PC stable with 2GB, Unstable with 4GB?

I'd like to start off by saying I am not overclocking anything (yet ) It's all stock speeds. But this seemed to be the most relevant place to ask.

I just built a new rig this week and with only 2GB installed, it's completely stable. But as soon as I add the second 2GB stick, the PC starts to intermittently freeze. Especially when running memory intensive progs like games. (Far Cry 2, Stalker CS) Yet I have no freezing with only 2GB in and out of games. (Stalker still crashes, but that's just a buggy game) I don't beleive it's OS specific as this happens in both Vista x64 and W7 6801 x86.

Sometimes it will run for hours problem free when just doing 'desktop' tasks, but as soon as I try to run a game, the PC freezes completely, requiring a hard reset. Sometimes it happens after after a few minutes, other times straight away.

Specs:

ASUS P5Q PRO mobo
Jetram JM4GDDR2-8K Dual Channel Kit
Welcome to Transcend Website

(rest of the details are in My Specs)


The BIOS memory timing is set to auto. For the DRAM voltage I have tried using auto, 1.8 and 1.82 and it still freezes. (I think auto is 1.8 anyway)

I have run memtest on both sticks individually, one with 20 passes and the other about 8. Both sticks showed no errors. However when I tried memtest with both sticks installed, it froze completely at just 28% on the first pass.

I have now googled my self silly reading up on this, to the point that nothing is sinking in atm The problem for me is that I'm still not familiar enough with which BIOS settings to alter / test without risking damage. Since a lot of you have a shedload more h'ware OC'ing experience, I'm hoping you can help point me in the right direction

Damn. I think the boards FOOBAR

I just tried running memtest on the second dual slot and I encountered 310 errors in 2 passes. This the same stick that had 20 passes on the first dual slot with zero errors. To double check, I put the other stick in and it wouldn't even POST, let alone let me try memetest.

So I think it's fair to assume the board was faulty to begin with?

Try adding some notch on your NB voltage.

I thought about that, but it's currently set to auto and i don't know what the default auto setting is :huh:

I realise each board is different, but what are NB voltages usually set at when in auto? I'm just trying to avoid setting it at a number that is WAY too much begin with

Hi Patrick,

I'm sorry, but I am inclined to agree with you as regards your motherboard.

smarteyeball said:

I have run memtest on both sticks individually, one with 20 passes and the other about 8. Both sticks showed no errors. However when I tried memtest with both sticks installed, it froze completely at just 28% on the first pass.

Click to expand...

smarteyeball said:

I just tried running memtest on the second dual slot and I encountered 310 errors in 2 passes. This the same stick that had 20 passes on the first dual slot with zero errors. To double check, I put the other stick in and it wouldn't even POST, let alone let me try memetest.

Click to expand...

This would seem to indicate that the second slot is faulty or that the memory controller for that slot is not working correctly. Either way, this is not repairable. I suggest that you try to RMA the board.

Dwarf said:

This would seem to indicate that the second slot is faulty or that the memory controller for that slot is not working correctly. Either way, this is not repairable. I suggest that you try to RMA the board.

Click to expand...

That's my gut instinct. I'm just trying to double/triple check that I haven't missed a simple setting before taking it back. I just checked Everest Ultimate and it's reporting that the NorthBridge core voltage is at 1.1v . Would bumping that up to 1.12 or 1.14v have any negative impact? I'm hoping that since the board is only a week old, they'll just let me swap it for a new one (and not have to go through sending it to ASUS etc) and when they ask me if I've adjusted anything I can honestly say 'nothing dangerous'.

I'm also hoping that since I installed the board myself, they won't try and say 'you installed it, you broke it' (even though I'm 98% sure I didn't). On the plus side, if the board does have to come out, it's a perfect excuse to replace the stock intel cooler with a better one.

Hi Patrick,

I wouldn't have thought that bumping the core voltage up by that tiny amount would have any negative impact. If it had, then why weren't both memory slots affected? Go ahead, and RMA the board.

I haven't adjusted the NB volts yet (I haven't moved it from auto), that was my last trick to try before RMA'ing it. (I haven't really played with the PC for a couple of days:bad a** electrical storms +it was giving me the sh*ts It's only tonight I've started fiddling with it again

I'm just about to try upping the NB volts and that fails, then I'm definitely calling it a dud and will take it back tmw.

Hi Patrick,

Forget about upping the NB voltage. If the board doesn't work correctly on automatic settings, then it is extremely unlikely to do so on modified ones.

Yep. The boards a dud. I was just being stubborn and not accepting the obvious

Thanks mate

Hi Patrick,

Let us know about your new system when you have finished building it. One thing that I would recommend is a surge protector to give protection from spurious high voltages on your power supply.

Will do mate

Funny you should mention that, after that ripper storm we just had (plus all the devices I have connected), a surge protector was next on the list

UPDATE:

I went to to the store where I purchased the parts, told them what was wrong and they had a quick check themselves and swapped the mobo without any hassle I even let them rebuild the system for me while I waited. Even though I did a better job in the first place. (one bloke had even left the cooler power cord lying right on top of the video card which nearly melted the cord

I also ditched the stock intel cooler and replaced it with a Noctua NH-U12P. Damn it's quiet

However when I got back home, i ran memtest 2.10 which gave me 10752 errors on the first pass

So I took out one stick, ran a pass in DIMM slot 1 and DIMM slot 3 which returned no errors. I then swapped the stick and repeated it again which also gave me no errors. (I know one pass is insufficient, but I was pretty peeved and impatient by this stage)

Since memtest 2.10 was reporting my memory as DDR3 and not DDR2, I ran the tests again with version 2.01. Still no errors. I then ran with both sticks installed for 4 passes and didn't have any errors :huh:

So far I haven't had any freezing this time, or POST problems unlike the previous board. I guess extended usage will soon show up any problems though.

IMHO, you are/were having a hardware issue related to the board's default settings not being strong enough to drive 4 DIMMS (twice the modules = twice the voltage drop).

The shop appears to have set (some of??) this up for you when they rebuilt, but for future reference should you have issues:

(1) Manually set all timings to the maker's recommended: 5-5-5-15 = Good "Auto"= Bad

(2) Manually set voltage to the high end of the maker's recommended. You can reduce later on, but in my experience for starters it's usually best to eliminate low/marginal voltage as a variable early on.

(3) Command Rate should be 2T

(4) ***Ensure "tRFC" (Refresh to Activate Delay - sometimes has different initials in the Bios) is set to 54 or greater. *** Seems a small thing, but can have a large impact on stability and voltage requirements


******

The above is normally sufficient, but if that's done and you're still having troubles, then:

(1) Are you running Performance memory?? Some Mobos, especially models not aimed specifically at Overclockers, do not lile running large amounts at speeds higher than 800. If this is the case, you may have to decide between 2GB at 1066 or 4GB at 800.

(2) Bump the voltage to the Front Side Buss (INTEL) very slightly - One or two notches. If you have to go further than this, then you should revisit the first steps.

Hi Scotteq,

I was only using two Dimms (2x2GB sticks). I've read that with this board, using all the four DIMMS was as you suggested, a cause for voltage drop, but i didn't come accross anyone having similar trouble just two Dimms. (Not saying that's not the case though

The shop had not adjusted any of the memory settings in BIOS. In fact, the only BIOS adjustmesnts that were made were to disable the ASUS splash logo, disable the onboard audio and disable the floppy. Everthing else was set to auto.

I manually adjusted the timings to 5-5-5-15 from (Auto) 5-5-5-16. I did that after I had the frirst lot of errors and before I ran the tests again. I'll double check what the Command Rate and tRFC is set to after posting.

The only voltage reccomendation I can find from manufactuer is 1.8 with no min/max levels. JM4GDDR2 8K (not performance memory) Meanwhile the ASUS P5Q Pro is geared towards overclocking.

The first board was dfinetely a dud. They even tried several sets/speeds/brands of RAM and it still froze. The replacement board is not freezing. I'm more optomistic that it's setting based this time, rather than h'ware. Settings can easily be fixed

Thanks for the reply mate.



EDIT:

I changed the timings to 5-5-5-15-3-55-63 and ran memetest 2.01 for 5 passes with zero errors. So far, so good.

(I couldn't find the Command Rate to check what it was set to. Does it go by a few different names?)

KK - Sounds like you have it sorted


The command rate is normally called exactly that - Might be CR, or somesuch. Your choices are 1T and 2T, so it should be simple to identify.

Scotteq said:

KK - Sounds like you have it sorted


The command rate is normally called exactly that - Might be CR, or somesuch. Your choices are 1T and 2T, so it should be simple to identify.

Click to expand...

Fingers crossed

I'm still getting familiar with this new BIOS, so it's probably staring at me right in the face

EDIT:

Everest Ultimate is reporting that the command rate is set to 2T

@ Dwarf : Got a surge protector


Thanks for help everyone

Asus is really picky about what memory works with it's boards. Are you sure your memory is on the QVL list for your motherboard?

blackram_air said:

Asus is really picky about what memory works with it's boards. Are you sure your memory is on the QVL list for your motherboard?

Click to expand...

As I have discovered I am pretty sure it's not on the QVL list. I originally had some some kingston 1066 hyperX that was completely incompatible (wouldn't even post). This set was just a 'test set ' that 'worked' lol

I plan on getting some better QVL 1066 RAM when my budget allows it, but for now, it seems to be happy enough

Good to hear. Just thought I would pass it along.