| Installed Memory vs RAM Usable by Windows | |
| Articles - Featured Guides | |||||
| Written by Olin Coles | |||||
| Monday, 14 June 2010 | |||||
Page 3 of 3
Troubleshooting System MemoryAlthough our test system uses an Intel X58-Express triple-channel motherboard with three 2GB DDR3 modules installed for a total of 6GB RAM, the same problem could exist on any other motherboard with more or less capacity. So when the BIOS on our test system began to report less RAM than was installed, the first step was to verify the amount in Windows. Since this 64-bit Windows-7 computer functioned perfectly stable, even with the memory and processor overclocked, it was hard to believe that a memory module had gone bad. In fact, this is where we learned our first clue. The easiest way to troubleshoot system memory is by replacing it with known-good components. When this isn't possible, the existing modules can be individually tested using free tools such as Microsoft Memory Diagnostic or Memtest86+. After the memory has proven itself operational, the blame shifts to the motherboard, with a remote chance of the processor causing problems. Diagnosing motherboard and processor defects is possible by using software to focus stress tests at a particular component, such as Lavalys EVEREST does. Unfortunately, there's only one guaranteed solution: replace the hardware. Using three difference triple-channel memory kits to troubleshoot, 3GB (3x1GB), 6GB (3x2GB), and 12GB (3x4GB), each time the system started the BIOS would report an available amount of memory that suggested one of the modules was undetected. Inspecting the memory settings inside the BIOS suggested otherwise, giving us the ability to set latency timings for each of the three modules. Here was another clue: the third module could only modify the CAS latency, while the other settings were unavailable. Visiting the Windows Resource Monitor delivered another valuable clue: 6144 MB of memory installed, but 2050MB was Hardware Reserved.
Windows 7 Resource Monitor - Memory TabDespite all of the obvious symptoms and indicators, our diagnosing tests and troubleshooting steps had removed system memory from the list of potential suspects. The next step was to replace the computer's motherboard, and thanks to our many Motherboard Reviews over the past few years there were plenty of replacements on hand. After taking the time to update firmware on each an every motherboard we tested, each installation ended with the same result: the BIOS reports less memory than installed. By this point, even the most die-hard hardware enthusiasts would have run out of options, but we still have one suspect remaining: the processor. As the introduction eluded, processors now include an integrated memory controller (IMC) built into the CPU, such as the triple-channel DDR3 controller equipped on Intel Core i7 Nehalem LGA1366 processor series. I don't imagine that consumers and technicians would ever get to the point of replacing memory, motherboard, and processor, but anything's possible. In this particular case, I happened to own two Intel Core i7-920 processors for use in our Best CPU Cooler Performance series. Removing the first processor and replacing it with a second identical CPU was all the proof needed: the BIOS reported 6144MB of memory available, and the Windows System properties displayed Installed memory (RAM): 6.00 GB (6.00 GB usable). Success! Not so fast... this victory would cost the average person hundreds of dollars in unnecessary expenses if they took to the stores for replacement components. Even a seasoned computer technician would not find the problem right away. So what could speed up discovery for other people trying to diagnose similar problems? Look for the signals. In this case, Windows-7 displayed two forms of proof that all of the memory installed was actually detected. If you find yourself with a system where all of the memory modules appear visible to Windows, but the full amount of memory is not available, begin checking individual memory channel settings (for latency timing) in the BIOS. The processor's integrated memory controller is a great technological advancement, and the dawn of a new era in troubleshooting RAM problems.
Related Articles:
|
|||||





Comments
Often it helps just to reassemble the CPU.
Or some Pins in the Socket are bend, could be fixed with a needle.
Seems that it can happen that the CPU has no proper seat or is twisted a tine bit.
Or by tightening the Cooler to much.
Quite often it only happen with OC and @ stock everything works.
So please make sure, if the CPU is really the case or bad contact in some way.
I am active @ PCGH-Forum and we have/had some issues like that,
most was solved without changing anything at least.
I hope you can get some more light on to it.
Best wishes
SteinSchocK
##evga.com/forums/
Sometimes in an X58 based motherboard the CPU must be lined up absolutely perfect or you will see the problem that you troubleshot. Never rush a CPU install on this type of motherboard. Double and triple check the CPU pin alignment BEFORE you close the CPU locking mechanism
best regards
rjohnson11
EVGA Forum Moderator
#lmgtfy.com/?q=x58+triple+channel+n ot+working
Simply bump up your QPI voltage and it'll be fine.
Worst comes to worst, play with the ram divider to try a few different frequencies (slower is not always more likely to work).
(i.e. Mine works in tripple channel at 1600, but won't at lower speeds.)
This issue is super common with people running 6 dimms (you can almost count on it).
Slightly less common with 3 dimms.
The real clue is that when you post, it will say 'dual channel'.
A tool like CPU-Z (and windows) will identify all 6 slots worth of dimms.
The bios will also list all 6 if you go under the section that shows timing data.
The dimms are working, and are in communication with the mainboard.
You are not in dual channel because a dimm is offline.
A dimm is offline because you are in dual channel.
Essentially, the IMC is overloaded and can't handle 3 channels.
continued ...
You can increase the robustness of the IMC by raising its voltage, or you can decrease the load on the IMC by finding a friendlier memory divider.
Try 1.35v, as that is the top end of the normal range for an i7 (intel spec).
OCers often go over this limit by a good amount.
OCing usually needs more QPI voltage.
Filling 6 slots usually needs more QPI voltage.
And, having a 'lemon' memory controller usually needs more QPI voltage.
When I initially encountered this, I went to the MSI forums.
When I got there, I found a wall of threads on the same issue.
Looking around, it was everywhere, on every i7 mainboard forum.
Hence, why just googling will turn up hits on missing ram or channels.
-scheherazade
My condolences.
( took about 4 full install/test cycles to get my first Athlon to connect reliably to RAM,
and the erors were showing-up only in the near-final controller-tests iirc? ).
Secondly, if you test a module, pull it & re-place it, you DON'T know it's reliable now.
( see the first item )
Thirdly, Memtest86/Memtest86++ should have identified that some memory wasn't being usable, directly, thereby showing the problem.
When testing, SIMPLIFY the system, so as few variables as possible are interacting, otherwise you're just doing voodoo...
Cheers,
That said, the easiest solution is simply to run the memory at 1600MHz. I know it's frustrating not getting every last iota of performance out of the memory you paid for, but really, you'll never see a whit of difference between 1600MHz and 1866MHz except in synthetic memory bandwidth tests: not a single extra frame per second in any game or anything...