Date: Thu, 14 Aug 1997 12:17:31 -0700 (PDT) From: Howard Lew <hlew@www2.shoppersnet.com> To: Mike Andrews <mandrews@termfrost.org> Cc: hardware@FreeBSD.ORG Subject: Re: K6-200 Has anyone successfully done a 'make world' ? Message-ID: <Pine.BSF.3.91.970814110134.25716A-100000@www2.shoppersnet.com> In-Reply-To: <Pine.NEB.3.95.970814121615.6637A-100000@mindcrime.termfrost.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 14 Aug 1997, Mike Andrews wrote: > I'll throw in my two cents on this... > > Running Intel Pentium 120 and 133 chips on the same boards worked fine. > I bought two Asus T2P4 boards and two K6-200's. The K6's were > backordered, and so I ran an Intel Pentium 120 and 133 in the boards. No > problems with "make world". Once the K6's were installed, I started > having the same problems everyone else here has: 'as' getting corrupt > input and dying, gcc internal errors, sig 11's, sig 6's.... you name it. > > One of the machines also had Windows 95 on it, and I noticed many more > GPFs than even usual... Even DOS-extender programs, like MAME, started to > misbehave and crash. I decided pretty quickly that it didn't have > anything to do with FreeBSD... > Hmmm... you definitely should not get GPFs in Win 95. Something else is flaky (MB?). > I had my vendor get me a third MB/CPU, this time as a PRETESTED set from > his vendor. Much to my surprise, the thing worked. Even more to my > surprise was that there wasn't any heat sink paste on it, and I'd also > forgotten to plug the CPU fan in. Yet it had survived half a dozen make > worlds. Weird... okay, sounds like it works when you first get it. 3 identical Asus MBs? > > I started swapping chips and fans around, and found at first that the > first two chips would run great in the new board, but after a few swaps > even the known good chip started to fail, even after everything was > restored exactly the way it had been on the pretest set. Argh. If this was all done scientifically and no errors were made in the experiment, here is a summary: 1) The first 2 original K6 cpus when paired with the 2 Asus MBs had problems. Could most likely be a MB or CPU problem. 2) The 3rd K6 paired with another Asus MB ran fine doing make-world at first. (control) 3) When you take the 2 "make world-failed" K6 cpus and pop them in this 3rd board, make-world works. Hmmm.... 4) But after swapping the cpus around none will work with even the 3rd MB. Let's look at the experiment. #2 tells us that we have one combination (mb + cpu) that seems to work fine. This is the control. #3 tells us that the first 2 cpus that failed the make-world test can pass the make-world test with a new motherboard (in other words, the cpu was not damaged, and most likely the problem with the first 2 cpu-mb sets were 2 bad MBs). #4 may be explained by reasoning that perhaps something has changed again with the 3rd MB (3rd MB is now damaged). I don't want to say that the K6 is a Asus MB killer, but it seems that everyone who has been complaining about the K6 failing make-world has been using Asus MBs. Yes, Asus is a high end board manufacturer, but there is probably a problem with compatibility with the K6 cpu. Although to some swapping the K6 cpu for a Pentium MMX seems to solve the problem, this may not be quite valid here because it does not hold all things equal. The 2 cpus have different electrical needs and are architecturally different. I have seen weirder cases that indicate that swapping a cpu is not a particularly good "control." For example, when the FreeTech F63T MB came out, the AMD K5, Pentium and Cyrix 6x86 were also available. With the onboard 256K Pipeline Burst Cache no one complained about any problems with the K5, Pentium, or Cyrix 6x86. But some people wanted 512K of PBC. For some reason the MB would work with the K5 and Pentium fine even with 512K, but the moment we swapped in a Cyrix 6x86, it would GPF in Win 95. It turns out this dual-banked cache memory was not compatible with the Cyrix even though it works fine on the K5 and Pentium. Because our distributor did not carry the Cyrix, they said it was a crappy cpu because it did not work with their mb. We even replaced the original board with identical ones and they too had the same problem with the Cyrix 6x86. I was suspicious if it were not true that we had another brand that was using 512K and did not have problems with the same Cyrix 6x86 chips. After talking with the MB manufacturer tech support, they said the cache modules we were getting are not compatible with the Cyrix 6x86 even though they work with the other 2 brands because of the way the MB is designed and cpu differences. The solution was to get LiteOn or Winbond modules (We originally used the UMC ones that came from our same distributor.) And after we replaced them with LiteOn ones everything worked rock solid with 512K for all those 6x86 customers. The moral is that I am not saying the Asus MB is bad; it just may have a compatibility issue with the K6 and swapping the cpu does not necessarily make all things equal. > > By the time I'd gotten done doing all this, Intel had dropped their > prices... so I (very grudgingly!) returned the pretest set and the other > two K6's, and got a pair of P200-MMX's. Even overclocked to 225mhz > (3x75), they've passed every torture test I can throw at it, including > many dozens of "make world"s. This probably tells us despite the fact the MBs no longer work with the K6, they still work fine with the Pentium 200MMXs. > > Specifics: > > CPU Model: K6-200 > Motherboard: Asus P/I P55T2P4 v3.10 > Chipset: 430HX > L2 Cache: 512K > System Memory: 64M true parity (tried with 32M EDO also) > HD Controller: Asus SC875 (NCR 53c875) > Hard Drives: Seagate ST15230N, Quantum Fireball TM3200S [System #1] > Hard Drives: IBM DORS 32160 [System #2] > Bios: Award v0203 (first K6-supporting rev) > Switching/Linear VRM: dunno, I assume linear based on what I've read A linear voltage regulator on your MB will most likely look like a big heatsink on the MB. A switching voltage regulator on the other hand looks like a coil around a ceramic washer the size of "Certs" candy. In the past, I have learned that if the voltage regulator (i.e. linear) on the motherboard (i.e. FIC PA2002) gets too hot (i.e. original Cyrix 6x86s), you will all kinds of weird stability problems due to its inability to get rid of the heat. That's why there was a big fus about having special cpu fans oriented in a way to blow air at the voltage regulator of the motherboard. This heat problem is unrelated to the heat problems of a cpu heatsink. But nowadays switching VRs solve this problem. > HS Grease Applied: usually; see above > Fan Type: a green one that came in a pink box :) Don't know > the manufacturer exactly, sorry > Make World: fails w/ sig 11's or with 'as' dying on garbage input > Also stability problems in Win95 > General FreeBSD Stuff: occasional sig 11's outside of make world > FreeBSD Version: 2.2-STABLE > > > -- Mike Andrews (MA12) network & systems guy, Digital Crescent, Frankfort KY > -- mandrews@dcr.net -- mandrews@termfrost.org -- http://www.termfrost.org/ > -- "Evil shall always prevail over good, because evil has better marketing..." > -- Sick of junk e-mail? Visit http://www.cauce.org/ or http://spam.abuse.net/ > > --------------------------------------------------------------------------- Shoppers Network (Support) AMD K5/K6s, Cyrix 6x86, Intel Pentiums/Pro Phone: (415) 759-8584 Email: howard@shoppersnet.com ==============================> WWW - http://www.shoppersnet.com ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.3.91.970814110134.25716A-100000>