Oldschool hardware gurus: Help with K5 system?

nd

Golden Member
Oct 9, 1999
1,690
0
0
Full story posted below. If you don't want to read it all, the basic question is: What's a good motherboard/upgrade path for an old K5 system with a shot motherboard (replacing case, p/s, motherboard at minimum)?

Hi,

I had an AMD K5 PR-100 box (with a Biostar motherboard) functioning as my LAN server. This worked great, it ran Linux, did IP masqerading for me and provided a remote shell account and handled all the server tasks without a hitch. Yeah, I know K5's are ancient, but with Linux it wasn't uncommon to stay up for several months, and served me well for many years.

I recently started getting crashes/kernel panics that are definitely hardware related. After some general diagnostics I'm convinced that the IDE controller on the motherboard is shot. The system has a desktop case (unknown "form factor", if any - it was originally a 486/dx2 box), so I want to replace it too. Basically, I'm gonna need to replace at least the motherboard and case for this box, and I'd like to go ATX. Now, the AMD K5-PR100 is for P54C (socket 5?) motherboards, but I'd rather not buy an ancient mainboard for this. From what I've been able to gather, it seems the (relatively) newer Socket 7 boards will take the P54C K5-PR100 ? I don't really want to replace the CPU, since it seems silly to throw away a perfectly good processor.. but I think getting a Socket 7 board would allow a decent upgrade path if I do decide to replace the CPU.

I'm looking at the AOpen AX59 pro as the motherboard. Am I insane to consider using a K5 in this? Is it overkill? Will it even work? Is there a better cheap, stable motherboard that will work for my K5 server?

And finally, any recommendations on a good cheap ATX mid-tower-ish size case? I really don't want to spend a lot of money (less than $200 total) for reviving this seemingly ancient box, but it needs to be fixed ASAP (downtime sucks)

Thanks in advance guys,
 

GL

Diamond Member
Oct 9, 1999
4,547
0
0
If the only thing that is shot is the IDE controller, I'd much rather purchase a cheap IDE controller like a Promise Ultra33 instead of a new mobo. I'm not even sure if the K5 will work in the mobo you mentioned (sorry to be of so little help!). I'd look for a good ol' Asus T2P4 if you can find one that supports the K5 and use that.

-GL
 

nd

Golden Member
Oct 9, 1999
1,690
0
0
GL,

Thanks for the response. Getting a new IDE controller does sound like the most reasonable idea. This is actually what my dad suggested. The only problem is that I'm not sure that it's where the problem lies. The crashes regularly occured with lots of harddrive activity (they weren't freezes, the kernel usually reported some problem with no visual connection to any particular driver).. I swapped hard drives, IDE cables, disconnected all unneeded cards, moved it from primary to secondary, etc.. still had the problem. I'm no expert, but the gut feeling here points to IDE (these problems would occur during a clean OS installation even).

Another possibility is RAM, since lots of harddrive activity usually means RAM activity.. but it seems less likely.

Maybe if those Promise controllers are really cheap (<$10) I'll just try it and find out..

P.S, like the sig :)
 

nd

Golden Member
Oct 9, 1999
1,690
0
0
Also, aren't the Promise Ultra 33 controllers PCI only? My box only has 3 PCI slots, all of which are taken (ethernet, video, sound, though if necessary I guess sound could go, but I don't know if it's worth it).
 

GL

Diamond Member
Oct 9, 1999
4,547
0
0
The Promise controllers are PCI-only. I believe the Asus T2P4 supports the K5 (please double check first though...and what revision of the board supports the chip). This was the board to have back in the day, and Tom's Hardware recently got a 500+MHz chip working on that board (considering the board came out in 1996 that's amazing). You can probably buy this board cheap off somebody in the FS/T forum or from eb*y. Then, if you feel the need for some more processing power, get yourself an AMD K6/2+ like the one in the article and you'll have a souped up 500MHz machine on your hands.

-GL
 

nd

Golden Member
Oct 9, 1999
1,690
0
0
hmm.. the T2P4 doesn't appear to be on pricewatch, so I'm guessing it's old enough that I'd have to get it elsewhere.

NaughtyusMaximus:

Yeah, I could.. except I already have 2 nics in the box (1 ISA, 1 PCI), and both were needed. I'd prefer to ditch sound (i never touch the box once it's up and running) over moving ethernet to ISA.

So it looks like my choices now are take my chances with a new IDE controller or find a motherboard still.

Yet another edit: Tom's Hardware does seem to really like the T2P4, and it seems like the perfect board for the box.. too bad I can't buy it &quot;new&quot; anywhere anymore. I'd rather not get used hardware. I'm reading the original review of the board here, and it does say it supports the K5 PR-100 :)
 

GL

Diamond Member
Oct 9, 1999
4,547
0
0
The T2P4 was the board that started Tom's Hardware. It was the first to offer the 83MHz bus setting and good ol' Tom asked everybody to send in benchmarks of Monster Truck Madness with overclocked CPUs on these T2P4s. Essentially, the T2P4 was the first mainstream overclocker's motherboard...and it started the whole Internet hardware/reviewing/overclocking movement. Besides that, it was a kickass piece of hardware and even supported USB!

If you want to find this board, I'd suggest requesting it in the &quot;For Sale/Trade&quot; forum here at AnandTech. My friend has one of these in a computer that I built for him back in 1996. Unfortunately, he had an early revision that didn't overlock. If you ever want to upgrade to an AMD 500MHz chip, make sure you get the revision 3.0 of this board.

Good luck.

-GL
 

dszd0g

Golden Member
Jun 14, 2000
1,226
0
0
Why don't you post some of your kernel panics and any related errors in the system logs so that we can confirm or contradict your proposition that it is a problem with the IDE controller.

It does not make sense to me to suggest solutions until one knows what the problem is. It could be a bad hard drive itself for all we know in which case replacing anything other than the hard drive would not make sense.
 
Oct 9, 1999
15,216
3
81
why dabble for that old proc when you set a budget of 200. Get a new processor. Your memory if 168Pins(if 72, look for 72 pin motherboard support)

You can get a:
233 Mhz Pentium <$50
FIC PA-2013 for <$80
CaseEdge case ATX <$50
grand total = $180 without RAM - assuming ram is 168pin.
shipping <$20

the reason is on the long run its better.. sure the K5 has seen better days, but you might as well get something a tad faster.
 

nd

Golden Member
Oct 9, 1999
1,690
0
0


<< Why don't you post some of your kernel panics and any related errors in the system logs so that we can confirm or contradict your proposition that it is a problem with the IDE controller.

It does not make sense to me to suggest solutions until one knows what the problem is. It could be a bad hard drive itself for all we know in which case replacing anything other than the hard drive would not make sense.
>>


The kernel panic messages would not be helpful in determining this problem to us. I've looked at them, and as I said earlier there is no obvious connection to any hardware. It's basically &quot;null reference&quot; and a buncha numbers. I'm fairly proficient in dealing with the Linux kernel so there's very little chance I would get more info about that here (unless Alan Cox or Linus Torvalds is lurking around here :))

And also, it's not the hard drive. As I said before, I tried using other drives, and even used the main hard drive upstairs in my Athlon box to back it up.. all hard drives are affected.

I'm all for narrowing down the problem, but everything so far points to the motherboard/IDE controller.



<< why dabble for that old proc when you set a budget of 200. Get a new processor. Your memory if 168Pins(if 72, look for 72 pin motherboard support) >>


Yeah, I knew someone would mention this. I really don't want to throw away perfectly good working hardware that does the job well enough for its purpose. I certainly wouldn't mind the extra speed, but I suppose I meant the $200 as a maximum budget. Spending less is certainly ideal. Old CPUs are cheap, old motherboards are cheap, but I'd rather not go crazy here.. and keep the CPU if at all possible.

Thanks for the input so far, I appreciate it.
 

dszd0g

Golden Member
Jun 14, 2000
1,226
0
0
Sorry, I must be blind. I still don't see anywhere that said you tried different hard drives. You said that you did &quot;general diagnostics.&quot; Maybe it would help us if you told us what diagnostics you did?

Also, did you try your linux drive in another system (One compatible with however you have your kernel compiled.) What kernel are you running anyways? Some of the developmental kernels have had some IDE issues. A null reference points to a driver issue typically not a hardware problem (unless its bad memory). If you are running a developmental, have you tried running a stable and seeing if the issue goes away?

Mind posting the kernel panic just to humor me if nothing else? I take it there were no errors in any of the system logs? Do you have a cron job that dmesg >> /var/log/messages?

If you are sure it is the IDE controller I would disable the IDE controller on the mobo and stick in a Promise card. Although, my personal opinion is it may not be worth putting any money into a machine that old.
 

nd

Golden Member
Oct 9, 1999
1,690
0
0
dszd0g,

Here's where I said it.. it just wasn't in the first post of this thread, sorry:



<< The crashes regularly occured with lots of harddrive activity (they weren't freezes, the kernel usually reported some problem with no visual connection to any particular driver).. I swapped hard drives, IDE cables, disconnected all unneeded cards, moved it from primary to secondary, etc.. still had the problem. >>


I never wrote down the message, and the box isn't ready right now so I can get it (it's obviously hard to get at because you can't exactly cut n paste), so I can't humor you right now unfortunately.

I did try the hard drive in another Linux box. As a matter of fact, as soon as I noticed problems I took the harddrive upstairs to my Athlon system and backed up the entire hard drive.

I was running 2.2.14, though I would get the error in 2.2.16 as well. It's VERY unlikely a software issue, as the box was up for 45 days before this crash started occuring.. and considering that the same (similar) things would happen when running the Red Hat 7.0 installer (I'd watch the virtual consoles during the Red Hat 7 install), it's probably not a case of lingering corrupt files cured by fresh-install.

I'm not aware of any IDE issues in the development kernels that didn't exist previously (I follow kernel traffic, etc), although it wouldn't matter since I'm not using them.

I monitored the system logs during diagnostics. The easiest way to crash the box was by running e2fsck (something I ran an awful lot since it was crashing.. I actually would run e2fsck inside the Red Hat 7 install from the F2 virtual console install.. had to mknod the hdX devices first of course - rest assured, this DOES work fine as I've done it many times). The only system message I recall getting occasionally once it starts going crazy is a message like &quot;XXX lied&quot;.. can't remember the exact wording other than &quot;lied&quot;, only that there wasn't any visible correlation to any specific hardware. The whole &quot;lied&quot; thing just seems to indicate an inconsistency in a sanity check, usually hardware (or memory like you said).

I'd rather not waste you guy's time with this though... so really all that's important is any upgrade path recommendations as far as motherboard/case or any ideas (like the IDE controller card, but as I said I have no PCI slots available).

later..

EDIT: I just grep'd through the kernel sources and found the message that would appear:

SIG: sigpending lied

Now, let's look at the context of the source.. this is from signal.c in the kernel/ directory

/* Sanity check... */
if (mask == &amp;current->blocked &amp;&amp; signal_pending(current)) {
printk(KERN_CRIT &quot;SIG: sigpending lied\n&quot;);
current->sigpending = 0;
}

AHA! I was exactly right about the sanity check :)

You know, this looks more like memory than IDE now. Thanks for getting me thinking about this again, dszdog. Tomorrow I'll experiment with the memory (something I shoulda done in the first place had I not been so lazy)
 

GL

Diamond Member
Oct 9, 1999
4,547
0
0
If your memory is starting to crap out on you, try to put the memory timings on a more conservative setting and see if that alleviates any problems...it'll mean a slower computer but it's not like you need a performance computer in this box. For all you know, you may just need to re-seat the SIMMs in their sockets.

-GL
 

nd

Golden Member
Oct 9, 1999
1,690
0
0
Well, I have 80 megs of RAM in there.. 64 MEG EDO and 16 of something else (older), with the EDO being much newer. The first thing I'll do is try it with just the EDO... swap around a little and then see what I can do in the BIOS. Right now I'm at work though :(