The Old New Thing

My uncle had a similar problem with SATA drives on his motherboard.  One brand of hard drive tested worked fine.  Another brand didn't.  I never could figure it out.  Both were SATA 3 drives with 16MB cache.

hard drives will not spin if dat cable is connected the wrong way around.

Should have gotten the motherboard I got.  :-P

(It has both PCIe and AGP slots.  It's a Socket 939/DDR1 by default, but it also has a "FUTURE CPU UPGRADE" slot, and you can buy another card for that slot to make it a Socket AM2/DDR2.  I'm not going to say it's the best design, but it does allow you to replace parts piecewise.)

On the hard drive thing:  It wouldn't surprise me if it *was* power.  Just having one hard drive plugged in still draws more power than having none plugged in, and having the IDE cable plugged into the motherboard draws a bit more power than with it disconnected (no matter which end is unplugged).  The drive still does some kind of self-test at power on if it's told to by the motherboard (I hope anyway -- at minimum, the BIOS should read the SMART data out of the drive, but I suppose it may be that not all BIOSes do that).

If it still does it without the video card but with the hard drive plugged in (or some other appropriate test: removing the video card but leaving the drive in means your OS might boot, which is probably bad), then I'd say it may not be power.

I used to assemble my own computers because it was, once upon a time ago, cheaper to do so. My last computer I bought off the self because it was cheaper to do so than to assemble it myself.

By the time I get around to upgrading, everything I have is obsolete - CPU, memory, power supply, form factor. I consider myself lucky to be able to move my mouse and keyboard. I suppose it is different for those who need absolutely the latest and greatest components, but in my case, the computer is more than fast enough.

I agree with Jimbo. Check the cables are in the right way round. Don't ask me why IDE cables are "ended", but they seem to be.

>> I agree with Jimbo. Check the cables are in the right way round. Don't ask me why IDE cables are "ended", but they seem to be.

All modern 80 wires cables have Cable Select that is they automagically configure masters and slaves depending on which hdd is on the blue plug and which one on the grey one.

Unless you are going to completely trick out your PC, $600 at CompUSA goes a long way, and saves a few trips to the store

I second the vote - check the hard drive cable to make sure it's not backwards or upside down (upside down is impossible if the IDE cable is keyed, but many cables aren't) -- the IDE cable ends for the two hard drives should be very close to each other, with the end for the motherboard much farther away.

I've done that a zillion times before.

>>All modern 80 wires cables have Cable Select that is they automagically configure masters and slaves depending on which hdd is on the blue plug and which one on the grey one.

IIRC, the blue goes on the motherboard, the grey is the slave and the black is the master.

I guess I've been around too long. I got tired of installing hardware back in the early 80s and pretty much just buy a new computer when one dies. Oh I'll install a new drive, some additional memory or a new keyboard if one breaks. But as soon as a motherboard goes or more than one or two parts need replaces that is a signal to go shopping for a new system. Computer companies spend a fortune on good hardware engineers who do the work to make sure things run well together. I'm a software guy so I figure I'll worry about that and let the hardware companies solve the HW interaction problems.

I've had exactly the same "connect the IDE cable and the computer won't boot" problem. It was in an old computer in which I needed to put a new hard drive. Old hard drive: fine. New hard drive: no boot. The power supply had been replaced, although not super recently. I never did figure out what the problem was - wound up replacing the whole case - for about $200, I got a brand new computer.

The drives might have master/slave jumpers set rather than cable select.  If one or the other works fine alone, I start thinking jumper conflict.

Some IDE drives have settings for stand-alone use, while others have secondary jumpers for some BIOSes to see them.  (Western Digital drives come to mind.  Some systems need "Master" and others need "Master" + "special Jumper."

>>The Northbridge fan would buzz and sometimes spin really slowly.

Those sleeve-bearing fans are annoying for that. They are a bit quieter but don't last long. And if your fan does this "vibration+slow rotation", overheating and crashes are very near!

I guess your northbridge wasn't as robust as the processors I had, since they just crashed and didn't deteriorate.

Anyway Raymond, the next time you hear this sound, quickly replace the fan with another one.

In my case, I have replaced all the fans with either watercooling (CPU) or passive cooling (GPU, northbridge, PSU). The fans that cool the water are much bigger and can rotate so slow you can't hear them (and they are on ball-bearings so they last longer)

I think they made progress in the silence domain (the recent PC we have at work are quite quiet), so maybe *now* I wouldn't go so far to silence my PC.

You replaced the entire computer, just because 1 fan died. That's the excuse I use too ;-)

You can get new fans for your northbridge chips - zalman make some nice ones that do not need fans, and one model in particular (the NB32J) that comes with thermal paste that is glue so it'll fit any chipset. I use it myself - no more buzzing from that tiny, whiny fan.

Anyway, my advice is to buy two - you can sell the old PC on ebay to recoup some of your outlay, and also install the other on your new mobo to make it quieter.

Note that Raymond says the drives won't spin up if the cable is connected.  M/S jumpers or the drives switched on the two cable ends won't stop a drive from spinning up.  You may not be able to access them, but they will spin up.  Flipping/reversing a cable end (i.e. pin 1 connector is not on pin one on a drive or motherboard) will cause a drive not to spin up.  It's very easy to do if all cable ends aren't keyed.  Ditto if it's hard to get to the motherboard connector and you keep accidentally shifting the connector by a pin.

As others said, if your cable isn't keyed check that the same side (i.e. wire) of the cable goes to pin 1 on all drives and the motherboard connector.

"I'm thinking I would have been better off just buying a Dell."

That's the approach I took the last time I very nearly went through this and I got so much more for so much less (money, time and hassle). I think the days when it was advantageous to build your own PC are long gone.

This post started as a comment on The Old New Thing blog where he talked about the trials of upgrading...

"I used to assemble my own computers because it was, once upon a time ago, cheaper to do so."

Whenever I've bought machines I've found this this to be true with one glaring exception: software. Even now, buying a licence for Office and Windows outright is $500-600. In my experience, the difference between that and the discounted versions of Office/Windows available through Dell and Gateway eats up the cost advantage of buying hardware outright. (The same thing is true for Apple machines: Adding Office to a MacBook adds $400 to the total cost. Adding Office to a Thinkpad Z61m starts at $79 and goes up to $229. You can buy a Thinkpad Z61m with Office for less than the price of the most basic MacBook.)

I JUST went through the same exact situation.  I mean identical.  What I expected to be a $100 mobo/proc swap turned out to be a $700 upgrade of the PSU (because the newer mobos need 24 pin power PLUS a 4 pin just for the proc), RAM (because the old RAM was too old and slow to work), graphics card (because the new mobo needed PCIe) and an HDD  (because hell, I've already spent 4x the money I anticipated, might as well leave EIDE behind for SATA-300.)

I feel your pain.  I wish I'd have known this would happen this way, I'd have saved myself money and more importantly time and bought an HP at Best Buy or something.

I had a similar experience when my PSU decided that 4.1V was sufficient on the 5V line. Unfortunately I hadn't realised that current PSUs don't provide -5V, and the old socket A m/b wanted it (although I can't see what for, as it doesn't have any ISA slot). It did give the excuse to upgrade the machine to the new socket 939 board though (which as Roymond points out, is now obsolete - replaced by AM2).

I don't think that motherboards with replaceable CPU cards are a solution either. At the time of the 386/486, a number of manufacturers (ag ALR) offered them, and it was never cheaper to upgrade with one than just buying a new board for the new socket.

My PC-assembly experieince is very outdated, but I think every time I had a PC not come up at all and the hard drive not spin up with the IDE cable was a backwards IDE cable...are both ends keyed?

The only other thing I remember off-hand is that maybe the hard drive screws are too long, and grounding out on the circuit board. Sometmes that happens, but I'd don't remember the symptoms...I assume the manufacturers are still (dumbly) using the same thread pitch for hard drive screws (which should be shorter) and motherboard/case screws so that they will happily screw right in, but occasionally cause problems...

"At the time of the 386/486, a number of manufacturers (ag ALR) offered them, and it was never cheaper to upgrade with one than just buying a new board for the new socket."

My family had an ALR Powerflex in the late 80's... it was the machine we had when Windows 3.0 came out. The Powerflex was a 286 machine with a slot for daughtercards containing ALR-specific CPU upgrades. We initially bought the machine with one of these installed: a cacheless 386sx/16 daughtercard. The result was a functional but relatively slow 386sx box.  

ALR also made other daughtercards, including one containing a (L2 cacheless) 486DX/25. This was actually released at the same time as the machine itself, and made the Powerflex one of the first and cheapest 486 class PC's. It was also one of the slowest, being limited to 5MB of 16-bit RAM. By the time it was time to upgrade from a 386sx, everything else about the machine was too dated to be useful. My hunch is that the higher end daughtercards were mainly sold to people looking for inexpensive 486 machines.

I don't know for sure, but I always got the impression that ALR's later PowerVEISA systems were better thought out than the PowerFlex. They were definately not AT-class machines and could support 486-class chips.

Another somewhat related note: one of the early PC peripheral vendors (I don't remember who) made a 386-based daughtercard that fit into 80286 sockets. There were a few PAL's on the board that enabled the 386 to honor 286 bus protocols, but otherwise it ran at the same clock speed as the replaced 286 and without any form of caching. Unlike an accellerator product, PC's benchmarked slightly _slower_ after this was installed.  It was targeted at people who had a high need for 80386 specific features but lacked the money or desire to buy a new PC.

Sounds like the Heisenberg Hard Drive Principle.

While we're on the subject to DIY PC upgrades, I guess I could throw in a link to my tale of woe...

http://honestillusion.com/blogs/blog_0/archive/2006/04/24/One-thing-after-another_2E002E002E002E00_.aspx

In case someone hasn't already mentioned this:

I had a similar problem when I was building a computer for my daughter.  It turned out that I had to use Cable Select in order to get the drives to work

(I am normally kind of a control freak, and prefer to manually configure the Master and the Slave, but I absolutely could not get that to work with the MSI board that I was using).

Are you using 80-wires IDE cable ? I had problems trying to use recent hard drives with 40-wires IDE cable.

For some reason, they have the same number of pins in both cases so you need to count the wires.

Raymond, your last computer was obsolete.  Things change, and old 40-pin IDE cables may have other issue.

Please get a new IDE cable.

Just a thought, but check all the pins on both ends. Strange things can happen when a pin is crushed/bent.

Hate to see what happens when the water pump goes out on your car.

Raymond,

Are you sure you're not a victim of overclocking?  :)

http://blogs.msdn.com/oldnewthing/archive/2005/04/12/407562.aspx

Sorry, couldn't resist.

James

Ouch, ok.  Have you started simple, putting only one drive on the cable and seeing if it works first. No, try a different drive.  Yes, add a drive.

Someone mentioned screws and that reminded me of two problems I faced before.  One was with a motherboard that randomly failed... turns out I'd dropped a screw behind it and it would randomly short.  Second was with a cd-rom that worked, but not properly... one of the metals posts that supported the motherboard near the IDE cable plug was shorting against a trace.

If all else fails, try removing any nearby screws and lifting the motherboard slightly.

<aol>Me too</aol>

In my case it was the IDE controllers on the motherboard that were dying -- I'd gone through about three hard drives that failed more and more often, added an extra fan in case it was cooling, but no joy. Then the dvd writer started to flake out, and asking around _everyone_ said 'it's the IDE controller'.

So I got a new motherboard, and, yup, Socket A doesn't exist any more, it's PCIe not AGP, , new power supply, etc..

On the other hand, the old hard drives/keyboard/burner etc all worked fine, so I have a much spiffier new computer for not as much money as it might be.

<i>You replaced the entire computer, just because 1 fan died. That's the excuse I use too ;-)</i>

Me too.  The CPU fan on my motherboard is going, and I'm taking that as a sign to upgrade.

And Raymond's right, it's a quick spiral to a whole new computer.  Might as well get a sexy new SFF case while you're at it, right?  And now you start thinking about Vista, and Aero, and that's a whole 'nother problem.

Have you tried removing other parts, like the floppy, one of the memory sticks (if you have more then one), optical drives, etc...? Maybe even the video card (esp if you have an integrated video option).

The fact that they worked once seems like worse news then it never working - it sort of removes hope that you are the source of the problem. Considering you're a smart guy and this isn't rocket science, it may be time to return the motherboard / everything and start over. (or buy a laptop and be done with endless tweaking/upgrading like I'm considering!)

I had the same situation just a while ago: my motherboard was dead (capasitor leaked) and I had to buy a new one. I chose a normal Core 2 Duo motherboard, which meant I had to go from AGP -> PCIe display card, upgrade memory, change from ATA->SATA harddisk, change processor, and get rid of one PCI card, since there were only 3 PCI slots on the new mobo.

The only thing I didn't replace was the power supply: there are cheap adapters that you can use to connect the new 24-pin power connector on the new mobo to your old power supply (note though that the psu needs to be good enough and supply enough power for the new equipment, so it can be wise to upgrade it.)

Perhaps I should have bought http://www.asrock.com/product/775Dual-VSTA.htm it has an AGP slot, more PCI slots and PATA ports, and is even cheaper.

I know it's a bit late, but maybe you could've gotten a Mac - quelle irony!

Try turning the cable around.  Plug the end that is currently plugged into the hard drive into the motherboard and vise versa.

Have you tried to reset the BIOS? I don't know about the model you have, but usually there is a set of pins that you can jumper, power off, wait a bit, remove the jumper and power back on.

Since you mention that you were able to get the drives visible once, but not after, maybe the BIOS settings are screwed up?

Maybe your motherboard cannot boot from an hdd plugged into a pata port. This kind of mb has only 1 pata port and a lots of sata port.

you may return your mainboard to your retailer and get a functioning mobo...

I am SO glad I stopped doing this eight years ago when I got my first ThinkPad.

Are you sure your MB has the latest BIOS installed.  Maybe it's a firmware issue.

Throwing in my vote that the motherboard is defective.

The last straw will be that when, after 3 days of stuffing around, you actually do get the computer running, you have to sit on the phone with Microsoft Product Activation for 10 minutes to get XP activated again.

Some thoughts:

A lot of the AM2 motherboards (all of them?) are using Serial ATA now for the HD's. If your old motherboard was Socket A, you probably had one of the Ultra ATA 66/100/133 hard-drives.

The motherboard was probably built around the assumption that you'd be using Serial ATA for your hard drive. It's possible (I can't see why, but throwing it out) that the motherboard isn't properly supporting IDE protocol - perhaps it's providing E-IDE support that's not compatible with Ultra ATA. Can you try connecting a CD-ROM drive instead of a hard drive, and seeing if it boots/recognizes it? If it does, you may have to get a Serial ATA drive too.

- I recently did the same type of endless upgrade you just described, and ran into a problem with the BIOS shipped with the motherboard. A conflict cause by the type of memory and CPU installed (not sure if it's the speed, timing, or just manufacturer) was causing the BIOS to complain about . . . a USB port that was using too high a voltage!?. Anyways, couldn't boot. Solution was to turn off "stop on errors" in the BIOS, then later flash it with a later version. Point being is that there's no guarentee your MBoard manufacterer didn't ship you some defective software. Perhaps a BIOS update might help you out.

Pascal: Are you using 80-wires IDE cable?

Raymond: This exact same cable worked fine on the other computer.

Any answer other than "Yes" or "No" is a non sequitur: this is not a question about the history of the drives.

Imaginary Policeman: Had you accelerated to 40, or 80?

Imaginary Raymond: I was using the exact same gas pedal as in my last car.

Now, if they are sleeved, and have three different cooured connectors, then fine, they're probably 80-wire cables, and cable age probably won't be an issue.

But if they are flat ribbon cables, are there 40 or 80 wires in the ribbon? Count the wires, or at least ten of them, and see if you're about quarter of the way across. And if they're two small to count, then it's 80-wire!  (The number of pins in 80-wire and 40-wire cables remain the same: only the number of wires in the ribbon changes. "80-pin", though common, is a misnomer).

Odds are that at least the cable with the mobo is 80-wire, so its *probably* not the problem - but since it's a simple check, it is definitely the second thing to check after making sure that cables are keyed and connected correctly (which you've already done).

If the answer is "No" (ie, you're using 40-wire cables), then you'll need to upgrade yet another component, as your mobo probably won't speak to your drives through an oldstyle cable. I recommend those sexy, non-ribbon cables. Much better for air flow, worth the $5 or so.

Failing that, I'm with the people who reckon power problems. If you've a good multimeter, check voltage levels when you connect. Look for sudden low voltage spikes when the drives try to spin up. Try disconnecting unnecessary stuff and see if that helps. Again, powering the drives off a separate PSU from the rest of the machine should have eliminated this. But it's the third easiest thing to check.

Last option is, as you have suggested, the mobo is borked. Check for similar problems by googling the model number and checking any forums (and user comments on online reviews/shops) that discuss it. Generally these things are design or bad-batch issues, and will be known about and widely discussed.

Have you tried the drives in other machines, since you connected them to this one? A bad IDE interface borking the drives is always a bit of a risk. Finding that the drives are toast might save some time.

Just a wild guess, but it could be an APM/Powersave/Hibernate thing in the BIOS. (Because Windows can turn off your drive to save power, maybe the BIOS wants to help you save power too...)

Rgrds Henry

P.S. And yes, Brad's idea in the prev. post is good, i.e. checking with a CD-ROM. Also, try connecting both one of your IDE drives (as the master) and a CD-ROM (as the slave) on the same IDE wire.

I have a never-used Socket 462 motherboard I woulda sold you for about $50-$60.  (EPOX 8KRAI Pro.)  Oh well.

There's nothing wrong with using the excuse of a fan going bad to replace the entire guts of a computer.  Sometimes I end up replacing the case also!

">>All modern 80 wires cables have Cable Select that is they automagically configure masters and slaves depending on which hdd is on the blue plug and which one on the grey one."

Yes, but you have to make sure that you have NOT jumpered a disk or CD-Rom drive to be Master or Slave -- you need the drives to be jumpered Cable Select for this to work right.

Even if the drives are jumpered wrong, that usually doesn't keep the drives from spinning.  I haven't encountered drives not spinning before (for reasons other than dead power supply, etc.).

"perhaps it's providing E-IDE support that's not compatible with Ultra ATA."

It's a new motherboard, so why wouuld it only have E-IDE support (ATA-3, obsolete since 2002 at the latest)?

PingBack from http://www.eddiefast.com/?p=142

Raymond: Does the new motherboard have a separate (square) 4-pin power connector somewhere?

Brad: I doubt if Raymond can get the computer in a state where he can do a BIOS update... unless he has a floppy drive or can attach one temporarily.  (Although if he ever needs to add SATA or SCSI drivers, he'll need a floppy drive.)

I go with Neal's comment at 11:03 AM.

Having read all your various comments I have to say it sounds like the motherboard is shot. I know it's not really helpfull to get new hardward, but that may just be the solution.

At this point my bet is that Raymond comes back to tell us the motherboard was bad OR he comes back hanging his head in shame because it was something very simple that worked even though he thought he'd tried it 10 times already.

I don't think this is really going to help, but the IDE problems are very strange (well known technology, been pretty reliable for a long time and all...). So then, it's going to be something silly. Some silly suggestions, you never know (and no, I don't think you're stupid ;-)

- are the cables connected to the right connectors on the MB?

- are the cables connected correctly to those connectors? I remember when they did not have plastic around them, you could easily put on the cable one or 2 pins 'off', so 2-4pins of it would not connect to anything.

- they are labeled and connected according to the labels... So maybe the labels are incorrect (or what you think are labels aren't?) Just try to connect them the other way around? Can't hurt much, since you're on the edge of dumping them anyway...

- IDE isn't disabled in the bios is it?

Sometimes you can get a dud MB. Do you have a spare PCI/PCI Express IDE controller that you can try? Don't plug in the cables the wrong way, as you could short/damage both the HD and the MB. They have those keys on both ends for a reason.

I have never had a problem with drives not spinning up when connected to a controller. Perhaps you have a short on your motherboard. All the evidence you presented so far suggests the motherboard/IDE controller is broken. Let us know what you find out.

This sounds exactly like the behavior of an IDE cable that is rotated 180 degrees.

I know you said the cable and sockets are all keyed, but I've encountered cases where they are keyed wrong...check that the keying is correct - pin 1 should be labeled on the motherboard and hard drives, and the cable should have markings on one of the wires that corresponds to pin 1.

I second what Ken just said.  I've frequently come across cases where the keying was flipped, so don't trust it -- make sure you track pin 1 all the way through and ensure that pin 1 on the HD is connected to pin 1 on the MB.  (This frequently involves weird gyrations of the cable.)

Also, as some others have said, if you have a CS-type cable then you must configure your drives as CS, and vice versa.  (Though that being wrong won't cause a failed spinup, only an inability to access the drive.  Usually.)

If you are using the same drivers which were in the old motherboard, it's possible that the old motherboard (or old PSU) failed in such a way that it burned up something on the hard drive. It's also possible that the component which burned up in the hard drive failed in such a way that it burned up the new motherboard's IDE port when first connected to it.

Such "viral" hardware problems do exist, and are a pain to diagnose.

OTOH, if I were you I'd first check if there aren't any extra metal spacers shorting pins in the back of the motherboard. It's a common mistake to make, and causes all sorts of misterious symptoms.

The problem sounds bizarre...  I am curious about whether the motherboard might not spin up the drives if the IDE adapters are disabled in the BIOS?  I don't know enough about BIOS operation at that level, but it's either that or you have faulty IDE connectors...  are you able to take the PC to the retailer that sold the motherboard to try a second unit?

I feel your pain by the way... I had non-stop hardware issues when I tried to switch to a dual core Intel.  Went through two motherboard replacements and a video card swap, before switching back to Asus hardware and an AMD cpu...  worked first time :\

Nice site!

Well done!

Thank you!

yeah not spinning up for me was caused once by backwards cables, another time was not setting master/slave jumpers correctly <-- on that last one I had 2 different hd's and I didn't realize the jumper settings could vary from manufacturer to manufacturer!

I hate to say this, but I have a similar problem.  Different system, but same issue presents itself.  Sadly, that system is now sitting in the bottom of the storage closet.  I gave up and ended up buying an entirely new system...

I am hoping someone finds something that works!

Punt.

Raymond said in one of the comments that his hard drive was not screwed in. That may imply that the hard drive is not grounded to the chassi.

I have noticed lots of drives that behaves strangely and fail intermittently if they are not grounded.

So ground the drive to the chassi with metal wire or something. Any metal part of the drive should work (except the circuit board of course).

Sorry if this has been mentioned. I did some text searches through the comments but did not read them all.

Henrik: The disk drive power connector has a ground wire for 5v and a ground wire for 12v.  I don't think the drive is required to be grounded to the computer case.  

Is the computer case actually required to be grounded to earth?  Is the computer case required to be grounded to the power supply ground?  I don't think so.  Not all cases are metal, anyway.  Also, some people use rubber anti-vibration disk drive mounts.

I don't think that's Raymond's problem.

David,

I have a computer (a file server) with 8 drives. The case I have them in have special places to put the drives in. But these are plastic, so the drives did not make contact with the case. I did not think too much of it when I first put it together.

Then I started using them (by creating a raid volume and doing some testing). One drive would fail and I just thought it was broken. So I swapped it for a new one. But then that would fail too or another one would fail. I would reuse failed drives and they would work.

It was obvious the drives were not faulty. But something was. So I blamed the sata pci cards, the drivers, the software.

I don't remember exactly how I realized that the grounding was the problem. But when I did, I used metal wire to ground all 8 drives to the case.

I have been using the computer for months now without any more problems.

I don't know why it works. But it does. So I will always make sure that my hard drives are grounded to the case in the future.

Whether that was Raymond's problem I have no idea. Seems that he has given up now anyway.

I've gone over to "pay the shop $100 to move the bits across" because I have a RAID card and a pile of disks but no interest in spending a week moving stuff over. What I end up doing is building a batch file that 7-zip's whole directories with encryption, running that for a week when I'm not otherwise using the computer, then handing the box over for the upgrade. It takes a tiny bit of specifying exactly where I want stuff (mirrored boot disks, RAID5 data disk, PS scratch disk) but the overall time saving is huge.

The PC shop charges competitive prices for the hardware and $100 for the installation. All up I'm buying a week of my spare time back for ~$5/hr. HappyJoy.

(the 7-zip is because I don't 100% trust *anyone* not to run off with stuff like regsitration codes, (bought) music, customer files, but I'd like to avoid restoring 800GB from DVD).

The same thing just happened to me recently. Except for mine was socket 478. Cool to read about it somewhere else. Luckily my Power Supply had quit on me about three months before the motherboard so I was ready to go there. Also, I was ahead of the curve in 2003 and got the first generation SATA drives with that computer so I didn't have to upgrade the drive immediately.

Using newegg.com and zipzoomfly.com, I was able to upgrade to a Core 2 Duo system for around $840.

Heh.  This all just reminds me why, when my motherboard died recently, I bought an identical replacement motherboard off Craigslist, rather than upgrading everything.  Total cost: $60, and I got a whole computer for the price.  Swapped the boards over, and (except for some problems where I didn't have the heat sink on the CPU mounted quite right), it just worked -- no new setup required.

Yes, I know already.