BSOD only when all slots full

C

courteous

Weaksauce
Joined
Aug 31, 2009
Messages
110
Relevant pieces:
  • Asus P5K (PAE enabled)
  • 4 x "identical" 1 GiB [ame="http://www.amazon.com/Patriot-Signature-PC2-6400-Desktop-PSD21G8002/dp/tech-data/B000MV15G0/ref=de_a_smtd"]Amazon.com: Patriot Signature 1 GB PC2-6400 DDR2 800MHz Desktop Memory PSD21G8002: Electronics@@AMEPARAM@@http://ecx.images-amazon.com/images/I/21fZRFbJNrL.@@AMEPARAM@@21fZRFbJNrL[/ame]

I know that MB's slots are OK, because I tried all combinations; I get BSOD only with all 4 slots filled ... PC boots into OS, then crashes at random (I'd guess when OS gets garbage bits from one module).

I've ran MEMTEST+ (v4.10) on them: on one "pass" there were errors on lower byte (FF), on another "pass" no errors.

So it would seem that
(1) the power supply isn't good enough
or
(2) the voltages (which are on "Auto" in BIOS) aren't correctly set? Though CPU-Z shows them at correct voltage of 1.8V ... so I presume this should be left untouched (as Patriot says to have them @1.80V)? :confused:
or
(3) should I change the voltage of the "north" bridge (MCH)?

PS: I know it's a common issue ... I've read (too) many threads already, alas no solution. :(
PPS: I'll be replacing power hungry GPU with a passive one: any chance this will soothe the problem? :)
 
Last edited:
Mike (Official Corsair Tech Support :cool:), thank you for the reply. I'll follow your advice ... but I'm also interested in why this should/will work? For instance, does upping the voltage increase the S/N ratio (and therefore decrease the chance of garbage bits)?*

*Just an (un)educated guess from a novice (but curious) CE. :p

EDIT: Also, shouldn't MB's "Auto" setting automatically set everything accordingly when it "sees" the 4GB (or any other number) of RAM? If not, is it because of the variety of memory modules (and other pieces of hardware)? But if so, couldn't they rely on SPD data (in case of RAM; similar info. surely exist for other hardware)? Or does it all just come down to the fact that the (drivers) programmers are too lazy and/or the art too hard? ;)
 
Last edited:
Most BIOS' come tuned to run 2 modules. 4 modules places 2x the load on the memory controller. In some cases, you need this small voltage bump to be stable at the top speed of the memory controller with it fully loaded.
 
You said to "bump 1-3 steps". Well, BIOS only has 4 voltage levels for MCH (1.25V, 1.40V, 1.55V, and 1.70V). :) Also, the BIOS' setting on "Auto" doesn't show which setting is currently used ... which would you say it is? Is it 1.25V? (EDIT: Just found out it is indeed 1.25V using Everest trial version. Hope I can rely on this info ...)

And can I go (if need be) all the way up to 1.70V? I mean, is it "safe" ... of course, I expect that 1.40V (or 1.55V) will already make RAM stable.
 
Last edited:
OK, I've just increased the NB voltage to 1.40V in BIOS. The Everest (or Aida as it is now called) still shows only 1.25V! Can OS (Windows 7) override BIOS settings? :confused: Or is this just the program's misinformation?

PS: If the NB indeed is @1.40V it sure made a lot of change in temperature: MB went from ~40°C to ~46°C (could this difference be more reasonably explained by the added 4th RAM module?), while CPU went from ~30°C to ~26°C. :eek: (All data per HWiNFO64 program.)

Is it correct, that the 4th RAM module added the majority to MB's temperature (and not the NB increased voltage from 1.25V to 1.40V)? :confused:
 
Last edited:
:( Even after increase to 1.55V I still get BSOD. And I doubt increasing it further to 1.70V will help either (if it is safe anyway; even the 1.55V voltage increase certainly heightened the CPU temperature for about 6°C :)).

PS: AIDA64 still reports "North Bridge voltage" as 1.25V, though.
 
Have you tried running a memtest with all of the slots full to rule out a bad bank?
 
courteous said:
I've ran MEMTEST+ (v4.10) on them: on one "pass" there were errors on lower byte (FF), on another "pass" no errors.
Click to expand...
Yes, once there were errors ... on second "pass" no errors. And I also switched modules back-and-forth, so whichever 1 of these two "new" modules I put in (i.e. running with 3 slots filled) ... everything is OK. There is (almost) certainly something wrong only when running with 4 slots.

I took a picture of my relevant BIOS settings*:
dsc01140tw.jpg


* The somewhat low CPU voltage was set after OCing E6420 @ 3.2 GHz. But this BSOD-when-4-slots-full problem started way before the OC! Though (regarding the OC) I must say, that PC boots only the second time I press the "power button". :confused: Again, this isn't related to RAM problems ... well, it could be now, but this problem started before.
 
Last edited:
Why are you increasing the northbridge voltage again when it's a ram issue? You should try increasing the DRAM voltage above the stock 1.8v and see if that fixes your stability issues. Try something more extreme like 1.875 or 1.9v and if your problem goes away then you've solved it. Then back the voltage down one setting at a time until you start blue screen again to find out which setting is stable without overvolting the hell out of it.
 
Try 1.85v DRAMv and 1.25v NBv. If that fails, try other combos. 1.9v DRAMv, 1.25v NBv, or 1.85v 1.4v NBv and you'll find the spot you need. Your RAM is rated at 1.8v so 1.8v in BIOS "should" be enough but a small bump may help. But, I think you definitely will need 1.25v or 1.4v to get 4 modules stable. And, 1.7v is too high on the NBv. Don't go above 1.55v.
 
I've increased from Patriot's 1.80V to 1.85V (smallest step available) and NB back to 1.25V: it seems somehow better ... for now; BSOD hits you when you least expect. :p Will certainly let you know if it goes unstable again ... but, definitely an improvement from before.

I must say that I'm very happy (and lucky) that I got the help from Official Corsair Tech Support ... Mike Clements, thank you. And to all of you here on HardForum!
 
BAD_POOL_CALLER @ 0xC2 :(

BSOD image:
dsc01145m.jpg


Now I increased DRAM voltage to 1.90V (I shouldn't go any higher, huh?) and left NB voltage @1.25V ... hope it'll behave. :)

If this fails, then I'll (as Mike Clements advised) lower the DRAM voltage back to 1.85V and increase NB to 1.40V.

Should even this fail, then what? :confused::(
 
Mike Clements said:
Remove 3 modules, leave 1 in the 0 slot. Test for 3-5 passes each with Memtest from www.memtest.org

Repeat for all 4 modules. Post the results.
Click to expand...

+1. I was having an issue like this where I seemed to get BSOD only when I used more than 1 stick of memory. It turns out that I was just super unlucky and had to RMA my memory 3 times before I got a working set. Originally I ran mem test and got no errors so I figured it wasn't the memory, but after a couple months of random BSOD I tried it again one stick at a time. Sure enough, one of the four sticks was bad.
 
Yes, but the strange thing is that any combination of 3 modules runs fine for weeks ... then put in the 4th one ... KABOOM in less than an hour!

Nevertheless, I'm still going through the MEMTEST one-by-one ... can you tell me where the logs will be saved? Or do I have to take pictures of the monitor? :D
 
If the memory modules aren't bad, could it be that the VRM for the memory are going bad on the mobo?

Do you see any burn marks or anything out of the ordinary on the board?

That's really odd that three will work of any combo, but adding a forth will make it fail. It almost sounds like the VRMs my be overloaded imo.
 
courteous said:
BAD_POOL_CALLER @ 0xC2 :(

BSOD image:
dsc01145m.jpg


Now I increased DRAM voltage to 1.90V (I shouldn't go any higher, huh?) and left NB voltage @1.25V ... hope it'll behave. :)

If this fails, then I'll (as Mike Clements advised) lower the DRAM voltage back to 1.85V and increase NB to 1.40V.

Should even this fail, then what? :confused::(
Click to expand...

Check out this thread, and see if it helps.

http://www.sevenforums.com/crashes-debugging/67198-bsod-bad_pool_caller.html

Did you get a windows dump of the problem?
 
Last edited:
Red Falcon said:
If the memory modules aren't bad, could it be that the VRM for the memory are going bad on the mobo?

Do you see any burn marks or anything out of the ordinary on the board?

That's really odd that three will work of any combo, but adding a forth will make it fail. It almost sounds like the VRMs my be overloaded imo.
Click to expand...
Red Falcon, thank you for pointing that out. Is this the VRM (Voltage Regulator Module) you pointed out: Asus P5K series - VRM / Power ciruit cooling? With my current electronics knowledge, I can only check, say, if capacitors were blown out. :) Again, thank you, will certainly check this out.

BTW, are the processor voltage regulators (these 8 MOSFETs on P5K Deluxe) doing a similar function? Perhaps a layman explanation will do (haven't passed my electronics course ... yet :p).

msny said:
Check out this thread, and see if it helps.

http://www.sevenforums.com/crashes-debugging/67198-bsod-bad_pool_caller.html

Did you get a windows dump of the problem?
Click to expand...
msny, yes I did (have to download .dmp viewer to look at it). Though, I'm not sure why the majority of BSODs doesn't get dump log? Perhaps you know why? :confused:
 
courteous said:
Red Falcon, thank you for pointing that out. Is this the VRM (Voltage Regulator Module) you pointed out: Asus P5K series - VRM / Power ciruit cooling? With my current electronics knowledge, I can only check, say, if capacitors were blown out. :) Again, thank you, will certainly check this out.

BTW, are the processor voltage regulators (these 8 MOSFETs on P5K Deluxe) doing a similar function? Perhaps a layman explanation will do (haven't passed my electronics course ... yet :p).


msny, yes I did (have to download .dmp viewer to look at it). Though, I'm not sure why the majority of BSODs doesn't get dump log? Perhaps you know why? :confused:
Click to expand...

Yeah, those are the very ones.

The ones for the CPU do the same exact thing, only on different voltage levels and variations.

I'm not sure about the log files, it sounds like the system on a whole locks up when the voltages become incorrect/overloaded/etc and cannot write to a log file.

Do you have any other systems to test all 4 modules in at once? If all 4 can work properly in another system, then it is definitely the VRMs on the mobo going out, or they were just faulty to begin with and cannot handle the load of all 4 modules.

Just my opinion from what I am seeing. I hope this helps in some way.
 
This idea does help (I wouldn't thought of it). Indeed, even if I don't have a 2nd comparable (with DDR2-800 support) MB, I could cheaply get a used one (and while am at it, use it to build a "new" PC for my father). I'll go and see what's on sale.
 
Mike Clements said:
Remove 3 modules, leave 1 in the 0 slot. Test for 3-5 passes each with Memtest from www.memtest.org

Repeat for all 4 modules. Post the results.
Click to expand...
Just done MEMTESTing (v4.10) as Mike Clements instructed me to.
All MEMTESTing was done with DRAM voltage @ 1.85V and MCH/NB voltage @ 1.40V (stock @ 1.80V and @ 1.25V, respectively).

  1. module (1st "old"): _5 passes - no errors. [ photo ] (All photos @ imageshack.us.)
  2. module (2nd "old"): 10 passes - no errors. [ photo ]
  3. module (1st "new"): _4 passes - no errors. [ photo ]
  4. module (2nd "new"): _5 passes - no errors. [ photo ]

Also took some pictures around the area of RAM slots:
Anything obviously out-of-place? :confused:

Where do I go from here? I am really stumped. :(

EDIT: To sum up what voltage settings (on DRAM and NB; stock @ 1.80V and @ 1.25V) I've already tried (and failed to bring stability) with all 4 slots full:

NB _________ DRAM
*****************
1.40V _____ stock
1.55V _____ stock
stock _____ 1.85V
stock _____ 1.90V
1.40V _____ 1.85V
 
Last edited:
You must log in or register to reply here.
Back
Top