Subject: Re: Giving up on the evil 486... :/ [Sympathy]
To: Mason Loring Bliss <>
From: D. J. Vanecek <>
List: current-users
Date: 09/15/1997 21:36:48
> First off, I think I really, really despise old Wintel hardware. I know, I
> know, I'm working with old stuff, and I can't compare it with what's out
> there now, but it's still pretty obscene.

Yup. Save your pennies for a Sun.

> Check this out:
> #!/bin/sh
> echo "The following checksums ought to all be the same:"
> cd /
> cp netbsd foobsd
> cksum foobsd
> rm foobsd
> cp netbsd foobsd
> cksum foobsd
> rm foobsd
> cp netbsd foobsd
> cksum foobsd
> rm foobsd
> cp netbsd foobsd
> cksum foobsd
> rm foobsd
> cp netbsd foobsd
> cksum foobsd
> rm foobsd
> echo "I bet they were all different. Right?"
> I ran this in a chroot environment, incidentally, just for kicks. The
> results are the same outside of the chroot. I was just in a chroot because
> I wanted to pretend for a moment that I had a real, working system, with
> everything in place. So, anyway, here's the output:
> The following checksums ought to all be the same:
> 516552295 1915692 foobsd
> 959436651 1915692 foobsd
> 2634406170 1915692 foobsd
> 984535928 1915692 foobsd
> 1163808251 1915692 foobsd
> I bet they were all different. Right?
> Pretty cool, huh? I hate this hardware. It took an extreme amount of
> control to not lob it out my second-story window.

hang fire, there, pardner... Sell it to a spammer.

> If I manage to dredge up employment any time soon, I'll probably end up
> getting a Pentium-based box, at which time I'll try again. I don't blame
> NetBSD for the problems... I blame this ragged collection of decrepit, evil
> hardware that I've got. If/when I get a Pentium, I'll make sure it has a
> SCSI controller. This IDE stuff is pretty horrible. Or, I don't know. Maybe

I, too, am a recruit to this club. Better the unknowns of SCSI to the
familiar headaches of IDE. Think non-Intel. Sun? Dec Alpha? (eh, not the
lousy ones). how about a hot-stuff rack mount...

> it's just my controller, or something. It's old and generic.

                                 So am I. But I still boot every day.

> I'm seriously wondering, now... Should I give up? I don't know what's
> changed, but I can't successfully copy large files any more. Obviously
> SOMETHING has changed, but I can't say what.

It has been getting the heaviest use it has had in its life, probably.

I've seen the cksum bogosity under Linux. It occured with a machine
with what the Linux Community calls the "Signal 11" syndrome, from
its symptom during heavy CPU use, particular with lots of DMA, disk I/O
and cetera. ("gcc: cc1 exited with Signal 11 (SIGSEGV)" or words to
that heart-wrenching effect.) Bus Error is another popular variation.

This is caused, so they say, by one or more of the following:

	a) Lousy memory, with "cheap" (i.e. "logical") parity, or
	no parity at all. This is the usual reason, I think.
	Memory optimistically speed rated. Very optimistically,
	sometimes. (Just what *do* they do with those trade-in
	120ms Simms...)

	b) Broken, misdesigned/misimplemented mobo DMA.

	c) Flaky and/or overclocked CPUs

	d) Marginal hardware in general. Crummy bus design/
	implementation. Stupid interrupt scheme (noise sensitive)
	on the (E)ISA so-called bus. ... Misimplemented PCI busses.
	(I think Intel has tripped over this one). Bogus power supplies
	that drop V or float AC, intermittently, like when your
	disk is writing that superblock... Most PC powersupplies
	are, well, not exactly Mil Spec.

	e) Bogus caches. Not the *ultra* bogus L2 cache, which
	consists of dummy chips, (Play-Skool chips, like the portholes 
	on a Buick), which at least do _nothing_, but with fakey or 
	"mislabeled" (often fraudulently so) cache chips, which do _evil_.

	f) Some iconoclasts say that over-aggressive Linux code
	is responsible for part of the problem. Apparently relying
	on the published spex for DMA is over-aggressive. But, Linux
	does rupture the bleeding edge from time to time. (Why I
	run 'BSD reason #12312)

Don't suppose that box is from Packard-Hell, is it?

Now don't get me wrong, I hate BG as much as the next man, but an
awful lot, I think, of those mysterious W95 hangs and reboots could
stem from this sort of problem. Just remember: PC's are sold as
commodity items, and are *all* as *cheap* as it is humanly possible
to make them. The ones that cost more have "brand names". Period.
I would love to be proven wrong. Compare the guts of a PC with the
innards of a Sun or a Decstation or a SGI. (Yugo vs BMW) Will your PC 
support a 19in monitor on its case? etc etc. This cheapness extends, 
alas, to the electronics. They want it compatible with the typical user 
and OS, I guess.