riddle: I have 3 boxes, each with 4x8GB = 32GB memory. 2 of them run daedalus, 1 chimaera. But one of the daedalus machines only shows 16GB in free or /proc/meminfo. I've checked vm.overcommit_* and they are set the same (kbytes =0, memory =0, and ratio = 50). | 01:04 |
|---|---|---|
| systemdlete | sysctl -a |grep vm.overcommit | 01:04 |
| systemdlete | I just double-checked the manual for the MB, and it does support 32GB | 01:05 |
| systemdlete | channel pair or the sticks in the channel are bad? I suppose I can run diagnostics on it, but I am curious why the system seems to work fine otherwise. | 01:06 |
| * systemdlete goes to check the BIOS on that system... | 01:06 | |
| systemdlete | even dmidecode shows 32GB | 01:07 |
| systemdlete | but I'll boot into the bios, and then run diags | 01:07 |
| systemdlete | well, this is interesting. The bios says it only has ~16GB installed / usable memory | 01:10 |
| systemdlete | how do I run into these problems? | 01:13 |
| systemdlete | the northbridge settings indicate that ECC and DRAM ECC are disabled. Not sure if I even have ECC RAM. | 01:13 |
| systemdlete | (I don't think it is) | 01:13 |
| systemdlete | Board is M5A78L-M PLUS/USB3 | 01:17 |
| systemdlete | Latest bios installed... | 02:06 |
| systemdlete | guess time for diagnostics... | 02:06 |
| systemdlete | it's time for | 02:06 |
| plasma41 | systemdlete: Swap memory between two systems and see what, if any, difference it makes? | 05:29 |
| systemdlete | plasma41, actually, I'm doing something similar, though your idea may be better. I took out a pair of the four, in alternating slots and testing each pair separately. The first pair is almost done with 3 passes of the memtest. Then I'll test the other two in the same two slots. If there are no failures, I'll try testing each pair in the other 2 slots. | 05:43 |
| systemdlete | taking down 2 systems atm would be inconvenient, but I'll keep your suggestion in mind! | 05:44 |
| systemdlete | I notice a tiny bit of dust on the metal shield, near the air vents on one of the sticks; it is the stick that was closest to the fan on the drive rack. It is pretty tight, and the power supply cables also pass by that way also. Heat? I'm guessing. | 06:03 |
| systemdlete | I mean, not the dust causing heat, but maybe trapping some badly needed air circulation. | 06:03 |
| systemdlete | So after spending the last 1/2 a day eliminating memory sticks and methodically trying different sticks in different slots, ,memtest86 could not find any defects in any of the configurations I tried. So I reinserted all 4 sticks, and now the BIOS is reporting the full 32GB again. | 16:28 |
| systemdlete | go figure. | 16:28 |
| systemdlete | I'm guessing that it might be that when I inserted a couple of disk drives in that box a few days ago, I might have accidnetally disturbed one of the sticks (nearest the edge of the board). Those memory sticks do lock into their slots, but there are a lot of cables nearby in a tight space. The multi-drive bay has a huge 140mm fan protruding from the back of it not far from the board either. | 16:38 |
| systemdlete | Maybe I jiggled a cable just enough to pull one of the sticks out of its slot. Or, maybe too much heat? Strange, I've been running this box for a couple of years now without a problem like this, so I'm leaning to the former explanation. | 16:40 |
| systemdlete | Anyway, I'm learning to be ever more cautious with my hardware... | 16:40 |
| systemdlete | Just to be sure, I'm running memtest86 once more this time with all 4 sticks inserted. | 16:41 |
| greenjeans | It's always the little things, i feel your pain...as my father once told me when I was just a boy : "Nature always sides with the hidden flaw" | 16:41 |
| fatal | is it just me or is the devuan installer always this sluggish? installing on arm64. | 18:18 |
| plasma41 | systemdlete: Whenever a computer is not seeing the full complement of RAM, often the simplest and quickest effective fix is to simply re-seat all of the DIMMs by just removing them completely, then inserting them right back into place. | 18:57 |
| fatal | it looks like adding "vga=normal fb=false" as a boot parameter to the vm-netinstall sped things up clearly | 20:04 |
| systemdlete | plasma: No kidding, lol! Unfortunately, I had to find that same fix the long way. | 21:24 |
| systemdlete | (but thanks) | 21:24 |
| systemdlete | When I began my investigation, those seating clips didn't even occur to me. | 21:25 |
| systemdlete | Anyway, it appears that all is well again--I'm on the 3rd pass of memtest, which should finish shortly. | 21:27 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!