24GB, 16GB in post/windows

watanabe

Limp Gawd
Joined
Nov 7, 2000
Messages
502
Have 2 sets of http://www.newegg.com/Product/Product.aspx?Item=N82E16820231356 in a Asus Rampage Gene III

So far I've:
1. remounted the cpu+HSF(stock while watercooling parts are arriving)
2. tested sticks individually (all work) in sets of 3 (all work)
3. upped QPI to 1.3~1.4v
4. upped vdimm from 1.5v
5. shuffled dimms around
6. Enabled a weird ASUS BIOS option: http://www.youtube.com/watch?v=5AOylHaCyUA&feature=channel
7. updated to newest BIOS (8/27)

but can't get all 24GB recognized at post...

well it worked a few times randomly after I noticed the vdimm had been at 1.65v from the start, and dropped it back down to 1.5v, but then after a reboot it went back to 16GB.

Ideas? Don't really want to RMA as it seems SOMETHING can get it working; just can't figure out what.

And yes, I can use the RAM, was running out at 12GB on my last x58 build.
 
Have 2 sets of http://www.newegg.com/Product/Product.aspx?Item=N82E16820231356 in a Asus Rampage Gene III

So far I've:
1. remounted the cpu+HSF(stock while watercooling parts are arriving)
2. tested sticks individually (all work) in sets of 3 (all work)
3. upped QPI to 1.3~1.4v
4. upped vdimm from 1.5v
5. shuffled dimms around
6. Enabled a weird ASUS BIOS option: http://www.youtube.com/watch?v=5AOylHaCyUA&feature=channel
7. updated to newest BIOS (8/27)

but can't get all 24GB recognized at post...

well it worked a few times randomly after I noticed the vdimm had been at 1.65v from the start, and dropped it back down to 1.5v, but then after a reboot it went back to 16GB.

Ideas? Don't really want to RMA as it seems SOMETHING can get it working; just can't figure out what.

And yes, I can use the RAM, was running out at 12GB on my last x58 build.

That will happen if you're using the Home Premium versions of Windows 7 or Vista. That edition is limited to 16GB maximum. You will need to run Windows 7 Professional or Ultimate or Windows Vista Business or Ultimate in order for Windows to see all 24GB.
 
sorry, I'm using win7 ultimate x64, anyways don't see how that would affect how much ram is detected when the machine POSTS.
 
Try reseating DIMMs 0, 2, and 4. I did this when a few gigs of memory (out of 6 x 2 gigs on my P6TD) weren't being made available by the BIOS to Windows, and it has worked since then.

Something else I was thinking about: did you have a video card lying around that contains much less on-board memory than your active one?
 
Last edited:
The presence of a GPU isn't going to decrease the amount of available RAM by 8GB.

Sounds like a RAM slot might be bad. Verify each stick works by booting the PC with each RAM stick individually. Then verify each slot works by using a single RAM stick and inserting it in each slot.
 
^ The BIOS *might* be incorrectly mapping I/O to the 16 gig mark. There's been plenty of BIOS updates from X58 mobo manufacturers correcting various issues (slow boots, missing system memory, etc) with systems containing lots of memory and video cards with lots of on-board memory.

It's my second guess, for what it's worth. :p
 
Last edited:
A lot of the suggestions have already been done.

night_2004 is on the right track I think.

Regardless of whether I'm trying 12 or 24gb now, only 2 out of 3 sticks (or 4 out of 6) are detected. I think it's the bank furthest from the CPU, as if I try to boot with a single stick in a red slot (2 or 4 or 6) the system refuses to boot with a slot in 6. Also if I take out the ram in slots 5 and 6, the system still boots up with 16GB detected, same if I swap the RAM that was in 5/6 to 1/2 or 3/4.

Can't figure out why it was working for a bit, but guess I'll have to RMA it while I can.

edit: newegg rma submitted. will update when replacement gets here.
 
Last edited:
If you still have problems with a new board, I'd look into the BIOS or CPU next. I highly doubt the CPU is the problem but I've seen stranger.
 
I sort of had the exact same problem on an EVGA x58 Tri-SLI

Had 12GB with an i7 920, decided I needed more, upgraded to 24GB. Worked great. Upgraded to a 980x, only 16GB would show up. Sometimes even 8 or 22.5. Did all the troubleshooting steps you did so many times that I lost count. The bug was occurring at a BIOS level (e.g. only 16 would show in the BIOS). Windows knew 24GB was physically installed but said only 16 was usable. The 16 that were available worked just fine. I once got 24GB to work randomly (Strangely enough, it was at stock settings after I've spent weeks tweaking) but it went back to 16 when I rebooted.

I have two threads here on [H] about the whole issue and one on the EVGA forums. All of the suggestions, while informed and comprehensive, did not help anything at all. I even returned the RAM and got a whole new set (Also G.Skill Ripjaws. 3 kits of 2 x 4GB) and had the exact same problem with that set too.

What ended up fixing the problem was I yanked all the DIMMs out, shuffled them around, and put them back in. Now, if you're like me, you've probably already done this at least a dozen times at this point. It was probably the 20th-25th time that I did it when it was finally fixed.

I know you said you did that already, but it was the exact arrangement/order of the sticks that fixed the problem. I have no idea why. All I can say is that I kept track of which sticks have been in what slots, and the arrangement that worked was one I never tried prior. I now have the sticks labeled with what slot they're in, in case I need to take them out. I've been running solid for somewhere around a month and a half now. I even have my memory overclocked to 1440MHz.

The problems you're experiencing are eerily similar to what I went through (yet so few people have ever heard of this exact problem), so who knows, it might work for you too. I'd set aside an hour or so just to try tons of different arrangements. I went through a solid month of trying to fix this problem. It was a huge pain in the ass. Hopefully it works out better for you! Good luck.
 
Last edited:
I wonder if we will ever know what's up with these missing memory problems. Mine went away after I RMA'd the whole set of memory I was using.
 
What ended up fixing the problem was I yanked all the DIMMs out, shuffled them around, and put them back in. Now, if you're like me, you've probably already done this at least a dozen times at this point. It was probably the 20th-25th time that I did it when it was finally fixed.

I know you said you did that already, but it was the exact arrangement/order of the sticks that fixed the problem. I have no idea why. All I can say is that I kept track of which sticks have been in what slots, and the arrangement that worked was one I never tried prior. I now have the sticks labeled with what slot they're in, in case I need to take them out. I've been running solid for somewhere around a month and a half now. I even have my memory overclocked to 1440MHz.

Considering that there's 720 possible arrangements for six sticks of RAM, that has got to suck...
 
Considering that there's 720 possible arrangements for six sticks of RAM, that has got to suck...

Yes. It sucked a whole lot of ass. But 4 or 5 weeks of troubleshooting had nothing on dicking around with my RAM sticks for an hour. :D
 
Yes. It sucked a whole lot of ass. But 4 or 5 weeks of troubleshooting had nothing on dicking around with my RAM sticks for an hour. :D

Since you tested each stick, and you tested each slot...

When you found a stick that didn't like a slot when all the other RAM was also in, did you test that stick in that slot alone?

I just realized that method would probably shorten that 720 possibilities way down if an individual stick in a specific slot just wouldn't work on its own?

And if it did work on its own in that slot, why didn't it work together with the other sticks there?
 
Well this thread gave me some ideas.

RMA board arrived today, threw all 6 slots in it, booted. 16GB... no worries, didn't do anything in the BIOS so there was some hope left: upped QPI, enabled the weird asus bios option "lock memory settings" or something like that, etc. Reboot. No dice.

AHHHHH


Decided to be systematic about it, so tested each stick in each slot, or was planning to. I already knew the black slots don't work if there's nothing in the red, so I started with the red. The serials on each stick end in 80, 81, 82, 83, 84 and 85. So I used those as my labels. Right off the bat I found 80 and 81 didn't work in the slot farthest from the socket, what I'll call "C." was pretty frustrated thinking this board had another set of bad slots. Got to 82 though and it booted! 83 was good to, 84 no, 85 good.

so 80, 81, 84 didn't work in the farthest slot, and 82, 83 and 85 did.

I threw 82 83 and 85 in the red slots, thinking those were my best sticks that worked in all slots, trying to get the board to recognize 12GB with 3 sticks, think it worked, but it may not have, can't remember at this point, but it may have cause I moved on to testing all 6 sticks/24GB. 16GB again. Dicked around in the BIOS for about an hour, upping/lowering QPI and IOH trying to find some ratio that would work... nothing.

Then it struck me that the black socket furthest from the socket may be 'weaker' than the other 2 like the red ones... took one of the good sticks that was in the red slot closer to the socket, plopped that in the furthest one, rebooted...

24GB!

Would lock up/etc but I think that's because I had upped the volts and the stock HSF couldn't handle the heat possibly. Going to move onto phase 2 now and set up some WC on the cpu/gfx cards; continued here: http://hardforum.com/showthread.php?t=1543862
 
Sounds like you want to get the memory replaced and the previous motherboard was fine?
 
I sort of had the exact same problem on an EVGA x58 Tri-SLI

Had 12GB with an i7 920, decided I needed more, upgraded to 24GB. Worked great. Upgraded to a 980x, only 16GB would show up. Sometimes even 8 or 22.5. Did all the troubleshooting steps you did so many times that I lost count. The bug was occurring at a BIOS level (e.g. only 16 would show in the BIOS). Windows knew 24GB was physically installed but said only 16 was usable. The 16 that were available worked just fine. I once got 24GB to work randomly (Strangely enough, it was at stock settings after I've spent weeks tweaking) but it went back to 16 when I rebooted.

I have two threads here on [H] about the whole issue and one on the EVGA forums. All of the suggestions, while informed and comprehensive, did not help anything at all. I even returned the RAM and got a whole new set (Also G.Skill Ripjaws. 3 kits of 2 x 4GB) and had the exact same problem with that set too.

What ended up fixing the problem was I yanked all the DIMMs out, shuffled them around, and put them back in. Now, if you're like me, you've probably already done this at least a dozen times at this point. It was probably the 20th-25th time that I did it when it was finally fixed.

I know you said you did that already, but it was the exact arrangement/order of the sticks that fixed the problem. I have no idea why. All I can say is that I kept track of which sticks have been in what slots, and the arrangement that worked was one I never tried prior. I now have the sticks labeled with what slot they're in, in case I need to take them out. I've been running solid for somewhere around a month and a half now. I even have my memory overclocked to 1440MHz.

The problems you're experiencing are eerily similar to what I went through (yet so few people have ever heard of this exact problem), so who knows, it might work for you too. I'd set aside an hour or so just to try tons of different arrangements. I went through a solid month of trying to fix this problem. It was a huge pain in the ass. Hopefully it works out better for you! Good luck.

When running 24 gigs of ram with certain motherboards/CPU's, you need to set the RAM speed to 1066, or 1333MHz. I know this has been an issue with eVGA boards, and 980x CPU's because of the stress on the memory controller. It's an odd problem to have seeing that the 980x is a new (should be better?) version of the i7's.

For OP do this:

1: Manually set the voltage to whatever the RAM specifies.
2: Manually set the speed to 1066MHz. If that works, then try 1333, if not, go back to 1066.
3: Make sure the HSF is not putting too much pressure on the board (if it's not the stock one). When after market heat sinks are bolted down to tight, it has been known to cause incorrect memory readings. I think Kyle himself said something about this in another thread.
 
6!
6x5x4x3x2x1

Yes but that is a static equation. Perhaps stick 1-2 will work with stick 2-3, but not with stick 4. So you have 6 sockets, yes, for 720 combinations, but then you also have combinations of RAM that may or may not work with each other. The variable is that 6 sticks of RAM may not be equal. Am I missing something?
 
Yes but that is a static equation. Perhaps stick 1-2 will work with stick 2-3, but not with stick 4. So you have 6 sockets, yes, for 720 combinations, but then you also have combinations of RAM that may or may not work with each other. The variable is that 6 sticks of RAM may not be equal. Am I missing something?

If you're saying you want your two matched pairs of three sticks to always stay together when tri-channeled:

3!^2 = 36

That's still a lot of combinations to fool around with...
 
well what I'm saying is that after you're done with the 720 combinations given that all RAM is equal when working with each stick, you would still need to figure out all the combinations of the combinations. For instance, you would have to check groups of 2 then perhaps stick 1 in slot 1 when stick 2 is in slot 3 and stick 6 is in slot 5, but only when stick 1 and 3 are are in slots 1 and 2. Something like that?
 
well what I'm saying is that after you're done with the 720 combinations

There's only 720 combinations of having all slots filled with one stick of the six sticks of RAM. There are no more combinations.

But he's trying to populate all slots with a stick of the existing six. 6! covers it all.
 
Last edited:
I've always had a hard time getting my head around that. Thanks for the explanation.

But I'm still confused. If the order matters, that is which ram sticks can work with others, that would not be a simple combination. It would mean there were permutations, would it not?
 
Last edited:
Okay, there's only three different possible combinations of how to fill RAM slots, assuming six slots and six sticks:
720 - (6!) - if looking at all RAM as equal, anything can go anywhere
36 - (3!^2) - if the two matched set of three sticks must stick together, but each set of three sticks can be in any order

5913 - (7! + 6! + 5! + 4! + 3! + 2! + 1!) - if any stick can go anywhere, and you can leave slots open (this method checks that

any combination of RAM can work alone or with any other combination of any other stick(s) anywhere, or the slot(s) are open)




36: Now, if you're talking the matched sets must stay together, he's a visual representation of the 36 combos:

(the first three would go in slots 1, 3, 5, the second three would go in 2, 4, 6)

123 123
123 132
123 213
123 231
123 312
123 321

132 123
132 132
132 213
132 231
132 312
132 321

213 123
213 132
213 213
213 231
213 312
213 321

231 123
231 132
231 213
231 231
231 312
231 321

312 123
312 132
312 213
312 231
312 312
312 321

321 123
321 132
321 213
321 231
321 312
321 321


This would keep both pairs of matched sets mixed within their own set, but the two sets never mix with one another.


720: I'm not typing all 720 combos, but he's what it would look like if there were only four slots and four sticks, so 4! = 24

combos:

1234
1243
1324
1342
1423
1432
2134
2143
2314
2341
2413
2431
3124
3142
3214
3241
3412
3421
4123
4132
4213
4231
4312
4321


5913: I'm trying to understand if you're talking about having slots open as well. If you have six sticks of RAM and six slots,

but you can either fill each slot with one of the six sticks or leave it empty, there's 5913 combos.


For a visual of this, let's assume just three slots:

4! + 3! + 2! + 1! = 24 + 6 + 2 + 1 = 33

(zero represents an open slot)

001
002
003
010
012
013
020
021
023
030
031
032

100
102
103
120
123
130
132

200
201
203
210
213
230
231

300
301
302
310
312
320
321


So, if you have six slots, it is 5,913: 7! + 6! + 5! + 4! + 3! + 2! + 1! = 5040 + 720 + 120 + 24 + 6 + 2 + 1 = 5913

Keep in mind that not all of these 5,913 combos will even POST anyway.

But that's all of the different ways you can fill (or not) six slots with six sticks (or two pairs of matched sets of three).
 
That looks a lot more like how I was envisioning the permutations. And if you broke the sets up, such as only set 1 stick 1 will work with set two stick 3 in X combination of slots when all six slots are filled, it gets even greater.

Practically speaking, trying to get a combo that works, if one even does, is a fruitless task, unless you have years on hand or you get lucky.

But the new equation would be something like 6 slots have to be filled in order to get 12GB, since all sticks are 2GB each. So that would narrow permutations down quite a bit. But the correct organization is a permutation because the order and slot is necessary to get a solution. This problem here, with 6 slots always full, is not a simple combination, similar to how a combination lock is not really a combination lock, but a per4mutation lock, given 0-29 numbers and a sequence that must be followed in order.
 
I still don't understand what you're suggesting.

Since you can't explain it well or give a visual and didn't understand the previous math, I'm just going to assume you don't understand what you're suggesting either and leave it at that?

What we're discussing here is highly unusual anyway to have to bother with. I've never had to bother.

But, looking back at the posts, if you determine a certain stick of RAM can only work in a certain slot, then the factorial just decreases by one. Is that what you're asking?

Six sticks, but you determine one only likes to go in one spot, then it's just 5! combos for the others. Determine two must stay in certain spots, 4! combos for the others.

I think you're putting too much thought into this.

I'm still thinking you should just be able to put a single stick into a single slot to see what it does and does not like and go from there.

That's what we did at HP when something was finicky. If that method doesn't work (stick just won't work anywhere), then something's wrong.

Built and tested thousands of systems each day this way. Never had to go to extreme measures of math and combinations to figure it out.

Then again, if something's finicky now, it's bound to end up being finicky again later. So, once a bit of testing's still coming up no good, time to just get different RAM or a different motherboard.

But Proxy made another point about maybe under-clocking the RAM. I've had mobos back in the socket 478 days that had six slots, and if you filled two or four they ran at 800, but you fill all six and they run at 667.

But, we've already put too much thought into this. If I was in these positions, I'd be talking to the RAM/mobo manufs about if an RMA should be considered. Even though they found a combo that works, I won't be shocked if one day it suddenly doesn't work again... MaxBurn's fix was an RMA...

(I do know I have 6GB in my new machine now, and Mushkin discussing issues with 12GB (6x2GB) and having all slots filled is why I stuck with 6GB... Think it was just about timing issues though, but since it's a game machine, eh...)
 
I still don't understand what you're suggesting.

Since you can't explain it well or give a visual and didn't understand the previous math, I'm just going to assume you don't understand what you're suggesting either and leave it at that?

No need to get all butthurt. So then if we have 6 sticks of RAM and there are 6 slots, and all slots must be filled, all combinations = 720 as originally noted? I think I get it now. And yes I was over thinking it.
 
Back
Top