Runaway Linux Clock

Darin · 02-09-2003, 11:38 PM

System:
Intel "Providence" 440fx dual PPro200 board
Slackware 8.1, 2.4.20 kernel custom comiled for i686 SMP.
Hardware clock set to local time, usa-pacific / los angeles / PST8PDT.

The problem is that when it's running Slackware the system time creeps ahead of regular time. When the system is powered off or running NT4 the system time stays as accurate as can be expected. The difference is about an hour a day. the only anomoly I can find is that the BIOS and Linux see the 200Mhz processors running at 210Mhz. It's been this way ever since I installed linux about 6 months ago, it's not an instantaneous hour jump that can be correlated to daylight savings time adjustments.

Anyone have any ideas / theories / suggestions / answers?

finegan · 02-10-2003, 01:34 AM

As in a cat /proc/cpuinfo shows the kids running at 210 a piece. I've seen a page with the "undocumented" jumped settings for the board for overclocking, but that's just weird... I've run the same board, same chips, same kernel... I always cut out any apm garbage... basically I run with "when in doubt leave it out", that could be it, but a fast running clock is usually a sign of a dying battery and after 7 years that board's due for a new one probably... maybe part of NT's settings is a scheduled task netdate? Beats me... BIOS rev maybe? But probably not, the Providence is a tank, heck LQ ran on one for years.

just $.02,

Cheers,

Finegan

Darin · 02-10-2003, 04:30 PM

cat /proc/cpuinfo
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 1
model name : Pentium Pro
stepping : 7
cpu MHz : 210.002
cache size : 256k
...
bogomips : 418.61
...
processor : 1
...
cpu MHz : 210.002
...
bogomips : 419.43

No apm in kernel or in startup scripts, heck nothing named apm exists in the filesystem anymore. I tried unsuccessfully to get APM or ACPI to power off on shutdown and after nothing worked I took it all out of the kernel and startup scripts and removed any apm or acpi packages.

Can't remember what I did with jumpers, but pretty sure it was jumpered for 200. I had jumper info I pulled off the intranet site when I worked there (different division than the one that made the providence tho) and a spec sheet off their external page for the board as well.

BIOS is the last one Intel released for it, 1.00.09.DI0 (PR_BIOS9.EXE)

NT didn't have anything scheduled for time, it wasn't even connected directly to the net since I had another win pc with ICS before putting linux on the providence.

Battery is an option but you would think that a bad battery would cause the clock to run slow or fast or reset when the machine was powered off and that doesn't happen.

Oh well, it's not critical, guess I can add a con job for NTP or something.

finegan · 02-10-2003, 04:47 PM

I found the link about the jumper settings:

http://www.burdell.org/technotes/pr440fx.html

Gotta admit, you got me too... regardless, nice machine. Somehow with Moore's law kicking us in the but every few years, these old dualie PPros somehow come in like a '57 Chevy.

Cheers,

Finegan

Darin · 02-10-2003, 05:00 PM

Oh sweet, according to that page If I'm really running 210 then I'm at 60x3.5 but a 200 is 66x3.0 so I can probably get away with 66x3.5

If I blow the chips then I guess I run with these 180s I have, but I've got huge heatsink/fans on them now and it's in a real server case with insane airflow.

Guess I'll see if the jumpers bring it to 200 and if that fixes the clock first. Not that overclocking should screw up the system time? Kinda reminds me of C64 days where you could bring a European C64 to the US and it would run faster but so would the clock (chip timing came from AC power with Eu based off 50cycles and US on 60cycles.)

finegan · 02-10-2003, 05:05 PM

Overclocking oldschool! Offhand are these 256k cache chips, .5meg, or 1megs?

I never knew that about the C64, that's just funny.

Cheers,

Finegan

Darin · 02-10-2003, 05:26 PM

I looked for some .5 or 1meg cache chips on eBay but too expensive for an old school home LAN server. These are 200/256 (I have a pair of 180/256s also) and were just something I got out of the outdated hardware pile at work. When I was there I had a quad PPro200 box with RedHat MP on it. That one was cool but the thing was HUGE, looked like an oversized minifridge, about 3 feet tall and 2 feet square. Ran seti pretty nice tho.

Darin · 02-13-2003, 05:40 AM

update: problem seems resolved.

The board was jumpered at 200 running chips at 210, so for testing I jumpered it at 180 and it came out at 190???

that link you posted listed a jumper for

Quote:

J12B2:
5,6 shorted = normal(60,66)
4,5 shorted = turbo(63,70)

Which happens to be the same jumper block J12B2 that is used for RAM timing (1,2 = 50ns and 2,3 = 60ns.) I must have mistakenly set both the jumpers down when I switched to 50ns RAM, instead of just the top one (pins 1,2,3.)

So the next step is to figure out a stable clock speed using the regular clock jumpers, i had it running briefly at 233 in Slack but it puked running at 266 (uncompressed kernel then hung.) I figure with the case buttoned up and the dual 4" case fans augmenting the CPU cooling it shouldn't run too hot. Gotta love server cases and those old workhorse PPros

finegan · 02-13-2003, 03:57 PM

It hung on decompress at 266? That is a little weird. I remember having that issue a lot with a PPro board, but I can't remember what in the heck did it... Regardless, its nifty, keep me informed. These things are tanks though, my first foray with a dualie was a paired 180 machine that kept horking every time I ran it SMP, just random hangs. I had never played with SMP before, and the board was weird, not a Providence, but this thing from Micronics that even today won't run SMP with FreeBSD. Eventually I got edgy and just left the case open while running and finally noticed BOTH cpu fans were dead... and nothing blew up at all!

Cheers,

Finegan

Darin · 02-13-2003, 05:29 PM

I played with SMP a lot when I worked in the server lab at Intel. There's nothing like watching a 8 way Xeon box run through web server benchmarks in Solaris.

I seem to remember an SMP Micronics board that was a little weird, I think I had one running dual P120s and didn't like it. I probably eBayed it off when I got the original board that was in this case, an old Intel board (Altair) that didn't have any problems until the FDC went out on it. The Providence I've had for awhile and I had posted it up on the floor with an old HDD, a bare PS and no case a few times. Eventually I had to work this case over with a new ATX PS and some soldering to replace the AT power switch and motherboard connectors.

I've got it set at 233, seems to be stable so far. Ran through some kernel compiles with "make -j6" switch which went comparably fast and smooth and it hasn't done anything suspicious yet so I'll probably leave it here. I checked all the fans with the case off, the CPU fans work though one makes a little noise at times and the case fans and PS fan work At 240 the CPU heatsinks are still cool to the touch during a kernel compile.

The good thing is if I fry the 200s I have a matched pair of 180s sitting here for backup. That and it's old, not as spendy as if I fry my new athelon trying to overclock it.