Memory test failure? What's that?

July 28, 2007 5:28 PM

Computingretardfilter: I've got a stubborn old desktop that I don't know what to do with. "Memory test failure", it tells me.

It turns on ok, and the bootup goes until I get to the part where it does a power/model/speed check on all the hardware. It lists my processor type and speed just fine, then it says: "Memory check" and then sequences through every kilobyte of my 524 megabytes of RAM, then restarts, then tells me that I have no primary or secondary master, one primary slave, and no secondary slave. Then it says "Memory test fail" and stops the bootup.

Can someone explain what I should do with all this in granny language, or link me to something? I've opened up a desktop before and swapped a few cards, but never to fix one. Do I have to do something with F1 and DEL and all that?

posted by saysthis to computers & internet (16 comments total)

If you are lucky, it will use cheap RAM that you can easily replace. If you are unlucky and it is really old, no one will make RAM for it anymore. Step one, though, is figuring out what kind of RAM it takes.

How old is the computer? Is it something custom built, or from a major manufacturer?
posted by Hackworth at 5:42 PM on July 28, 2007

Computer brand and model would be helpful here.

Can you hit the ESC key during the memory test and does that stop it from running through the test? If so, does the machine boot after that?
posted by Cyrano at 6:02 PM on July 28, 2007

Well the brand is 狮王科技/Lion King Technology. It says the BIOS is Award, Inc. "Award Modulator BIOS v6.00PG" and "6/11/2001 for i815EP Chipset Support". It's OOOLD. It's an offbrand Chinese thinger I picked up for almost nothing to use as a downloading computer. I'm googling that now.
posted by saysthis at 6:04 PM on July 28, 2007

Thanks for the tip about pushing ESC during the memory test. It goes past the memory test and tells me disk boot failure, boot from CD. I, um, don't have a CD. Should I buy one? I'd rather avoid wiping the HD if I can. It was having hard drive trouble before the memory died, and I'm guessing that might be causing the disk boot failure.
posted by saysthis at 6:09 PM on July 28, 2007

With the Intel® 815EP chipset, Intel delivers an external graphics solution for PC133/100 SDRAM platforms

I was curious about how much memory costs these days, so I looked it up. New stuff is absurdly cheap. The SDRAM you need is way more, like $90 for 512MB.

Swapping DIMMs is quite easy, or any decent computer shop should do it for you at no extra charge.

It was having hard drive trouble before the memory died

Oh well, there's always the chance it could be something more serious. Still, swapping memory is probably the first thing to try. But don't be too surprised if it doesn't work and you need a new motherboard.
posted by sfenders at 6:16 PM on July 28, 2007

Sounds like your primary hard drive failed. It would normally be set on the master. If it's not showing up... poof!
posted by SpecialK at 6:16 PM on July 28, 2007

During the first part of the Power On Self Test (POST), many BIOS versions offer the possibility of doing parity/address checks on the system memory, or of bypassing such tests. The POST test of memory is necessarily brief, consisting only of writing patterns of bits to available memory locations once, and reporting any errors as a failure. Just because system memory passes the POST test, doesn't mean it is "good," in stress operation, which is why repetitive free memory checking tools like memtest86 were written. But if memory is failing so badly that even the POST test routinely picks it up, something is clearly wrong.

You can do/check a couple of things:

Many old /Pentium II/Celeron/Pentium III machines used memory modules with gold plated leads in tin pinned sockets, or vice versa. Over time, vibration and age makes the dissimilar metals form insulating oxide layers that disrupt the signal flow. It's usually worth it to remove the memory modules carefully, and clean their edge contacts on both sides with a common pencil eraser. Then reseat the modules, which is often enough to get good contact in the sockets again for awhile.

If that isn't the issue, and you have more than one memory module, you can perhaps get a better understanding of where the physical problem is, by swapping the locations of memory modules, and watching to see if the memory test terminates earlier or later. However, some BIOS memory test continues to run, even if it encounters errors early in the test, so sometimes using the POST memory test is uninformative, or downright confusing.

In this case, go in to BIOS (F2 during boot, or Del, or whatever key combination is correct for your machine), and see if you can turn off the POST memory test, and set the machine to boot from floppy disk, or CD. This will usually be a setting in BIOS options, or "chipset features." This may allow the machine to boot to a floppy disk or CD-ROM with a basic DOS diagnostic system, and memtest86. You can run the memtest86 diagnostic then, which will pinpoint bad memory locations exactly. Armed with the locations, you can definitively swap module around, re-run memtest86 and discover for sure which module, or modules are bad, assuming the problem is not the motherboard's chipset memory controller, or BIOS selected timings for memory that are out of spec.

Sometimes, older machines have been "upgraded" with memory of different spec than they were designed for, and the motherboard chipset and BIOS is not smart enough to auto-detect this, or deliver the right memory signal drive strength and timings. Thus, there may not be anything "wrong" even with memory that memtest86 identifies as "bad," other than it cannot work in your motherboard, due to limitations of that motherboard. A BIOS update may help, as would using modules of matched capacity and characteristics.
posted by paulsc at 6:20 PM on July 28, 2007 [1 favorite]

I love it when paulsc suggests everything I was going to - it saves me so much time! Do what he said.
posted by flabdablet at 6:23 PM on July 28, 2007

However, some BIOS memory test continues to run, even if it encounters errors early in the test, so sometimes using the POST memory test is uninformative, or downright confusing.

In that case, you might try leaving in just one module at a time.
posted by sfenders at 6:28 PM on July 28, 2007

This sounds serious. Thanks for the testing advice, and I'll try that tomorrow. Hope this works...
posted by saysthis at 6:35 PM on July 28, 2007

In addition to the memory, it also sounds like your hard drive may be on the way out, so you should probably just give off on booting from it, and get a new (or working spare) hard drive ready, so when you get the RAM problems sorted out, you can install a fresh OS onto it.
posted by Kadin2048 at 7:01 PM on July 28, 2007

Yeah, it sounds like in addition to a RAM problem there's a disk problem. The RAM error message might even be a red herring.

If it's your disk, it might just be a jumper that's in the wrong place, or a loose cable.

"no primary or secondary master, one primary slave, and no secondary slave" is telling you about what IDE disks the BIOS has found while booting up. A typical motherboard will have two disk controllers (primary and secondary) each of which can control two disks (master and slave, which are kind of misnomer for IDE disks, just think of them as disk #1 and disk #2). Normally you boot off of the disk that's connected to the primary controller as 'master'. But there is no such disk in your system so it won't boot.

If there's only one disk in your system, you can examine it for a jumper that tells it whether it should be 'master' or 'slave', and make sure that jumper is set to 'master'. (Every disk has a different arrangement of jumpers. Sometimes the instructions are printed on the disk. Sometimes you have to google for the disk's model number.)

If there are two disks, and one is already configured to be 'master', then that disk is bad. Or maybe its cable is just loose.

As paulsc says, the first/easiest thing to try is to re-seat all the connectors. Just go through and make sure every card, module, and cable inside the case is firmly plugged into whatever it's plugged into. Maybe unplug it and put it back in again, in case there's a speck of oxidation between the contacts (they're designed to wipe the connector clean as you plug them in, so unplugging/replugging will sometimes clear up a connection problem.) Be careful of static while doing this, of course.
posted by hattifattener at 7:32 PM on July 28, 2007

I'm not 100% convinced there is a memory problem, actually.. I mean, perhaps it is even likely that there is, but for some reasons I suspect the drive problem can sometimes cause this kind of message. Of course you should still run memtest ala paulsc's advice, when you can, and etc, but.. Figure out what is going on with your hard drives.

Also, complete reassembly of the system is often the best choice in these situations. I would do that before spending money, but not before any of the other investigations suggested so far. I took a quick look for a guide to reassembly in past questions, but didn't spot anything really good. Perhaps someone else has a good link..
posted by Chuckles at 10:22 PM on July 28, 2007

I'm with Chuckles - the "memory check fail" is just a generic post-POST failure message, indicating that something went wrong during the memory or boot device check.

Backing this up is the fact that no primary master device is found, and the (correct) "boot disk failure, boot from CD" message is shown after coming out of the BIOS setup screen.

PC hardware needs to boot from a device on the IDE Primary Master drive, and you don't have one...

There are ways around that in some BIOSes, which basically involve swapping what it thinks are the primary and secondary IDE busses but, guessing at the age of the machine, that's probably not an option. Possibly I'm wrong, and there's an option in the BIOS to "set boot device" or "set boot order" or something, which was originally set to allow it to boot from the secondary master which it does see. If so, maybe the CMOS battery has gone flat / rotted away and it's forgotten that setting?
posted by Pinback at 12:52 AM on July 29, 2007

... Primary Master drive interface ...
posted by Pinback at 12:54 AM on July 29, 2007

the "memory check fail" is just a generic post-POST failure message, indicating that something went wrong during the memory or boot device check.

Okay, there might be a disk problem, but this I find hard to believe. POST is not supposed to lie to you about things like that. It seems more likely that the disk error is caused by the part of memory that it uses to read the boot sector not working.
posted by sfenders at 4:19 AM on July 29, 2007