Based on experience, I can provide some generalizations (though this isn't an area of expertise)...
I wouldn't use the word 'burdensome', but...
- In general, performance is going to be best with a single DIMM per channel. If two DIMMs are used, they should be well (if not exactly) matched.
- The physical layout of the DIMM (i.e. where the ICs are located on the DIMM) isn't the issue; it is how the ICs are organized and presented to the memory bus. For Intel Desktop Boards, you want x8 organization; x16 organization is poorly supported (if at all) and is definitely a problem if you are attempting to use two DIMMs on a channel.
As usual, you are a fountain of information. Interesting. I ignorantly thought that x8 and x16 were a physical description of the chips on the stick. I learn something new every day.
I looked through my pile of sticks pulled from old PCs. Of two different Hynix DDR3-1333 sticks, one is 1Rx8 and the other is 2Rx8, with the former having chips on one side and the latter having chips on both sides. Two Samsung DDR2-800 sticks are 1Rx8 and two Qimonda DDR2-800 sticks are 2Rx8 sticks, with the former having chips on one side and the latter having chips on both sides. None of my DDR sticks, Crucial or Kingston, have a designation on them, with all of them having chips on both sides. My currently-being-used Crucial DDR3-1333 Ballistix Tactical sticks have nothing that is obvious. I looked at Speccy's data for RAM, but it does not list whether a stick is x8 or something else.
And then I noticed something I should have seen before. The Crucial sticks are marked 7-7-7-24, but Speccy lists them as 9-9-9-24. All of my systems, except for laptops, use 6-series Intel boards and Sandy Bridge CPUs (no W-10 for me!). I thought Intel boards (and those of other major manufacturers) read the XMP profile and use it. I went into BIOS for a DH67BL system and saw that it thinks the sticks are 9-9-9-24, Well, that sux. Do Sandy Bridge and 6-series boards not recognize CL=7? The DH67BL product manual does not say anything about it. I'll have to dig into the DZ68BC manuals now.
Intel's boards do indeed read and use the profiles in the SPDs. If configured for Auto, they will attempt to use the XMP profile(s) provided. If they work, great, but if they don't, the BIOS will first back off to the SPD's JEDEC defaults and then, if all else fails, to MRC (Memory Reference Code) defaults. In my experience, SPDs sometimes contain incorrect information (a quote from one of the BIOS engineers: "they <bleep>ing lie!").
It's time to take a look at the actual contents of your DIMMs SPDs. There are a lot of (mostly) free tools out there that can display this information for you - AIDA64 (free trial - but worth purchasing IMHO!), RW-Everything, CPU-Z, HWMonitor, etc. and etc. Pick one and take a look...
BTW, I also use Crucial Ballistix memory - the ones that support TSOD (Thermal Sensor On DIMM) - in a number of my systems, so that I can monitor the temperatures of the DIMMs. I am not sure which Ballistix line they were from (Tactical Tracer?), but it is possible that your memory also has this capability. In the final releases of Intel Desktop Utilities (IDU), I added support for exposing DIMMs with TSOD. If you have IDU installed and your DIMMs also have the sensors, you should see these sensors show up in the IDU display and alert monitoring. In my DZ87KLT-75K system, here is the Sensor Threshold report generated by IDU:
I chose CPU-Z because it sounds like a cheap zombie movie. It confirmed Speccy and BIOS, that the sticks are 9-9-9-24. I will ask Crucial customer support about this. I'll post their answer.
I chatted with Crucial, but the agent had no answer as to why all the sticks read 7-7-7-24, but Speccy, BIOS, and CPU-Z all read 9-9-9-24. I asked for a utility that Crucial trusts, but the agent kept side-stepping the question. He asked me if I enabled XMP, but I thought that was enabled by default by the AUTOMATIC setting in the memory page of the performance tab (I OCed a tiny bit, but I never messed with memory or voltages; and my DH67BL uses default BIOS settings). Is there a way to verify that XMP is in play?
I only see about half of that screen, so my sticks must be Walmart quality. By the way, how did you get the dialog to stretch to its full length? One of the things I never liked about IDU was how it did not allow resize. I just installed it on a DH67BL for the first time and it has that same old problem. The version is 3.2.8.
This is the dialog for setting sensor thresholds (go to Options then Set Sensor Threshold); it also uses a fixed-size window, but sets this size based upon what it needs to display.
Long story; typical result of outsourcing. I was not involved in its design or implementation. I always hated the fixed-window-size limitation but, when I eventually took over its ownership, I discovered I didn't have the resources to do anything about it (the implementation was so poor that a complete rewrite would have been necessary). My changes were limited mostly to the underlying H/W access layers, where I could add support for additional sensors (individual processor core temperatures, HDD/SSD temperatures (via S.M.A.R.T.), DIMM temperatures (via TSOD) and the extra voltage sensors included on the DZ87KLT-75K) without requiring any major changes in the GUI implementation. I did get scrolling added to the Hardware Monitoring scenes so that the additional sensors could at least be seen, but the GUI still sucks for the most part.
Easiest way to tell if your overclocking is working (remember that, technically speaking, implementing XMP profiles is a type of overclocking) is to look at the current memory speed from within the same set of programs used to display the SPD data - or just look at the System Information scenes in IDU.
Okay, I understand why 7-7-7-24 sticks show up as 9-9-9-24 ones. There are two different explanations.
DH67BL: In Performance -> Memory, I can only choose between Automatic or Manual for memory settings, but the Manual option toggles between DDR3-1333 and DDR3-1066. Remember this board only allows some minor tweaking of the graphics multiplier. This means that instead of actually reading the memory profile and using what it can offer, the board defaults to a hard-coded number of 9-9-9-24 for DDR3-1333 and the default for DDR3-1066.
DZ68BC: In Performance -> Memory, I have a few choices, Automatic, and two Manuals (sorry, I did not take notes), with one of them offering 7-7-7-24. However, BIOS will not remember the selection. If I set it to 7-7-7-24 and escape out of that screen, the next time I come back to it, the entry is reset to 9-9-9-24. If I set it and leave BIOS, saving the changes, the next time I return or run CPU-Z / Speccy the entry is back to default.
Yeah, I was on the same Set Sensor Threshold screen. My RAM must not have the cone-of-silence for temperature, so it has less data to present.
And I appreciate the GUI problems. It is a rare developer or manager who understands that a poor GUI makes even brilliant underlying software look bad.
I realize no one except Scott is interested in this, but someone might be reading this later and want the whole story.
I revisited my DH67BL and DZ68BC systems, all of which have the same Crucial DDR3-1333 7-7-7-24 Ballistix Tactical RAM, but the former has only two sticks and the latter has four (the two in the blue sockets are of one batch and the two in the black sockets are of a different batch). I tried to be more methodical in my approach. As readers might remember, both systems showed 9-9-9-24 via CPU-Z.
I first went into DH67BL BIOS and noticed that there were only two options for memory OC: Automatic and Manual. I selected the latter and changed 9-9-9-24 to 7-7-7-24. I saved the changes and booted. CPU-Z showed the correct timings, 7-7-7-24. Zounds!
Then I turned to DZ68BC BIOS and noticed that there were three options for memory OC: Automatic, Manual, and XMP-1333. I tried all three with no success. XMP-1333 changed 9-9-9-24 to 7-7-7-24, but when I saved and booted, CPU-Z showed 9-9-9-24. When I got back into BIOS, the setting was back to 9-9-9-24. Manual allowed me to change all values, but as with XMP-1333, the changes are not retained.
I tried one last trick, that of removing the pair in the black sockets, but no joy. DZ68BC does not retain XMP settings, but DH67BL does.
Yet another update. After getting the DH67BL PC to see 7-7-7-24, I ran WEI and got 7.7 for the RAM, one tenth higher than before. Both CPU-Z and Speccy listed RAM as 7-7-7-24. Then I returned to the DZ68BC PC and tried one last thing. As I said before, there are three options: Automatic, Manual, and XMP (1333 / 1066). Automatic always saw the RAM as 9-9-9-24. Manual allowed me to change each value to 7, but as I arrowed-down to access the next value, the first value would revert to 9. XMP (at 1333) would set all three to 7s. On a whim I used XMP, pressed F10 instead of Esc, and saved the value right then and there. I booted and ran CPU-Z and Speccy, both of which displayed 7-7-7-24. Zounds! So there is a way to use faster memory, but only one way. However, WEI only shows the RAM as 7.6 as before, so I'm still not totally satisfied, but I have totally run out of things to try.