[SOLVED] solving RAM incompatibility on this old board

spacemanspiff2000

Distinguished
Sep 8, 2009
32
0
18,530
I was inspired to post here because of THIS POST which is essentially the same thing but with different hardware and a different outcome (the advice therein didn't help me).

It's an old setup, but it's a workhorse HTPC I would like to improve b/c I have the h/w in-house to do it. Everything about this tells me it should work, but it just doesn't.

mobo: Intel BOXDH67GDB3
CPU: Intel Core i3-2100
PS: Antec EarthWatts EA-380D Green
existing RAM: G.Skill 4GB kit: F3-10600CL9D-4GBNS DDR3-1333 (PC3 10600) (2x2GB)
adding RAM: Crucial 4GB kit: CT2KIT25664BA1339 DDR3-1333 (PC3 10600) (2x2GB) --- NOTE: This kit came out of a long-working system it shared with other RAM.

How it acts:
  1. it still posts/boots with only the existing G.Skill RAM
  2. it posts with only the added Crucial RAM
  3. it does not post at all with both in at the same time
  4. it does not post with RAM in the secondary dual channel slots
(per DH67GD manual: "The Intel Core i7, Intel Core i5, and Intel Core i3 processors require memory to be populated in the DIMM 1 (Channel A, DIMM 1) socket." So I can't tell if the 2nd channel is bad, but I would think that unlikely (since I've never used it).

I have since upgrade the BIOS on this mobo all the way and there's no change in how it acts.

I tried to follow the guidance on the inspiring post, but my research says these are both 1.5V 9-9-9-24 modules. In looking at the actual settings in the BIOS area when each is in the HTPC alone, I only found 1 item at all different: tRFC was automatically 107 for G.Skill and 74 for Crucial, so tried forcing it to stay at the lower(?) setting of 107 but there's still no change from #3 above.

One last thing, I'm not an overclocker at all ... I'm not against it, I've just never done it and don't know what I'm doing there ... so if anyone has ideas I can try that play more with the BIOS settings, I'm just going to need specifics like I'm a 2-year-old. "Just play with the settings" won't work for my novice overclocking status.

Bottom Line: I have h/w that I think SHOULD work together, and I would really LIKE it to work (because it "should"), but in the end I can press on with the 4GB I had working anyway. So I would be most grateful if there's something even more advanced/detailed than the text from my post-of-inspiration that I can try and not throw in the towel on this.

Thanks!
Spiff |^D
 
Solution
Memory is guaranteed in the form sold. Other combinations you decide to make have no guarantee to be compatible together.

You can try setting memory to a slower speed 1066MHz, relaxing timings, increasing DRAM voltage.

You can also test for defective memory at
MemTest86
Here's the deal: getting un-matched ram to work together is problematic at best and impossible at worst. EVERYTHING about the two kits has to match; voltage, latency, timings, etc. Sometimes even the exact same kits(brand, make, model etc), won't play well together, especially if purchased at different times or vendors. They sell ram in matched sets for a reason.
 

spacemanspiff2000

Distinguished
Sep 8, 2009
32
0
18,530
Here's the deal: getting un-matched ram to work together is problematic at best and impossible at worst.

I hear what you're saying. The writing may be on the wall. Thanks for the reality check.

But also, a correction: at best, it just works.

My "offending" 1333 RAM came out of a system where it co-existed just fine for years of regular use with 1600 RAM of a completely different model. Now, while I'm not a pro builder and while this is my personal first time hitting the potential "impossible" side of this, I've mixed RAM plenty in my time without a problem.

Thus, my hopefulness that there may be a solution for me here.

Also, the most obvious reason to the uninitiated to sell RAM in matched sets is because it simply makes sense (Who would ever sell 1 RAM module for dual-channel systems and make you buy two? Some unscrupulous sellers, maybe.) But that doesn't preclude other reasons beyond that, of course.
 

spacemanspiff2000

Distinguished
Sep 8, 2009
32
0
18,530
Have you tried relaxing the timing to 10-10-10-24?

You can try setting memory to a slower speed 1066MHz, relaxing timings, increasing DRAM voltage.

Thanks for the tangible advice, guys. Now I understand "relaxing" to mean "increasing" in this context and I tried all of these things, but no dice.

I also found this odd but critical tidbit from the mobo manual which made me immediately stop trying:
"The board has four DIMM sockets and supports the following memory features:
...
- Unbuffered, single-sided or double-sided DIMMs with the following restriction:
--- Double-sided DIMMs with x16 organization are not supported."

And the Crucial RAM appears to have exactly that configuration: 8 chips on each side.

So that's probably the final answer, there.
 

spacemanspiff2000

Distinguished
Sep 8, 2009
32
0
18,530
Memory is guaranteed in the form sold. Other combinations you decide to make have no guarantee to be compatible together.

Guaranteed? Yes. Sure. Expected? Well, I simply disagree with the notion folks put out there that one shouldn't reasonably expect seemingly compatible RAM to work together. And I don't want to leave future readers with the impression that that is the only truth.

I'm reminded of a Star Trek episode where Scotty shows up in the Next Gen time (or similar) and schools the contemporary engineers on starship specs, because he wrote them to be safe for less skilled engineers of the future ... but Scotty knows that's just CYA and what it can really do. And of Dolores Umbridge who's Ministry of Magic wants folks to practice defense in a "risk-free way" by only reading and following the strictest of rules.

Back to reality, there are many valid realities here. One of them is: a home builder doesn't typically spec a mobo with 4 RAM slots and then always purchase 4x-whatever kits along with it because one should have no expectation of upgrading RAM in the future because it won't work. That's the prevalent idea being put forth here.

Much more likely, the home builder is spec'ing a 2x-whatever kit for the initial build with an eye toward upgrading by using those 2 unused slots later if desired. In my case here, I didn't even spec the 2nd kit, it just became available from another machine. In any case, this sort of approach to building & upgrading has worked for me for many years (or am I the only home builder here who does that?).

And I can't even say "my approach worked ... until today" because it seems I've ultimately found a solid, documented reason for this one case of mine. It forces one to wonder how many times the real reason is out there, but never found, and then chalked up to "don't expect any RAM to work with any other RAM." I would continue expecting it, while knowing the makers can always hide behind their lack of guarantee whenever it happens to not work.

In any case, thanks for the advice and exchange.
 
Guaranteed? Yes. Sure. Expected? Well, I simply disagree with the notion folks put out there that one shouldn't reasonably expect seemingly compatible RAM to work together. And I don't want to leave future readers with the impression that that is the only truth.

I'm reminded of a Star Trek episode where Scotty shows up in the Next Gen time (or similar) and schools the contemporary engineers on starship specs, because he wrote them to be safe for less skilled engineers of the future ... but Scotty knows that's just CYA and what it can really do. And of Dolores Umbridge who's Ministry of Magic wants folks to practice defense in a "risk-free way" by only reading and following the strictest of rules.

Back to reality, there are many valid realities here. One of them is: a home builder doesn't typically spec a mobo with 4 RAM slots and then always purchase 4x-whatever kits along with it because one should have no expectation of upgrading RAM in the future because it won't work. That's the prevalent idea being put forth here.

Much more likely, the home builder is spec'ing a 2x-whatever kit for the initial build with an eye toward upgrading by using those 2 unused slots later if desired. In my case here, I didn't even spec the 2nd kit, it just became available from another machine. In any case, this sort of approach to building & upgrading has worked for me for many years (or am I the only home builder here who does that?).

And I can't even say "my approach worked ... until today" because it seems I've ultimately found a solid, documented reason for this one case of mine. It forces one to wonder how many times the real reason is out there, but never found, and then chalked up to "don't expect any RAM to work with any other RAM." I would continue expecting it, while knowing the makers can always hide behind their lack of guarantee whenever it happens to not work.

In any case, thanks for the advice and exchange.
In reality, there are three parts to this puzzle. You have the ram, the motherboard, and the memory controller built into the CPU. Now, I agree that there are "tolerances" built into each of these legs. The problem is if the tolerances built into any one of these legs is not met by the other two legs, then that particular combination will not work. The difficulty is determining which "leg" is the one failing. Since the ram is the easiest leg to replace, that is the one that gets most of the blame for not working. This situation is exasperated as speeds become faster, the motherboard and CPU become less tolerant of any differences in the ram.

This in no way means that you can't mix and match ram from different manufacturers, Feel free to continue to expect it to work, although I suspect that the ability of that to "just work" will decrease as other parts of the system increase in speed and complexity.
 

spacemanspiff2000

Distinguished
Sep 8, 2009
32
0
18,530
In reality, there are three parts to this puzzle. You have the ram, the motherboard, and the memory controller built into the CPU. Now, I agree that there are "tolerances" built into each of these legs. The problem is if the tolerances built into any one of these legs is not met by the other two legs, then that particular combination will not work. The difficulty is determining which "leg" is the one failing. Since the ram is the easiest leg to replace, that is the one that gets most of the blame for not working. This situation is exasperated as speeds become faster, the motherboard and CPU become less tolerant of any differences in the ram.

This in no way means that you can't mix and match ram from different manufacturers, Feel free to continue to expect it to work, although I suspect that the ability of that to "just work" will decrease as other parts of the system increase in speed and complexity.

That seems very wise and tempers the message a bit on both ends. Thanks for your thoughts here!
 
This analogy
I'm reminded of a Star Trek episode where Scotty shows up in the Next Gen time (or similar) and schools the contemporary engineers on starship specs, because he wrote them to be safe for less skilled engineers of the future ... but Scotty knows that's just CYA and what it can really do. And of Dolores Umbridge who's Ministry of Magic wants folks to practice defense in a "risk-free way" by only reading and following the strictest of rules.
reminds me that contemporary hardware manuals, as compared to what we used to receive in the '90s and early-oughts, are utter crap.
Who fired the engineers in the technical writing departments?
I need technical data, not the "Huked-On-Fonix" edition.
 

TRENDING THREADS