RAM Testing

RAM (Memory) Testing

SVANDUSEN 19:13, 26 May 2010 (UTC)

RAM is a deceptively simple device to test. It appears to be simply a matter of gathering, sorting, loading, removing, labeling, and storing; repeated over and over again. In actuality, there are a multitude of details and issues involved that would take far more space to cover than is allotted here. This is a very basic overview of the essentials that are necessary to get through the process. For more information please consult other experienced volunteers, instructors, or research such sources as Wikipedia or forums on Google. STEPS NOTES AND DETAILED INSTRUCTIONS Collect Collect incoming RAM from the collection points. Advanced Testing Incoming Rack - the yellow tub on the shelf labelled 'Incoming RAM' (use this to collect from the other points as well). System Evaluation - the four boxes on the work bench labelled 'Memory' . Receiving - the box on the top shelf labelled 'Unsorted Memory'. Sort Sort the collected RAM into several categories as follows: 1. Sort by Type SIMMS - Automatic Recycle. RAMBUS - Automatic Recycle. CRIMMS (spacer sticks for RAMBUS boards) - Automatic Recycle. SODIMMS (Laptop) - Take to "Laptop Incoming RAM" bin in Laptop Build. SDRAM - Automatic Recycle. For now, anyway. DDR - Set aside to sort by speed and capacity. DDR2 - Set aside to sort by size speed and capacity. DDR3 - Set aside to sort by size, speed, capacity-When tested put aside for highend builds/store. Other - There may be Printer RAM or other types of RAM or small cards that may or may not have a use. Ask instructor. 2. Sort for ECC Separate out all ECC RAM from the DDR, DDR2, & DDR3 that has been retained for further processing - take to "ECC Incoming RAM" bin in Server Build. Most ECC sticks will say ECC on them. Others will have a letter at the end of their speed with a letter. If it is a U, it is a regular RAM stick. If it is an R, E, or F, it is some type of ECC RAM. If not, look at the number of embedded chips in them. If it is four or eight with no special chips in the center, it is a normal stick. If it has nine chips, or there are several smaller chips in the center, the RAM stick is SOME type of ECC. 3. Sort by Speed SDRAM - PC 66 - Automatic Recycle. PC 100 - Automatic Recycle. PC 133 - Automatic Recycle. DDR - PC 1600 (200 MHz) - Automatic Recycle. PC 2100 (266 MHz) - Set aside to sort by capacity. PC 2700 (333 MHz) - Set aside to sort by capacity. PC 3200 (400 MHz) - Set aside to sort by capacity. DDR2 - PC2 3200 (400 MHz) - Set aside to sort by capacity. PC2 4200 (533 MHz) - Set aside to sort by capacity. PC2 5300 (667 MHz) - Set aside to sort by capacity. PC2 6400 (800 MHz) - Set aside to sort by capacity. PC2 8500 (1066 MHz) - Set aside to sort by capacity. PC2 9600 (1200 MHz) - Set aside to sort by capacity DDR3 - All - Set aside to sort by capacity 4. Sort by Capacity SDRAM - 32 thru 128 MB - Automatic Recycle. 256, 512 MB, & 1 GB - Automatic Recycle. DDR - 128, 256 MB - Automatic Recycle. 512 MB, & 1 or 2 GB (only 1 GB+ for PC 2100/266 MHZ) - Stack by capacity in designated area on table (wooden rack labeled untested) to be loaded into test board. DDR2 - 256 MB, or under - Automatic Recycle. 512 MB, 1 or 2 GB - Stack by capacity in designated area on table (wooden rack labeled untested) to be loaded into test board. 5. Unspecified speed and/or capacity Frequently RAM will not be labelled as to speed or capacity by the manufacturer. In such cases it is necessary to determine these characteristics by other means. DDR, DDR2 & DDR3 - Use a test board. Several boards give an accurate reading regarding capacity and speed (see also "Loading & Testing). Load a single stick. Power on at the power supply. Wait until the system enters "MemTest". Read the capacity and speed from the display. Power off at the power supply. Remove the stick. Label and stack in designated area on table (untested box) to be loaded into test board. Research online. Several websites dealing in RAM specifications have been bookmarked and other websites can be accessed by "Googling" using the manufacturer and or model/part number (if available). Locate a website. Locate the RAM in question in the website. Note the capacity and speed in the specifications. Label and stack in untest box on table to be loaded into test board. Ask an instructor. Some of them know ways to determine information by examining the embedded chips. 6. "Kit of 2" Some manufacturers, especially Kingston and Super Talent have a marketing ploy of packaging RAM in pairs totalling a specific capacity, usually 512 MB and 1 or 2 GB. In these cases, each stick has only one half of the total capacity; eg. a 1GB "Kit of 2" consists of two 512 MB sticks. This will be indicated somewhere on the label, either as a distinct statement as such or, in the case of Kingston, within the model number; eg. KVR400x64C3AK2/1G. In this case the "400" indicates the speed, the "1G" indicates the total capacity of the kit, and the "K2" indicates that the stick is part of the kit. The actual capacity of the stick in this example is one half of 1 GB or 512 MB. Both halves of such kits are not always present when they come into our possession but they are testable separately. When labelling such sticks after testing, treat each one as a separate stick and label it with its actual individual capacity (see "Labelling and Routing"). Test Test the sorted RAM in the wall-mounted test boards as follows: 1. Select -Select the RAM to be loaded according to the following criteria: Must all be of the same type. Must all be of the same speed. May be of mixed size; ie. 256 and 512, etc. Should be balanced by size across channels (all test boards are "Dual Channel"); ie. 512 MB in channel 1 and 256 MB in channel 2 (see "Load"). Should be all of the same brand or paired by brand, if possible, for compatibility and to resolve "latency" issues (particularly with DDR2). Balance and brand can be worked around if single sticks are all that are available (see "Boot Up"). 2. Load - Load the RAM into the test boards as follows: Be aware --- Banks (the slots into which the sticks are inserted) are designated 0, 1, 2, & 3 moving from inboard [closest to the CPU] to outboard [farthest from the CPU]). Channels consist of two banks (all test boards are dual channel but will operate in single channel mode). Banks may be color coded; ie. two black and two blue. Clips on the banks (hinged plastic parts at the ends of the banks that hold the sticks in place) may be color coded; ie. two black and two white. Some boards may have a separation between pairs of banks, some do not. Some boards may alternate color coded banks or clips; ie. blue-black-blue-black, or pair them; ie. blue-blue-black-black. In any case, Channels are always composed of pairs of banks. Either the banks themselves or the clips will be of matching color. Intel cpu boards will nearly always match banks 0 and 2 as channel 1 or A, and banks 1 and 3 as channel 2 or B. AMD cpu boards will nearly always match banks 0 and 1 as channel 1 or A, and banks 2 and 3 as channel 2 or B. Load from inboard to outboard. Load largest capacity inboard to smallest outboard; eg. 1 GB in banks 0 and 2 and 512 MB in banks 1 and 3 or 512 MB in bank 0, 256 MB in bank 1, and 128 MB in bank 2. Insert the stick into the bank. Align the notches on the stick with the "key" (the small bar) in the groove of the bank. Make sure that both ends of the stick are in the groove of the bank. Press firmly and evenly until the stick moves completely into place along its entire length and the clips click into place in the notches at the end of the stick securing it in place. Manually check all clips to confirm that all are closed properly and the stick is securely in place. Power on using the switch on the power supply and, if necessary, the power button switch attached to the test board. Set the KVM channel to the corresponding number or letter of the test board (orange label w/ arrow on power supply). 3. Observe boot-up process The system should go through a boot up process appropriate to that particular board and its corresponding BIOS version and arrive at the Dynamic Host Configuration Protocol (DHCP followed by a spinning slash mark), indicating that it is booting from the network. Changes in total RAM capacity, lack of balance across channels, or issues peculiar to a particular board (eg. "System Fan Not Functioning") may result in the appearance of a prompt message directing you to hit a function key (F1, F2, or F4), ESC, or "any key". in order to continue. Do so. The system should then continue on to DHCP. Some boards, for whatever reason, may require you to pass through the BIOS. Simply hit the key indicated by the prompt message, hit ESC, and then hit Enter. The system should then continue on to DHCP. Following the completion of DHCP the system should arrive at a program called MemTest without any additional manipulation. 4. Observe MemTest This will be a blue screen with white lettering (occasionally monochrome) that will provide information about the RAM being tested (and the CPU as well). A box (outlined in white, dotted lines) in the upper right hand corner will indicate "Pass %", "Test %", "Test #", "Testing", and "Pattern" ("Testing" and "Pattern" are not relevant for now). This will indicated what point in the progress of each pass and test has been achieved. A line labeled "Settings", beginning on the far left of the screen, will provide information about RAM speed and "latency" (latency is not relevant for now). RAM speed is not present in all versions of MemTest. A row of column headings will appear about halfway down the screen and extend all the way across. Only "WallTime", "Cached", and "Pass" are relevant. WallTime indicates how much time has elapsed since MemTest began. Cached indicates how much total RAM has been loaded onto the board. Pass indicates what pass the testing process is on. 5. Monitor MemTest Indications of Failure - No video, sometimes accompanied by a series of beeps or video is distorted - The system will not boot up or show any other video signal or- The video is streaked, broken up into blocks, completely fragmented into dancing pixels, or wildly discolored. One or more sticks may be bad and/or incompatible; ie. ECC. See "Remedies" "Red Screen" - The most common indicator of failure. The bottom half of the screen turns red, a new set of column headings appears, the section fills with data and may begin scrolling rapidly. See Remedies. Usually occurs within the first two tests of the first pass but may, occasionally, happen much deeper in the testing process. One or more sticks have failed. See "Remedies". "WallTime" is frozen - The clock stops ticking ... period. This may happen very early in the testing process or may occur much deeper. One or more sticks have failed. See "Remedies". "Cached" column indicates a different total capacity than the known amount physically loaded - The total capacity of RAM indicated by this column must equal the known total capacity of the RAM physically loaded onto the board. If not ... this indicates that the system is not reading all of the RAM that has been loaded. It is permissible for the speed indicated by MemTest to differ from the speed indicated on the stick (it will simply run at a slower speed ... this is not optimal but acceptable). This may NOT be a problem with the RAM. This may be a problem with one or more of the banks on the test board. One or more sticks and/or banks have failed. See "Remedies." Remedies - This can be tedious, time consuming, and tricky, but with the technology available it is the only way. Red Screen - here there is an important clue. A new row of column headings will appear above the red area (the bottom half of the MemTest screen). Only Test, Pass, and Failing Address are relevant here. Test and Pass will indicate at what point in the testing process the failure occurred. If it occurs early (eg. pass 0, test 1), all well and good. If it occurs deeper in the testing process, it may be that a failing stick will begin testing well and then fail, making it more difficult to identify. Failing Address is in two parts which indicate the pattern (not relevant here) and the point in the total capacity at which the failure occurred. Moving from outboard to inboard, subtract the capacity of each stick from the total until you are in the range at which the failure occurred, eg. four sticks of 256 MB RAM for a total of 1 GB have been loaded. A failure occurs at 368.2 MB on pass 0, test 1. The failure has occurred between 512 MB, and 256 MB. Subtracting 256 MB for each of the two outboard sticks (banks 3 & 2) leaves 512 MB. The failure occurred at a point below this capacity. The failure occurred at a point greater than 256 MB, the capacity of the first inboard stick (bank 0). Therefore, it is most likely that the stick in bank 1 is the stick that failed. Power off at the switch on the power supply. Remove the suspected stick. Restart the testing process and monitor MemTest again for failures. No Video or Cached column indicates lower total capacity than physically loaded. This may simply be a matter of the sticks not being seated properly, or the system does not recognize the RAM in the order in which it was loaded, or it has compatibility issues with brands, latency, etc., OR it has some other problem all its own and refuses to give you any indication as to what it might be. Power off at the switch on the power supply. Remove and re-seat all of the sticks. Change the order, re-balance the load across the channels, reverse the channel position by pairs, etc. Restart the test process and monitor MemTest again for failures. WallTime freezes or the location steps for Red Screen, No Video, or Cached column differential don't resolve the issue. Unfortunately ... it's strictly trial and error from here on. Power off at the switch on the power supply. Remove the suspected stick or sticks, singly or in pairs. Restart the testing process and monitor MemTest again for failures. Repeat as necessary until issues are resolved. If a bad bank or other system failure is suspected for any reason - Power off at the switch on the power supply. Remove all RAM sticks. Test each bank with a known good RAM stick. If all banks pass, all well and good. If not, power down, remove, stick or sticks, mark the suspected bank or banks with a label, and notify an instructor. If something other than a bad bank or banks is suspected, ie. video or network issues, ask an instructor. 6. For the duration of the testing process - Check on the progress of the testing process. The time necessary to complete the testing process can vary greatly, depending on the capacity and speed of the RAM that has been loaded, the FSB (Front Side Bus) speed of the CPU or other factors specific to the test board. The time can be less than one hour, with low capacity and/or high speed, or can take an incredibly long time with high capacity and/or low speed (eg. 2 GB of SDRAM at 133 MHz can take nearly 24 hours). Generally an average load of around 1 GB of DDR or DDR2 will take between 2.5 and 3.5 hours. A complete testing process consists of three passes, labelled 0-2. Each pass consists of 'eight tests, labelled 1-8. MemTest will continue to run indefinitely until is manually discontinued. If the pass number is 2 or greater, the testing process is complete. Watch for indications of failure, as indicated above, and correct as necessary. 7. Repeat these steps for each test board as long as there is RAM to be tested. 8. Do NOT load any boards after 5:00PM. Label & Route After the testing process has been completed, label and route the tested RAM as follows: 1. Unload the test boards - . Power off at the switch on the power supply. Remove all RAM sticks and take them to the table to be labelled. If possible, remove any old, handwritten labelling to avoid confusion. 2. Label as follows: DDR, DDR2 & DDR3 - type and capacity and speed. Eg. DDR2 512/667. DDR and DDR2 have more subtle visual cues (other than the label) and can be easily confused by after-testing users such as build. Therefore, it is highly advisable to make them as distinguishable as possible. Kit of 2 - Label each stick with its true capacity and speed. Eg. a 1GB kit of 2 - each stick would be labeled as 512 MB with the speed. If you have both halves of such a kit, make a small nonstick paper strip label, write "Kit of 2" and "xxx MB (or GB) each" on it, wrap it around the kit, and rubber band it together. 3. Route as follows - The box on the table labeled, "Tested DDR2" Place the vast majority of the labelled RAM in the appropriate slots in these two boxes (see Store for exceptions). This will be harvested regularly by build and occasionally by the store to fill their needs. Lock box in build room now holds RAM overstock The boxes on the table will populate quite rapidly. As they fill (usually once a week or more), empty the slots into the appropriate static free bags in the lock box in the build room. As the bags fill, tape them shut and place them in the RAM Overstock in the lock box in the build room. Replace the bag with an empty one labelled with type, capacity, speed (in MHz) and PC or PC2 code (PC 2100, PC 2700 ... PC2 4200, PC2 5300 ... etc.) These will be harvested by build and the store to meet their needs. Store - Some tested RAM should be taken directly to the store (do not put in the outgoing "Store" box on the rack shelf next to the TARDIS door). SDRAM - 1GB/133 DDR - 1GB/any speed. 512/400 - pairs (or more ) of quality brands (Hynix, Nanya, Samsung, etc.). Leave singles in the box for build. DDR2 - 1 or 2 GB/any speed. 256 or 512/667 or higher. All PAIRED & BUNDLED Kit of 2 sets. Any sticks, paired or not, any capacity and speed, with "heat spreader" technology ... metal plates over the embedded chips that run the length of the stick on one or both sides. End of DAY POWER OFF ALL TEST BOARDS AT THE SWITCH ON THE POWER SUPPLY WHETHER THE TESTING PROCESS IS COMPLETE OR NOT!