Thoughts On the Missing Memory Issue

[magnetar]

Yes, the infamous missing memory issue, as seen with Intel Socket 1366 CPUs. Curiously, most people don't see it that way, it's either the memory modules or the mother board that is the problem. Do we see this problem on other platforms? This issue remains a mystery, since if an absolute fix or identification of the problem were known, it would be posted here and everywhere.

Consider for a moment, how it is that we know we are missing memory? Usually if not always because something is telling us that. That something being either our mother board's BIOS, a hardware monitoring program, or our OS. It's not because our software is failing due to lack of memory, or something can't load due to a memory shortage. Sure, it is surprising and frustrating to see CPU-Z display 4GB when we have 6GB installed, while the BIOS displays something else, but how do we know we can trust these programs? What is the most trustworthy measure of the amount of memory installed?

I have seen on my socket 1366 CPU PC with 3 x 2GB memory modules installed (yes, G.SKILL), differing amounts of memory displayed in different monitoring programs at the same time. One shows 4GB, the other 6GB. I've seen a program display a total of 4GB of memory, and then have all three of the 2GB modules individually listed in a list that shows the details of the module. The BIOS will show 6GB, while a Windows Gadget will show 4GB right after I exited the BIOS and the PC finished booting.

I've noticed this happening when using memory in triple channel mode as I described above, and have not seen it when using 2 x 2GB modules. That has me wondering, but then that is just my experience.

My point here is that I don't feel it is fair to only blame the memory and mother board manufactures for this problem, when it could be an issue with the i7-9xx CPUs with their integrated memory controller. Windows may also be causing this as well. I wonder why everyone just blindly believes the monitoring programs, how do we know they are right? I recently compared two well known hardware monitoring programs, which showed significant differences in CPU and Core temperatures, and also relatively large differences in voltage readings. How can that be? Are they reading different things, or interpreting them differently? A third program was in agreement with one of the others, but can we trust that?

I'm surprised that given all the grief memory and mother board manufactures are experiencing with this issue, that they have not created a program to check the amount of memory installed, if it is installed correctly, if it is broken, and what value for the amount of memory is being written to the usual places. In another forum I was discussing my point, and one person replied that if one hardware monitor out of many reported missing memory, that is what they would believe. Given that, I wish G.SKILL and others, Good Luck!

[Tradesman]

Good thoughts and observations

[GSKILL TECH]

Brought up some good points, but let me add.

There is not absolute fix because there are multiple causes. If you read the sticky about missing memory, that explains many possible fixes. Now, the reason why there are so many fixes is because the issue may lie within any of the hardware. (Motherboard, memory, CPU, or user error) It is probably the most complex issue, especially when there are no settings needed for capacity, it should be automatically detected!

How do we trust programs? You don't. Use the BIOS as your base, what is stated in BIOS is what the system is operating at. Although, for voltages, you need to look in the "hardware monitor" section of the BIOS. That is the section that shows actual voltages. Where you set voltages, is not what the system will actually use. Depending on motherboard brand, the actual value supplied will vary by a certain amount. Most programs will not show actual voltage, so just use BIOS.

The LGA 1366 CPUs (triple channel) can be very picky with memory. QPI and DRAM Voltage plays a large role in stability. A slight difference can cause a problem, and that is an issue on the CPU side of things. But, it is not the most common reason which is why the CPU is not "blamed" as much.

Programs, programs, no need for programs. Just stick to the basic, "old school" way, BIOS. You can use the EasyTune, TurboV, etc. to configure settings, or just to verify results. Specifically the missing capacity issue, just check BIOS, and Windows. CPU-z will always show the detected amount, but only Windows or BIOS will tell you exactly how much useable memory is available.

Happy New Years to all
GSKILL TECH

[magnetar]

Thanks Guys

Thanks for your comments. I am just sick of people immediately declaring the memory or motherboard is bad without spending any time researching the issue and trying things to fix it. As you know, you still see people complaining they can't see their 8GB of memory on a 32 bit OS.

I certainly agree the BIOS is the first and best place to find all the values we have mentioned. My point is that many people do not use the BIOS for that, they believe the programs above all else, since that is what the "pro's" on the hardware review web sites use. Actually, you are the first moderator or official web site person I have seen that states these programs can be wrong!

You are also correct of course about the other settings related to memory, such as the QPI speed, IOH voltage, Vtt, etc. These things are different and a bit more complex than socket 775 CPUs and their chipsets.

Frankly and no offense, but there is not much information in the sticky. Not that you can find more anywhere else.

[GSKILL TECH]

Programs can surely be wrong, that's why updates, revisions, patches are always available.

As a perfect example, CPU-z and memtest always have glitches and such that need to be updated with newly released hardware. Detecting incorrect timings, frequency, model numbers, all are very possible and have happened before. So although most programs are fairly accurate, when there is a serious issue, the best thing to do is check multiple programs. Also, make sure you know what you are looking at.

In CPU-z, ET6, TurboV, all the clocking programs will show DRAM Frequency in "MHz". This is the real frequency, not effective frequency. If you use a benchmark test, then it will show effective frequency as "DDR3-xxxx, DDR2-xxx, DDR-xxx" Many people confuse this even though I try to post it every where.

Also for "Max Bandwidth", DDR3 motherboards will always show PC3-10700 (667MHz). DDR2 will always show PC2-6400 (400MHz). This shows the max default for the motherboard, not of the memory. People always get upset about this, and some even claim we are scamming people and all kinds of ridiculous things, but they just don't know what they are looking at.

The sticky includes a link, the link has all the possible causes for missing capacity.

Thanks for adding to the forums, information is only good if it is shared. Luckily we have forums and technology to store it so everyone can see it. Hopefully people will do their part and just search for answers.

GSKILL TECH

[magnetar]

Thanks for chatting with me G-Tech, I'm sure you are busy.

I have realized that in the PC world, the more that I learn actually shows me how much I really do not know. The details of memory speeds and types is not simple and most people, including myself, don't spend enough time studying it.

Recently in a different PC forum, a CPU-Z user could not understand why, when he changed the speed of his memory in his BIOS, that he did not see any change in CPU-Z, under the SPD Data tab. He was frantic and thought something was broken. I explained to him that SPD data is static and populated by the memory manufacture with the recommend speed and timings settings, and hopefully contained EPP or XMP data too.

You did bring up one of my pet peeves, that is the reality of memory speeds. NOT that G.SKILL or any other memory manufacture created the standards and terminology, you and they just work and live withing them. Such as, DDR2 "800 MHz" memory that actually runs at 400 MHz, but due to the "double data rate" of the memory is equivalent to 800MHz at a single data rate. I've wondered how much memory has been RMA'd simply because the user did not truly understand how it actually works.

Ok, finally something that may be useful. Recently I purchased some G.SKILL F3-12800CL9D-4GBRL (2x2GB) modules at a great price. I thought I would try them in my ASUS SaberTooth motherboard with an i7-930 CPU. The point was to try running in dual channel mode rather than triple channel mode, which I could have done with my other memory, but I wanted to try some modules that were not on the QVL list (the 2GBRL modules are, but not the 4GBRL.) I wanted to see if the missing memory issue would occur in two channel mode.

Once installed, it worked just fine, ran immediately with the XMP setting and was all recognized by the BIOS, etc (Windows 7 64 bit.) I ran some memory tests on it, including one that is part of the Intel Extreme Tuning Utility, although I know nothing about it specifically. I used that memory in my PC for two weeks, the PC used everyday, and usually put in sleep mode at night and shutdown a couple times too.

I never saw any indications of missing memory in dual channel mode. In three channel mode, the discrepancy with the amount of memory readings would appear in a day or two, on and off.

I don't know if others see missing memory in dual channel mode as I have not scanned the forum enough, but I did not in two weeks of testing. Dual channel mode is easier on the IMC, etc, than triple channel, so that is a good indication of what the issue really is all about. Of course my experience is only one, but we can at least add it to our data points.