Messages in this thread | | | Date | Mon, 29 Apr 1996 11:33:35 -0500 (CDT) | From | "Daniel A. Taylor" <> | Subject | Re: Another oops (1.3.93) |
| |
On Sat, 27 Apr 1996, Linus Torvalds wrote: > > On Fri, 26 Apr 1996, Matthew White wrote: > > > > I do have a Cyrix chip. 486-DX2/66. And you know, I got another oops on > > Apr. 23. As well, the 00000004 came up. I can send that to you too, if > > you'd like. If it makes a difference, I also a victim of the sig11 bug > > (I even got one while making clean!) but I 99.44% sure it's my cache memory. > > Ok, this thing is more or less confirmed - there are some serious > problems with Cyrix chips. Not nice. > Mainly a heat problem with kernels >1.3.5x. Nothing that a good fan shouldn't fix quickly. I have seen similar reports on undercooled AMD chips as well, and I expect that even Intel chips aren't immune.
The symptoms that I have noted are that a kernel compilation (or similar operation) will fail at a particular point with some consistency.
I have personally gotten 1 Oops, 10 sig11's and a couple of hangs this way. Underclocking the chip is also a viable solution if you cannot get a usable fan (I ran a CX DX2/50 as a DX2/40 for a couple of months until I was able to get a good replacement w/fan.. it worked).
So it is in part a kernel problem, but not one I would trade away. The higher performance we are getting out of the new kernels is quite simply harder on much of the hardware out there. We _could_ make Linux run as reliably as DOS, but then it wouldn't be as fast. (maybe #ifdef BROKEN_HARDWARE #define SLOW_EVERYTHING_DOWN #endif ?)
Well, enough of a rant, thank you for reading this far.
Dan Taylor I fear no weevil.
> That doesn't mean you should all throw out Cyrix - there have been other > manufacturers with buggy chips (including intel), and I'd be happy to try > to come up with a fix for this problem. The real problem right now is > that we don't know _what_ causes the Cyrix problem, so trying to fix it > is pretty much hunt-and-peck unless somebody has some cyrix contacts? > > If anybody knows anybody that works for Cyrix, could you please speak up? > Doesn't have to be technical personell - we can try to get the addresses > of some really technical people from the marketing people if anybody > knows those scumbags ;-) > > I wouldn't like to ask for this information on the announce newsgroup (if > we can do this silently without upsetting a lot of people I'd be > happier), but that would be one way to reach a lot of people... > > Linus >
| |