Troubleshooting the System Memory

This section of the Troubleshooting Expert deals with memory-related problems, referring specifically to the main system memory

I suspect that there is a failure with the system memory

• I have a new or recently-upgraded system and my system memory appears to have failed or is not working in general
• I have an existing system that has not been recently worked on or upgraded and my system memory appears to have failed or is not working in general

I am having problems with parity errors

• I am receiving a parity error as soon as I boot up the PC
• I am receiving a run-time parity error during operation of the PC, after booting

I am having a problem with some of my memory not being recognized by the system

• I have over 64 MB of memory. All of my system memory is recognized under Windows, but some of it is missing when I boot DOS
• When the system boots up, it reports either 128 KB or 384 KB less memory than is in the system
• Not all of the memory that I have installed in the PC is being recognized. The system thinks I have megabytes less memory in the PC than I do

I am encountering errors indicating that I am out of memory when I think I really have enough

• I am trying to run a DOS program and it says it doesn't have enough memory to load, but I have 16/32/64/etc. MB of memory and it only needs 600 KB!
• I am running Windows 3.x and am getting "Out of Memory" errors, but when I go the Program Manager it says that I have many megabytes of free memory

I have a problem related to the performance of the PC after adding more memory to the system

• I added more memory to the system but I didn't see any improvement in performance
• I added more memory to the system and performance got worse!

 
I have a new or recently-upgraded system and my system memory appears to have failed or is not working in general

Explanation: There is an apparent failure of the memory on a newly built or upgraded system. This means either that diagnostic tests or troubleshooting procedures have identified the memory as being suspected of being bad, or the system memory is one possible cause of an unknown system problem.

Diagnosis: There are three common categories of causes of memory problems on a new system. The first is improper configuration, or using the wrong type of memory. The second is incorrect installation. The third is hardware failure of the memory itself; since there is so much cheap junk being sold to people who only look at cost when making a purchasing decision, it is more common than ever to end up with defective RAM.

Recommendation:

• First, check for possible general problems associated with new systems. One of these could be causing the memory to appear to be bad when it is not.
• If you happen to have a duplicate set of memory modules available that are just like the ones you are trying to diagnose, then try swapping the spare modules in to see if the problem goes away. If it does, put the original memory back in to see if the problem returns. If it returns, you can feel very confident that the original memory itself was bad. If you don't have any spare memory for this sort of test, just continue with the suggestions below.
• Make sure to carefully check the memory modules, which sometimes appear to be inserted correctly when they are not. Make sure they are all sticking up at the correct height from the motherboard. Ensure that the modules have been pushed all the way into their sockets and that the spring clips have snapped into position properly to hold them.
• Check for loose connections within the PC.
• Check the inside of the case for possible overheating problems.
• If you have just installed a new 32-bit operating system where before you were just using DOS, this may bring a memory problem to the surface. When this happens, the bad memory was probably there the whole time, but DOS is much more forgiving of bad memory than Windows 95, Windows NT or other 32-bit operating systems are.
• If you installed used memory, or installed into an older motherboard, the problem may be dirt either on the modules or in the sockets. Try removing the modules and gently cleaning the contacts with a soft cloth. If the motherboard is in an existing system then it has had a chance to collect dust in the sockets. Clean the dust away using either compressed air or a PC vacuum.
• If you are using SIMMs, ensure that you have installed a full bank of memory. For a 486-class motherboard using 30-pin SIMMs, you must have 4 identical SIMM modules. For a Pentium-class motherboard using 72-pin SIMMs, you must have 2 identical SIMMs. For DIMM memory a single DIMM is a bank so this does not apply.
• Some 486-class motherboards that have both 30-pin and 72-pin SIMM sockets have restrictions on the extent to which both can be used. Some Pentium and later motherboards that have both SIMM and DIMM sockets can also have restrictions. For example, you may only be able to use either one set of sockets or the other, or you may only be able to fill some of the sockets depending on what you are using. Consult your motherboard documentation for exact guidelines matching what you are using.
• Make sure that you have used the correct sockets. Motherboards have multiple sockets and putting modules in the incorrect ones will often cause problems. For example, most Pentium motherboards have four SIMM sockets, which make up two banks. If you put the SIMMs in the middle two sockets then you have accidentally put memory into half of the first bank and half of the second bank, and the PC will not boot. Putting the memory into the full second bank instead of the first won't work on many motherboards either (but it will on some).
• Ensure that the size of memory modules you have selected is supported. Some motherboards will not support certain sizes of modules; in particular, 2 MB, 8 MB and 32 MB SIMMs are generally composite and will not work in some machines. Consult your motherboard manual.
• Watch out for 16 MB non-parity SIMMs that have chips on both sides of the SIMM (parity SIMMs usually do). Some of these are showing up on the market that are actually composite (even though 16 MB SIMMs aren't supposed to be) and they will not work properly in a system expecting non-composite 16 MB SIMMs only. Ask for the SIMMs to be replaced with non-composite versions.
• Check the technology of the memory you are using. Whether a motherboard supports memory types such as EDO or SDRAM for example, depends on the chipset used on the motherboard, as well as how the motherboard itself was implemented. Using EDO memory in many older machines will cause them to lock up.
• If using SDRAM, make sure you are using the right type. SDRAM comes in 2-clock and 4-clock varieties, and some motherboards require only one kind or the other. Some motherboards also require SDRAM that has the "serial presence detect" EEPROM on it. See this section for more details.
• Support for EDO memory is not found in 386 and earlier machines. 486 class machines fall into three categories: some will allow it and use it properly as EDO (these usually have a BIOS parameter to enable it); some will allow it to be used but will use it as regular FPM memory; and some will not work with it at all.
• Some 430HX motherboards come with DIMM sockets. Most DIMM form factor memory is SDRAM, but SDRAM will not work in these boards because the 430HX chipset does not support it. These slots are intended for DIMM EDO memory.
• If your system uses DIMMs, make sure that you are using the right kind. DIMMs come in different voltages, and buffered and unbuffered versions.
• Many motherboards that support both DIMM and SIMM memory will malfunction if both are used on the board at once. This is because most DIMMs require 3.3 volt power, while SIMMs run at 5 volts. When both are present, the DIMMs are fed 5 volts and problems can result. Try the system with only one type of memory.
• If you have an older motherboard, especially in a 486 or earlier system, there is a chance that your motherboard requires jumpers to be set when adding memory to the PC. If this is the case then failing to change the jumpers may result in either the memory not being detected, or in incorrect operation.
• Some systems require a special BIOS setting to be enabled when using more than 64 MB of memory. If you are trying to use more than 64 MB, check for one of these settings and enable it if necessary.
• Mismatches between different banks of memory can cause problems. For example, some systems only like when all the memory in the machine is identical, even in different banks. If you use memory that is of different speed or if one bank is using FPM and another EDO memory, your motherboard may not be happy. Many newer systems allow this with no problem, but some older ones do not. You may need to try changing around the memory configuration. Try to see if the machine works with either bank installed but not both together. If this is the case, then the two banks of memory are incompatible.
• Some PCs use proprietary, special modules; for example IBM's PS/2 systems. Using industry standard memory in a machine that requires special modules, or vice-versa, will cause problems.
• Older 386 and 486 motherboards use 30-pin SIMMs. These SIMMs generally come in two versions: 9-chip SIMMs or 3-chip SIMMs. The 9-chip version uses all the same DRAM chips, while the 3-chip version uses chips of different sizes. Some PCs do not work properly with the 3-chip SIMMs; you need to consult your motherboard manual or technical support for the system. See here for more.
• There could be a problem with the relative speed of the memory modules compared to the timing settings (memory access timing or wait states) that were entered in the BIOS setup program. Double-check the speed of the modules you are using to make sure it is fast enough. Some PCs will work with slower memory modules, but you may need to increase wait states or slow down the memory timing. See here for more.
• Some motherboards will work with EDO memory installed, but only if EDO support is specifically enabled through a BIOS parameter. You may need to boot the system with regular fast-page mode memory to get into the BIOS setup, change the setting to EDO, and then shut the machine down and replace with the EDO memory.
• There could be something wrong the memory modules themselves. Note that bad memory will often pass the BIOS memory test at boot time, and will also often pass the tests performed by those small module testers that many vendors use. Those tests are very superficial and will not catch all memory problems. If you can, try the modules in another PC that uses the same kind of memory. If you have performed all the checks listed in the points above, and the memory works in another PC, the memory itself may very well be bad. Try to replace the memory and see if the problem goes away.
• There could be a problem with the power supply, though this is unusual. A bad power supply can cause strange memory errors that crop up because the memory is not getting enough power.
• Check in this section that discusses memory problems with existing systems, for more possible ideas that could account for the failure (though they are less likely).
• There could be a motherboard problem. If double-checking all the settings and replacing the memory does not fix the problem, there may be a bad motherboard or a problem with how it is configured. Troubleshoot it here.

Explanation: There is an apparent failure of the memory on an existing system. This usually is implied by troubleshooting procedures that have identified the memory as one possible cause of an unknown system problem.

Diagnosis: Outright memory failures coming up out of the blue on existing systems are unusual; most memory problems occur when a system is first installed or when it is upgraded. A failure on an existing system usually means that there is something wrong with the memory itself, because there aren't as many other possible explanations for the problem as there are with a new system. Overheating of the memory modules is a common cause of hardware failures.

Recommendation:

• One good first step to try when encountering apparent memory failures is to try running a comprehensive memory test. Using a diagnostic software tool, run a continuous loop memory test for a long period of time and see if it finds any problems with the memory. If it does, this may give you a suggestion of where the problem in memory is.
• Overheating of the system is a common cause of sudden failures of previously-working memory. Read this section, which discusses different ways the system can overheat. If the system case or the memory itself is overheating, you will need to address this or the problem will recur.
• If you have just installed a new 32-bit operating system where before you were just using DOS, this may bring a memory problem to the surface. When this happens, the bad memory was probably there the whole time, but DOS is much more forgiving of bad memory than Windows 95, Windows NT or other 32-bit operating systems are.
• Make sure that your BIOS settings have not been changed. Double-check the ones that are related to the system memory and make sure that they are correct.
• Something inside the PC may have come loose. Check for loose connections within the PC.
• If the memory modules being used in the PC do not use the same metal (either gold or tin) as the sockets they are in, it is possible over time for a chemical reaction to develop that can lead to poor contact and eventually, memory problems. This will typically take months or even years to show up. If you suspect this problem, power down and unplug the PC and take out one memory module. If its pins are gold and the socket is tin (a silvery color) or vice-versa, this may be the problem. If so, remove all the modules and clean them and the socket.
• There may be a problem similar to what is seen when new memory is installed; see the troubleshooting section for memory problems in new systems. Many of these ideas don't apply to an existing system because for example, if you use an unsupported memory type you normally won't get past the first boot of the PC without having a noticeable problem. However, some systems can react in strange ways and problems may not become noticeable until later on, in theory.
• There could be something wrong the memory modules themselves. Note that bad memory will often pass the BIOS memory test at boot time, and will also often pass the tests performed by those small module testers that many vendors use. Those tests are very superficial and will not catch all memory problems. If you can, try the modules in another PC that uses the same kind of memory. If you have performed all the checks listed in the points above, and the memory works in another PC, the memory itself may very well be bad. Try to replace the memory and see if the problem goes away.
• There could be a problem with the power supply. A bad power supply can cause strange memory errors that crop up because the memory is not getting enough power.
• There may be a problem with a component on the motherboard, or another part of the PC

Explanation: A parity error is occurring on a PC as it boots up. The parity error will typically occur as soon as the system completes the power-on self test; the memory count that is performed will typically pass without any problems being recognized. In some cases, when booting up Windows 95 or Windows NT for example, the parity error will not appear until the graphical user interface of the Windows operating system itself comes up on the screen.

Diagnosis: The most common cause of parity errors when first booting up the PC are incorrect configuration or using the wrong type of memory. It is unusual for an actual memory failure, of the type that parity checks for (meaning, you wrote one value into memory and read back another value with a bit changed) to be encountered at the start of booting, although it is possible.

Recommendation: Look on the screen to see if the system is giving you any sort of memory address that indicates where the parity error is occurring. Reboot the system and see if the same address comes up again, and then reboot a third time. Take note of whether or not the memory location changes, and then continue below:

• If the parity error is coming up as soon as the BIOS tries to boot, and especially if it fails at memory address "0000" consistently, this is a dead giveaway of trying to use non-parity memory in a parity system. Make sure that you have used real parity memory if you have parity checking enabled.
• If you are or were running with parity checking disabled, double-check the BIOS setting to make sure that it is still set as disabled. If it is enabled accidentally, parity errors will result.
• I have encountered a defect in the Abit IT5H motherboard, version 1.5, which will cause it to fail when using parity memory--it just does not work in this board due to a design flaw. Every time parity is enabled a parity error is generated immediately at boot time. It is possible that other motherboards may have similar problems. The only solution is to replace the motherboard or run with parity checking disabled.
• If you are trying to run with ECC enabled on a motherboard that supports both parity and ECC, change the ECC/parity BIOS setting to straight parity and try to reboot. If the parity error goes away, and returns when ECC is re-enabled, the chances are high that you have false parity memory in your PC. False or "logic" parity memory is designed to fool the standard motherboard parity circuits but will not work if you set the machine to run with ECC. The only good solution is to replace the memory.
• Make sure that the modules you are using are appropriate for your board. In particular, there are some motherboards that will not support the newer ECC-only modules. See this section for more in the differences.
• On a new PC, it is possible that you have an actual hardware failure. If the parity error address remains the same or is in the same general vicinity each time it is encountered, this implies a failure of the memory itself. Troubleshoot the memory itself here.
• Look in the section on parity errors encountered during operation. It is possible that one of the causes there could be responsible for a parity error encountered during the boot (just less likely than the causes listed above).

Explanation: A parity error is occurring on the PC as it is being used. The parity error may occur consistently or intermittently. It is also normal to see parity errors in Windows but not in DOS because Windows exercises the memory much more than DOS does.

Diagnosis: Parity errors are often a signal that something is working incorrectly in the system. This is commonly a problem with the memory itself, but can be caused by many other sorts of hardware problems as well. You may be surprised at some of the hardware problems that can lead to memory corruption, including expansion card issues, resource conflicts, etc. This is why parity checking is so important.

Recommendation: Look on the screen to see if the system is giving you any sort of memory address that indicates where the parity error is occurring. Reboot the system and see if the same address comes up again, and then reboot a third time. Take note of whether or not the memory location changes, and then continue below:

• You may be surprised to hear me say this, but if your system has been working well for a long time, you haven't changed anything recently, and you only encountered a single parity error, I often recommend just ignoring it. Well, don't ignore it, but don't take any action other than seeing if it happens again, and if so, under what circumstances. The reason is that fluke corruptions of memory occur, sometimes due to stray radiation or odd power glitches. You may have a single error and never another one. If the error recurs, of course, you have a problem that needs to be addressed.
• Check for resource conflicts, particularly if you have just added any hardware to the system.
• You could have a real problem with your memory. Diagnose the memory itself here.
• There are many different possible causes of intermittent system lockups and glitches. Many of the problems that manifest themselves as lockups, crashes or spontaneous reboots on a non-parity system will show up on a parity system as parity errors.

Explanation: All of the system memory in the PC can be accessed and used under Windows, but only 64 MB of it can be seen when booting to DOS or using DOS programs.

Diagnosis: Standard DOS cannot access more than 64 MB of memory due to limitations with the extended memory manager it uses.

Recommendation:

• There are very few DOS programs that really require 64 MB of memory. You may want to see if a Windows version is available.
• Apparently, some memory managers such as QEMM can address over 64 MB of memory, but I have not verified this myself

Explanation: The system is reporting approximately the right amount of memory when it boots, but is leaving out either 128 KB or 384 KB in its report.

Diagnosis: This is normal for many types of machines. The "missing" memory is the area of RAM between conventional memory and extended memory, called the upper memory area. It is not available for use by most standard programs, and therefore some BIOSes choose not to report it.

Recommendation: No action required

Explanation: When the PC is booted up, it does not show that it is detected all of the memory in the machine. Several megabytes, usually an amount equal to the size of a memory module or half a memory module, are missing. For example, you may have two 8 MB SIMMs in the PC but the system is only reporting 8 MB of system memory.

Diagnosis: This is almost always caused by using the wrong type of memory, incorrect installation, or incorrect configuration. Putting the modules into the wrong sockets, forgetting to change jumpers that may be present, using unsupported technology or unsupported module sizes are common reasons that memory would be not detected.

Recommendation: Most of the usual causes of memory not being recognized can also be causes of more general memory failures, although certain potential causes are more likely to be at fault in one situation than the other. Assuming that this is a new PC build or upgrade, you may want to look here to troubleshoot the memory as a general failure. This is especially so if the system isn't booting or is exhibiting strange behavior. Below I have several of the causes of memory problems from the general failure section, modified and rearranged so the more likely candidates of memory non-recognition are listed first. I have also excluded some of the causes of outright memory failure that are not generally found when you have memory simply not being recognized:

• Check for possible general problems associated with new systems. One of these could be causing the memory to appear to be bad when it is not.
• Make sure to carefully check the memory modules, which sometimes appear to be inserted correctly when they are not. Make sure they are all sticking up at the correct height from the motherboard. Ensure that the modules have been pushed all the way into their sockets and that the spring clips have snapped into position properly to hold them.
• Check for loose connections within the PC.
• One of the most common causes of unrecognized memory is using an unsupported module size. Ensure that the size of memory modules you have selected is supported. Some motherboards will not support certain sizes of modules; in particular, 2 MB, 8 MB and 32 MB modules are composite and will not work in some machines. Consult your motherboard manual.
• A partial bank of memory, if it is not the only bank in the PC, is often ignored by the system (although some systems will refuse to boot in this situation). If you are using SIMMs, ensure that you have installed a full bank of memory. For a 486-class motherboard using 30-pin SIMMs, you must have 4 identical SIMM modules. For a Pentium-class motherboard using 72-pin SIMMs, you must have 2 identical SIMMs. For DIMM memory a single DIMM is a bank so this does not apply.
• Some 486-class motherboards that have both 30-pin and 72-pin SIMM sockets have restrictions on the extent to which both can be used. Some Pentium and later motherboards that have both SIMM and DIMM sockets can also have restrictions. For example, you may only be able to use either one set of sockets or the other, or you may only be able to fill some of the sockets depending on what you are using. If you put memory into all of the sockets on such a machine, you may find some of it ignored. Consult your motherboard documentation for exact guidelines matching what you are using, and look on the motherboard for any jumpers that control memory configuration. Try using only one bank of memory at a time.
• Watch out for 16 MB non-parity SIMMs that have chips on both sides of the SIMM (parity SIMMs usually do). Some of these are showing up on the market that are actually composite (even though 16 MB SIMMs aren't supposed to be) and some systems will only recognize them as being half of their true size. Ask for the SIMMs to be replaced with non-composite versions.
• Make sure that you have used the correct sockets. Motherboards have multiple sockets and putting modules in the incorrect ones will often cause problems. For example, most Pentium motherboards have four SIMM sockets, which make up two banks. If you put the SIMMs in the middle two sockets then you have accidentally put memory into half of the first bank and half of the second bank, and the PC will not boot. Putting the memory into the full second bank instead of the first won't work on many motherboards either (but it will on some).
• Check the technology of the memory you are using. Whether a motherboard supports memory types such as EDO or SDRAM for example, depends on the chipset used on the motherboard, as well as how the motherboard itself was implemented. Using EDO memory in some machines can cause the memory not to be recognized.
• Some 430HX motherboards come with DIMM sockets. Most DIMM form factor memory is SDRAM, but SDRAM will not work in these boards because the 430HX chipset does not support it. These slots are intended for DIMM EDO memory and DIMM SDRAM memory may be ignored.
• If you have an older motherboard, especially in a 486 or earlier system, there is a chance that your motherboard requires jumpers to be set when adding memory to the PC. If this is the case then failing to change the jumpers may result in either the memory not being detected.
• Some PCs use proprietary, special modules; for example IBM's PS/2 systems. Using industry standard memory in a machine that requires special modules, or vice-versa, will cause problems.
• There could be something wrong the memory modules themselves. Note that bad memory will often pass the BIOS memory test at boot time, and will also often pass the tests performed by those small module testers that many vendors use. Those tests are very superficial and will not catch all memory problems. If you can, try the modules in another PC that uses the same kind of memory. If you have performed all the checks listed in the points above, and the memory works in another PC, the memory itself may very well be bad. Try to replace the memory and see if the problem goes away.
• There could be a motherboard problem. If double-checking all the settings and replacing the memory does not fix the problem, there may be a bad motherboard or a problem with how it is configured.

Explanation: Despite having a great deal of memory in the system, DOS programs are complaining and saying that they need more memory in order to load.

Diagnosis: The problem is likely caused by insufficient conventional memory. The conventional memory area is the first 640 KB of system RAM, which is what DOS uses when running standard programs. This is the most "hotly contested" memory real estate in the PC, and a program that needs more of it in most cases cannot use the megabytes of extended memory that are in your machine for this purpose.

Recommendation: Run memmaker or other such application to increase conventional memory.

Explanation: Windows 3.x is saying that it is out of memory, but when I check the available memory by looking in the "About" section under "Help" in the Program Manager, it says there is a lot of memory free.

Diagnosis: This should win an award as one of the most ridiculously misnamed error messages ever. Most times when you get an "Out of Memory" error you are really running out of Windows 3.x system resources, not actual memory. When you get this problem, take a look in the "About" box and near the amount of memory free you will see the current level of "system resources". You will usually see that the number is small, less than about 20%.

Recommendation:

• Close one or more programs to free up more system resources for the program you want to run.
• Exit Windows 3.x and restart it. Windows 3.x will "leak" system resources over time and usually has to be rebooted frequently if it is heavily used.

Explanation: Additional system memory was added to the PC, but performance did not improve.

Diagnosis: Depending on the circumstances, this may be true and may not indicate any sort of a system problem. Assuming that the memory is configured properly and is recognized by the system, it may just be that you have reached the point of "diminishing returns" if the PC already has a fair bit of memory in it. Most people who have 8 MB of RAM and add another 8 MB, will experience a much greater relative performance improvement than if they have 48 MB and add another 48 MB. How much performance is affected by memory total depends to a large extent on what operating system and software you are using, and also how tight your memory total was before you upgraded. Recognize that small changes in performance, less than 10%, are generally not noticeable to most users. This is discussed in more detail here.

Recommendation:

• Make sure that all of the memory is being recognized in the system. If it isn't, then the upgrade isn't working properly.
• Beware of being misled by benchmarks. Adding more memory will not improve the score of a benchmark that only tests the processor, for example. If possible, run an identical system benchmark before and after adding the new RAM. There should be some increase, even if it is minor.
• Check your BIOS settings related to memory to see if any of them need to be changed if you add memory or if you cross one of the common memory size "borders" (64 MB is the most common one).
• Often, the problem is that the operating system you are using is not being set up optimally to manage its memory efficiently. If you increase the amount of memory in some systems, the operating system may allocate the extra memory poorly, and you may see no improvement in performance

Explanation: Additional system memory was added to the PC, but performance actually decreased.

Diagnosis: There are three common causes of this problem. One is misconfiguring the new memory so that it isn't being used properly. Another is adding slower memory to an existing system than what is already in it, which can cause the memory subsystem to slow down slightly. Adding more memory than can be cached by the secondary cache can also lead to a significant performance hit.

Recommendation:

• Make sure that all of the memory is being recognized in the system. If it isn't, then the upgrade isn't working properly.
• Double-check the setup of the new memory to make sure that it is configured correctly. Check the BIOS settings, because some BIOSes need settings changed when the memory size changes.
• If you have added enough memory to take you over 64 MB of RAM, say to 80 MB or higher, your problem may be that your secondary cache can only cache the first 64 MB of memory. The Intel 430FX, 430VX and 430TX chipsets only cache up to 64 MB of RAM. Any memory above that is uncached, and accesses to those memory locations will be much slower than accesses to memory below the 64 MB barrier; see here for more. There is nothing you can do unless you want to upgrade your motherboard.
• If you have added memory to the system that is either slower in speed than what was there before (70 ns vs. 60 ns for example) or uses a slower memory technology (adding FPM memory to a system using EDO already) then your motherboard may slow down all memory accesses to match the speed of the slower memory, especially if it sets memory timing by autodetecting the memory speed. This will cause a small performance decrease. You may be able to change BIOS settings to get around this.