From owner-freebsd-stable Tue Sep 19 8:29:36 2000 Delivered-To: freebsd-stable@freebsd.org Received: from gndrsh.dnsmgr.net (GndRsh.dnsmgr.net [198.145.92.4]) by hub.freebsd.org (Postfix) with ESMTP id 551E737B42C for ; Tue, 19 Sep 2000 08:29:31 -0700 (PDT) Received: (from freebsd@localhost) by gndrsh.dnsmgr.net (8.9.3/8.9.3) id IAA39533; Tue, 19 Sep 2000 08:29:21 -0700 (PDT) (envelope-from freebsd) From: "Rodney W. Grimes" Message-Id: <200009191529.IAA39533@gndrsh.dnsmgr.net> Subject: Re: Make world is dying... In-Reply-To: from "Paul A. Howes" at "Sep 18, 2000 08:36:16 pm" To: pahowes@fair-ware.com Date: Tue, 19 Sep 2000 08:29:21 -0700 (PDT) Cc: freebsd-stable@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL54 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG [Charset iso-8859-1 unsupported, filtering to ASCII...] > All- > > I think I found the problem. I have used the memory before, it was known > good. I swapped it out, and had the same problem. I tried a second, > identical, CPU with a new heat sink, and still had the same problem. I > received a thought-provoking email from one kind FreeBSD user that pointed > me to the CPU I was using: The Cyrix P-166 itself, was the problem. The > email is reproduced below: > > If it's occurring in the same place consistently, it's not a RAM failure; > I used to have a Cx-6x86L-PR166 (ie. Pre-MMX Cx) and it died at roughly the > same place in a makeworld consistently; at the time I figured that I'd > cooked the CPU in it's lifetime (was prone to overheating), ditched it in > favour of a Pentium-100 and it all worked well.... no changes made other > than the CPU. That RAM is still doing sterling service in a dual PPro system > now. Our experience (which is somewhat more than an end user) as a computer repair facility with the Cyrix PRxxx line of CPU chips is that they seem to develope strange problems over time. Under FreeBSD it is usually signal 11's, 10's and 6's (often in that order of frequency), under Windblows 9X it is usually fatal exception 0xE's and or lockups. We often find that by slowing the CPU down one step (ie run a PR166 as a PR150 or PR133) the problem goes away. I suspect that the the chip has a minor internal thermal problem that leads to electromigration related timing problems. > If that's still the case regardless, then it's either a Cyrix CPU bug or a > code generation bug in GCC that affects them specifically. > > This turned out to be exactly my case. I installed a spare Pentium-100 I > have, and buildworld is working flawlessly, so far. Evidently, there's a > CPU-specific bug in "cc1" that causes it to crash on this particular Cyrix > processor. I do not know enough about the differences between the > older-generation Cyrix processors and the Pentium processors of the same > vintage to draw any conclusions, but I would be learn more about this > phenomenon. Try putting your PR166 back in, set up to run as a PR133 and see if the problem is still there if you wish to confirm my hypothsis of the cyrix problem. > Paul A. Howes > pahowes@fair-ware.com -- Rod Grimes - KD7CAX @ CN85sl - (RWG25) rgrimes@gndrsh.dnsmgr.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message