Sign in with
Sign up | Sign in
Your question

Diagnosing a system that locks up ...

Last response: in Systems
Share
November 16, 2007 5:58:09 AM

Built a new system recently, but i didnt add the new HD yet because im waiting for Vista to arrive, so I just used the hd from my old system with its old installation of WinXP. Took a little "fixing" to get it rnning but it was doing ok. My new system was nice but the two geforce 8800 gt cards were noisy so I put them on their own water cooling rig seperate from the cpu cooling system.

my complete system is as follows:

Asus P5N-E SLI, nForce-650i SLI motherboard (reserator 1 cooled)
Intel core 2 (3ghz) (reserator 1 cooled)
4gb corsair ram, huge heatsinks included - placed in 1st and 3 slots (the yellow ones)
Sounblaster X-fi
2x asus geforce 8800gt (reserator 2 cooled)
Nesteq 600w external PSU.

I added zalman gpu block and zalman gfx ram blocks to the cards, installed them, tested for leaks, booted, started playing, noticed great temperatrues (cpu < 40 / gpu < 48 / mobo < 50 even under stress). Then just as I was planning the "I love my new computer" party, system just locked. And it was the first of many :( 

The lockups are complete and total freezes. No popups, no BSODs, no funny graphic effects, no warnings. It happened watching a movie, it happened running 3dmark, it happened when I opened explorer once. And when it happens the temperatures all look good. The 2 things I did before these problems occured was to strip the stock coolers off the GT cards and add water cooling (Zalmon gpu cooler, and zalman gfx ram coolers on both) PLUS i added the latest drivers to remove a shimmer effect in Team fortress 2 when running sli.

The graphics cards are on a complete seperate cooling system to the cpu, and the new pump is inline with the old but i switch the old pump off. My power supply tells me i'm drawing only 200w which i don't believe, but these crashes occur when doing non intensive stuff so im not sure its the psu as I would assume the more i used the gfx the more power id use and thats when the psu would fail.

The one hope I have is its software because I just moved my old harddisk with winXP on it to my new system and then "fixed" the million driver issues. I have a new HD and am waiting for vista to arrive. Im hoping all my problems will go away, but somehow I doubt it seeing as they didnt appear immediately, but rather appeared after i changed the gfx cooling and drivers.

My current plan while waiting for the new OS to arrive and do a fresh install is this:

1) Remove 1 of the 8800gt
2) Remove 1 of the 2gb dimms
3) remove the soundcard
4) Run the system, if it crashes remove the last 8800gt and reinstall the 7800gtx i have (once i find the stock fan i removed to put water cooling on.
5) if it still crashes change the 2gb ram dimm with the one I removed earlier
6) If it still crashes move the dimm to a different slot
7) If it still crashes it got to be the mobo or CPU so my plan was to cry lots and swear never to go near technology again.

Any advice? I would appreciate any insight you guys have :) 

November 16, 2007 7:06:24 AM

Quick ammendment to this... i think the first thingg I will do is uninstall the drivers and go back to the ones that came with the CD. They had issues with team fortress 2 flickering on certain textures, but perhaps its the new drivers causing trouble, not the hardware. Seems "freezes" are something a number of 8800gt users are experiencing. I dont want to sound happy that other people have the same problem as me, but misery loves company :) 
November 16, 2007 7:22:41 AM

Clear cmos.
Run without the sidepanel on then you know if its a heatissue.

Total frezee is usually PSU-problems in my experience. Try another PSU, you have almost ridiculous power demands, or take away some things and see if the psu can muster them.
Related resources
November 16, 2007 7:31:03 AM

God i hope its not the psu .... was very expensive and i had to ship it from another country and pay the import tax :(  But I will try it if changing the drivers doesnt work. Oh and i havnt put the side panels on yet but I dont think its a heat issue. At least not if the temps im reading from the pc probe and the gfx monitor are to be believed:

idle:
cpu = 32c / mobo = 42C / gpu = 42C / gfx ram = 34C

load:
cpu = 38c / mobo = 48C / gpu = 48C / gfx ram = 40C

These are pretty low temps and the inside of the case feels very cool even considering its got no side panels. Watercooling where the radiator is external to the machine really seems to help.
November 16, 2007 12:36:17 PM

Been thinking about the PSU being the issue.... wouldnt that usually cause the system to crash when you load the system as opposed to being idle? One of the freezes occured when i opened explorer.exe and nothing else... surely the power consumption then was pretty low?
November 16, 2007 1:39:52 PM

In BIOS, set the memory voltage to Mfg. specs, PC6400 is 2.0-2.1v or a least bump up the memory voltage a notch from where you currently have it set which is (default?). Run memtest86 and check for errors if the system still crashes.
November 16, 2007 3:09:04 PM

Ok, quick update :) 

When i built the new system and took the old XP install into it, obviously it went bezerk and i had to reinstall drivers etc. I know I didnt do the 8800gt drivers well because I kept getting a message about of of the applications failing to load. But just before I water cooled my 8800s I updated to the latest drivers to fix a texture bug in TF2 ...... I guess you see where this is going :) 

Now I uninstalled the drivers and reinstalled the ones that came with the CD, carefully this time. And now the total freeze bug seems to be gone. But twice now I have closed 3dmark06 and the system gave the busy icon and locked.... not the same bug thou because the mouse still moves.

So I wondered about the memory, or perhaps virtual memory etc. So I moved the DIMM from slot 3 (I had one in slot 1, one in slot 3) to slot 2. Ran 3dmark no trouble, didnt freeze at the end. Now I just moved the second dimm back to slot3 and ran the test again .. still no bug. Confusing.

I am going to use the mem tester you suggest but I am beginning to think its a driver / "old install of OS" issue more than hardware. If memtester gives no joy I may up the votage as you suggest. But right now I am semi happy, it looks like it isnt the 8800s themselves thats the problem :) 
November 16, 2007 4:05:18 PM

Quote:
I just used the hd from my old system with its old installation of WinXP. Took a little "fixing" to get it rnning but it was doing ok.

LoL - ya think? mcbath! Come to reality.
You are using an old WinXP install from an old rig - massaged for your new hardware? And a complex, hi-perf ultra new rig it is too... you're lucky it ran for you at all. BSOD usually.
Quote:
Clear cmos.
Run without the sidepanel on then you know if its a heatissue.
Total frezee is usually PSU-problems in my experience.

perzy seems to think it's okay.
But I think it's just dumb and a waste of time, all the effort you're putting into getting it to run, and asking us about it.
You need a fresh OS install! WinXP SP2 should take about ~15 minutes on your rig, heheh... not too long. Didn't you say you had a new HD sitting there?
Quote:
Run memtest86 and check for errors if the system still crashes.

LoL!
November 16, 2007 4:06:07 PM

Most likely a fresh reinstall of windows will solve your problems.

In my experience problems like this have been related to memory problems. In two different instances I was able to run memtest86 for hours without a problem yet my computer would still lock up. In one instance I relaxed the timings and that solved my problem, in another instance i think one of my memory sticks was bad and it would run at DDR333 but not the specified DDR400.

If reinstalling windows doesn't help then I would try loosening your memory timings, increasing the memory voltages, and/or decreasing memory clock first.

Could be any number of other problems though from a weak/bad power supply to a bad hard drive to bad drivers. Overall it seems like you have a good troubleshooting plan. Sometimes it seems like things just don't get seated right and you need to reseat them. Try dual core orthos too: http://sp2004.fre3.com/beta/beta2.htm Might want to give a driver cleaner program a try too: http://downloads.guru3d.com/download.php?det=745
November 16, 2007 4:09:09 PM

pshrk...Yes!
November 16, 2007 5:38:55 PM

OK, 3 hours since I removed the driver for the 8800s, and its behaved pefectly :)  TeamFortress 2 has a bug in several textures which is annoying, but at least I know its all driver based and i dont have to mess with my system any further. Now ive got it working I will try to upgrade the drivers with newer ones, but not the newest and see where the instability occurs. It would be nice if I dont have buggy textures on TF2, but obviously ill put up with that until there is a stable driver for my system.

I do have a new hardisk (500gb), just mounted and cooled, and my copy of Vista arrived today so I can resinstall this weekend. Thats why I didnt reinstall already, just wanted to play around on the system and get the water cooling kinks all worked out. But my system seems stable (ill leave it mem testing over night to be sure), its damned fast, the temperatures in all components are impressively low and its almost silent even at full load .... Its a serious relief to know I can wake up tomorrow, install windows vista and then spend the next few weeks complaining about microsoft instead of my hardware :)  The things we do for dx10 and good frame rates eh?
November 17, 2007 1:20:09 AM

Seems I spoke too soon :(  I can now play 3d games etc, stress testthe cpu and so on ... but if i play a movie I wont get more than 5 minutes into it before the system locks. And this is on a fresh install of Vista or on the old install of XP. Now movies are not CPU or memory intensive... and I really think that means they dont push the psu either (am i wrong in that?). So its looking more and more like the graphics cards .....

dammit i was so close...
November 17, 2007 1:39:09 AM

memory voltage. Increase the memory voltage in BIOS. If system continues to crash, try one DIMM only in slot one. Then try the other as a single. You may have a defective DIMM.
November 17, 2007 6:22:04 AM

pshrk said:

If reinstalling windows doesn't help then I would try loosening your memory timings, increasing the memory voltages, and/or decreasing memory clock first.


Did you try these?

Could be a bug in the drivers too... are there different/older versions of 8800GT drivers available? does it still lock up when in non-SLI mode with either card as the primary? It's a long shot but if all else fails maybe try rebuilding the entire system outside of the case. Also wouldn't hurt to try a different PSU if you have another one available.
November 17, 2007 6:38:03 AM

mcbath said:
Seems I spoke too soon :(  I can now play 3d games etc, stress testthe cpu and so on ... but if i play a movie I wont get more than 5 minutes into it before the system locks. And this is on a fresh install of Vista or on the old install of XP. Now movies are not CPU or memory intensive... and I really think that means they dont push the psu either (am i wrong in that?). So its looking more and more like the graphics cards .....

dammit i was so close...


One computer I worked on would lock up every time I tried to burn a CD. Turns out it was a memory error, never showed up in Memtest86 either! Burning a CD should be just about as memory intensive as watching a movie. I had to turn the memory clock down from DDR400 to DDR333, have you tried doing something similar (i.e. make the memory run at DDR2-533 just to see if it is a memory problem, or increase the voltage to something slightly higher 2.1-2.2V, or maybe relax the timings to 6-6-6-18 2T)?

Hard to troubleshoot these sorts of things... keep your chin up, i'm sure you'll figure it out whatever it is!
November 17, 2007 8:40:28 AM

I think you guys are right, it was bad drivers AND a bad memory slot / voltage. After I posted last night I got to experimenting. MY motherboard has 4 memory slots: A1, A2, B1, B2 which are coloured yellow/white/yellow/white. I had memory in the yellow slots A1 and B1 so they had room to "breath". When I moved the memory to A1 & A2 instead I got total stability. The computer has now been on for about 7 hours playing a movie on endless loop at the same time as formating a harddisk.

No lockups :) 

If i get more lockups I will just push the memory voltage up a notch. This probably means my setup is not great for overclocking but with passive water cooling I really just want stability and silence (while still getting awesome fps).

I think we have a winner ladies and gents :)  Now im going to see what performance drop I got from Vista and go drool over Crysis .... THANKYOU guys for all the help and mental support. Much appreciated.
November 17, 2007 7:40:27 PM

If you put the memory in A1 & A2 does the memory still run at dual channel?
November 19, 2007 7:49:47 PM

Hmm... how do I check that?
November 19, 2007 7:52:00 PM

And a better question ... will it make much real difference?
November 19, 2007 9:44:31 PM

Ok, well looks like if i use A1 and B1 it does not run dual channel. Now I put it back in A1 and A2 (the yellow slots), and ran 3dmark06 and got perhaps 100 more score which is negligable.... But when I ran a stress test that comes with the nvidia gfx card it crashed.

So I went to the bios and saw the memory voltage was set to auto, so I have no idea what it was at. I manually set it to the lowest (1.9v) and windows vista crashes once again (this time when booting actually), so I set it to 2.0 and successfully went through the stress test. I assume the auto setting chose 1.9v then. The max voltage for this ram is 2.1v so i have only one more step up before im maxed. Im also hoping that because the ram is now seperated by a slot the increased heat produced will dissipate easier.

I am going to leave it on all night with the stress test just to see if i can run it with dual channels. The I will try to figure out if its worth the heat it produces :)  I tried 3dmark06 with one 2gb dimm, then the two 2gb dimm together and got very little difference even when in th same colour dimm slots.
November 19, 2007 10:39:02 PM

Ok, good to hear everything is working out.

The only way i know of to check if dual channel is enabled is my bios tells me when it posts.
Edit: just realized cpu-z tells you if your running in dual channel or not

From the benchmarks I've seen dual channel does make a difference, just not very big.

I think 2.0v -2.1v is definitely acceptable since that's what your memory is designed for.
Edit: Hmmm... maybe not, this link says it is supposed to run at 5-5-5-18 at 1.8v
http://www.corsair.com/_datasheets/TWIN2X4096-6400C5DHX...
November 19, 2007 11:44:31 PM

i have the same mobo, and some users reported having to manually set their ram timings even on new bios versions. i set mine to the rated spec (4 4 4 12 t2 for corsair xms2) when only one chip installed. then turned the comp off and installed the other chip. everything worked fine. so try that.

other than that, i would def say power OR sli, which is in use even when watching movies, browsing etc. try with just one card and see if it works.
November 20, 2007 5:19:26 AM

Ok, 2.0v was unstable. I have it running on 2.085 now, its stress testing as we speak. The timings on the memory I have trouble checking as everything is auto. Ill find out what my memory should be and manually set it to that.

This only happens though when I use the 1st and 3rd (yellow) dimm slots, not when I use the 1st and 2nd, so I am thinking the motherboard has low voltage on the 3rd and 4th slots (btw slot 2 and 4 crashed even faster).
November 20, 2007 5:37:43 AM

I read up on dual channel memory and it occurs to me that if I have to push the voltage of the ram up by 10% to achieve stability for dual channel, I wont be able to overclock the system as much as if I left it in the non-dual channel slot as the voltage will already be close to the limit.

I could of course just get a better motherboard that doesnt have this problem but the thought of dismantling all the water cooling again gives me the screaming eebie geebies!
November 20, 2007 5:42:17 AM

I checked on corsair and my memory is TWIN2X4096-6400C5DHX which is 5-5-5-18 rated. In the bios memory timings which do those number correspond to?
November 20, 2007 6:03:29 AM

From CPU-Z:

tCL - CAS Latency
tRCD - RAS# to CAS# Delay
tRP or (tRPD) - RAS# Precharge
tRAS - Cycle Time
tRC - Bank Cycle Time
CR - Command Rate

memory timings are usually listed as:
tCL-tRCD-tRP-tRAS

also:
tRC should be at least tRP + tRAS
CR can be 1T or 2T, 1T gives better performance.

CPU-Z http://www.cpuid.com/cpuz.php can give you some info on your current settings as well as the SPD of the memory in each slot. It can also tell you if your running in dual channel i think.
November 20, 2007 6:36:42 AM

mcbath said:
I read up on dual channel memory and it occurs to me that if I have to push the voltage of the ram up by 10% to achieve stability for dual channel, I wont be able to overclock the system as much as if I left it in the non-dual channel slot as the voltage will already be close to the limit.

I could of course just get a better motherboard that doesnt have this problem but the thought of dismantling all the water cooling again gives me the screaming eebie geebies!


Hmmmm... not sure how much you know about overclocking so just ignore this and don't be offended if you already know what i'm telling you ;) 

I'm not sure what processor you have, however I am able to overclock my processor from 2.13GHz to 3.2GHz without raising the memory clock by changing the memory ratio to 1:1 and increasing the FSB. Of course 3.2GHz is the highest my proc can go without raising the memory clock because i am limited by my 8x multiplier. If your chip is better than an E6400/E6420 then your multiplier will be higher than mine.

Also, setting the memory ratio to 1:1 gives great performance, while overclocking the memory doesn't improve your overall performance that much (except in a few select programs like super-pi)

I have an asus board as well, and whats listed in my BIOS as a 1:1 memory ratio is in fact NOT 1:1! In fact a true 1:1 memory ratio is listed in my BIOS as "Sync Mode"

For example a 1:1 memory ratio works like this: 400MHz FSB Clock = 1600MHz FSB (because its Quad Pumped x4) FSB and 800MHz Memory (because its Double Data Rate x2)
November 20, 2007 7:18:21 AM

mcbath said:
I checked on corsair and my memory is TWIN2X4096-6400C5DHX which is 5-5-5-18 rated. In the bios memory timings which do those number correspond to?


You want RAM that does 4-4-4-12, these RAM timings are faster the lower the number. At CAS 4 latency it takes four CPU clock cycles to access data from the ram chips. Get the Crucial Ballistix because they can do 3-3-3-10 800MHz all the way upto 1200MHz.
November 20, 2007 7:28:21 AM

Looks like I have some reading to do :)  I used to overclock back when it was a matter of short circuiting pins to change multipliers and up the fsb and pretty much nothing else. Was several years ago (had a vapochill ... wow the pain!) It seems to have gotten a tad more complicated. All we used to worry about was cranking the volts up and keeping the thing cool ..... 1:1 memory ratios and Sync Mode .... ?? I feel very old right now :p 
November 20, 2007 7:47:38 AM

Yes, though i think he wanted 2GB sticks, in which case CAS 3 is impossible to find, and CL4 can be found but i don't know of any that will overclock to 1200MHz. Perhaps he wanted 2GB sticks to allow for an upgrade or perhaps he is concerned about heat? Timings are important if you have a choice but they aren't really a huge deal in my opinion.

If he wasn't concerned about upgrading to 8GB later (overkill in my opinion) would 4x1GB be a better option for 4GB of ram?

mcbath:

check out some benchmarks to see the difference timings make. This older article for regular DDR was the best I could find:
http://www.tomshardware.com/2004/01/19/ups_and_downs/pa...

I looked up some info on your memory and it is supposed to run at 5-5-5-18 at 1.8v so perhaps your RAM is bad. Link:
http://www.corsair.com/_datasheets/TWIN2X4096-6400C5DHX...
You may want to consider RMAing your memory if you feel it is bad or you want better performance, it is also not uncommon for motherboards to have bad memory slots.
November 20, 2007 7:51:41 AM

pshrk said:

I looked up some info on your memory and it is supposed to run at 5-5-5-18 at 1.8v so perhaps your RAM is bad. Link:
http://www.corsair.com/_datasheets/TWIN2X4096-6400C5DHX...
You may want to consider RMAing your memory if you feel it is bad or you want better performance, it is also not uncommon for motherboards to have bad memory slots.


Make sure and read this part, i'm beginning to think your memory is bad or some of your motherboards memory slots are bad. Any other opinions?

Edit: I bet you've done this already but update your bios if you haven't. (make sure the system is stable before you update the bios i.e. use one memory stick and/or decrease the memory clock)
November 20, 2007 8:29:14 AM

I havent updated the bios because I have no floppy :)  I suppose I can dig one out and install it.

And I ordered the 2 x 2gb of ram mostly because I didnt really give it much thought at the time. Bad move really. I may order 2 x 1gb strips instead. Returning this memory might be hard, its not like I can prove its bad. It works fine if i put it in 1st and 2nd dimm slots and forget the dual channel thing. And that sort of makes me think it may be the mobo anyway.

cpuid rates my ram as follows:

tCL 5
tRCD 5
tRP 5
tRAS 31
tRC 22
CR 2T

Channels Dual
November 20, 2007 8:44:20 AM

My Asus P5N32-E SLI lets my update my bios if i stick the file on a USB drive and use the in-BIOS flash utility. I think it's called EZ-flash in the BIOS. According to the Asus web site the P5N-E SLI has EZ-flash. BIOS updates often fix memory compatability issues.

Also, you said it had problems with the two "yellow" slots, did you try just the two "white slots"? Perhaps one of the yellow slots is bad.
November 20, 2007 9:02:16 AM

The other two slots are even less stable :) 

I think what I will do now is take out one dimm at a time and see what cpuid tells me on each one. Im a little owrried that it registers the tRAS as 31. Perhaps one dimm is bad. But the memory is rock stable at AUTO voltage when in slot 1 and 2 and in single channel mode. So its hard to figure out if its the memory failing in dual channel, or the mobo. And to be honest at this point I dont have the funds left to buy either again and sending it back as faulty good would take a few weeks as they tested it etc.... if dual channel gave a 10% performance boost id think about it. But so far im seeing little improvement.

I havnt overclocked the 8800gt cards yet, or the cpu, so I still have some performance to be squeezed out if I can just get it stable :) 
November 20, 2007 4:55:59 PM

Ok, both memory slots behave the same no matter what. It was when I put them in 2 slots that would allow dual channel that it became unstable so Im beginning to think more and more its a motherboard issue. I have flashed the bios to the last non-beta version and so far so good. Its been a few hours and no crash yet. But it has gone this long before and I got false hope ... :) 

There were some odd sounds on my X-Fi card after I flashed though. Not sure if thats something I just noticed or is new. But im on old drivrs for the X-Fi so its time to fix that anyways.

I did a fast 3dmark06, and it also seems the new biosis about 4% faster on 3dmark06 which is a good enough reason to have done it anyways.
November 20, 2007 8:49:36 PM

Ok, 6 hours and no problem. Played TF2, Splinter cell4, watched a movie, surfed a little ... no problems. Going to leave it doing a stability test all night, then ill leave it on all day downloading stuff. If it does that without a hitch im a happy camper :) 

I now have a stable, silent and damned fast system with everything working as it should. In a few days Ill start overclocking a small ammount and see what I can do, but on the whole, i can relax now.

Thanks again for the help.
November 21, 2007 5:24:13 AM

spoke too soon... again.

Woke up to a black screen that would not wake up. Looks like I had the power saving set up to go off after 20 minutes of idle, and it would not wake up so Im assuming it crashed. Logs show it switched off approx 8 hours after I turned it on. Its more stable than it was, but looks like im still going to have to put up with single channel rather than dual channel.
November 21, 2007 6:32:46 AM

I found this comment out there when some one was talking about the 1gb slightly faster version of my ram:

"Initially I had some troubles when paired these sticks with the Asus P5NE-Sli motherboard because some tweaking must be done in the bios: set voltage to 2.178v and timings to 4-4-4-12-2, rest in auto and disable sli-ready memory. I found this on the Corsair Forums, which is quite useful. "

Perhaps I should try something similar. 2.178v sounds high though.
November 21, 2007 6:44:20 AM

Ok now that im 99% sure its not faulty gear, but imcompatible gear I did a search on my mobo with my ram and got a million hits. Nice to know your not alone, but it seems I may have to live with this in single channel. Neither peice of equipment is broken so I cant return it (had both longer than 2 weeks), so if playing with timings doesn't work, single channel is all I get. Its more stable after bios flash, but not enough.

Look on the bright side, even in single channel mode my new system is 400% faster (sometimes more), than my old system. 2 or 3% performance from the dual channel is not going to bother me too much yet :) 
November 21, 2007 9:34:45 AM

Hope they fix it later with a bios update.

Sometimes incompatability is a perfectly good reason to return something. I don't know where you ordered from but Newegg can be lenient depending on who you talk to. Their website says that you can return for a refund after 30 days (with a 15% restocking fee) or return for replacement after 90 days.

May not be worth the time or effort though. At least now you know what the problem is, and it's not with your graphics cards! Let us know if you learn anything else ;) 
November 21, 2007 10:22:05 AM

Bought it locally here in Norway. I have 14days to return unless its broken. But like you said, it probably isn't worth it. Ill play with the memory timing and votages to see if I can get it stable. Otherwise ill go back to single channel which was rock stable and just overclock it. Stable single channel overclocked will probably out perform unstable dual channel anyways :) 

Ive learned a lot the last few weeks though.
November 21, 2007 6:37:59 PM

Ok, tried manual memory settings and cranking the voltage just below 2.2v..... still not stable :)  Now when you consider that in single channel its stable at 1.9v my brain is telling me that even if pushing it to 2.3v gets me stability in dual channel, id be better off leaving it in single channel and using the voltage to over clock.

Ill keep updating the bios over the next few months, and if it allows me dual channel then all well and good. Otherwise its going to be a pain figuring out exactly whether its the ram or the mobo (or both) and for what? I have to dismantle the water cooling, send stuff back, wait for its return, hope that the replacement has no compatibility issues ... for 1 or 2%? I can squeeze 10% just overclocking.

Ill chalk it down to experience, and in a years time get a new mobo, new ram and this time go through the forums for a few weeks doing my homework :) 
November 21, 2007 7:57:05 PM

pshrk said:
Most likely a fresh reinstall of windows will solve your problems.



This is VERY TRUE.
November 22, 2007 5:29:22 AM

Its a completely fresh install.
November 27, 2007 5:37:58 AM

I did a lot of testing last night. Decided to find out the minimum voltages my system was stable in "single channel" memory mode. I got down to these numbers:

Vcore = 1.175v
NB = 1.208v
ram = 1.92v

Ran 3dmark06, an hour or so of stability testing, played a 45minute movie, 2 hours of TF2 and some Splintercell 4. I'm currently running memtest-86 (6hrs so far no errors). The temps on the CPU / mobo are 36 / 40 when idle, 39 / 41 when under load. The only blip on the landscape is when I pressed escape in TF2 whilst it was loading a level (98% finished loading) it crashed the game. I could not repeat this though. This could be the game, but when presing the "fast load" button in Splintercell 4 it seemed to crash perhaps once every 3 times.

I then ran in "dual channel" mode and to get it stable during 3dmark06 and the stress test I needed the following voltages:

Vcore = 1.35v (actually i didnt lower this so not sure if i needed this high)
NB = 1.56v
ram = 2.175v

Temperatures idle were 40 / 44 and under load were 42 / 46. Again splintercell 4 buggy during fast load. Actually seemed worse but no scientific number on that. Raising the voltages briefly "felt" like it helped but still crashed on and off.

The loading problems could of course just be splinter cell 4 being buggy, and the TF2 crash may have been just one of those things, but my instincts tell me otherwise. And during the perhaps 40 to 50 times I rebooted the system, I had one time it booted with a sound glitch. Rebooting fixed that. Just one of those things or another symptom that my system has problems loading things?

But here is the oddest thing: 3dmark06 showed no difference between single and dual channel memory mode for games playing performance! Can that be right? Because for the temperature increases I saw its just not worth it. Id rather stick to single channel, low voltage and overclock.
November 27, 2007 7:42:18 AM

perzy said:
Clear cmos.
Run without the sidepanel on then you know if its a heatissue.

Total frezee is usually PSU-problems in my experience. Try another PSU, you have almost ridiculous power demands, or take away some things and see if the psu can muster them.


Enter the bios and set the ram speed as low as possible, manually set the timings to slack settings (5-5-5-15) and set the ram voltage to 1.9v.

My ASUS P5B Deluxe/Wifi App by default tries to run my Corsair ram at 800mhz C4-4-4-12 when it should be 5-5-5-15 - manually setting the timings fixes everything.
November 27, 2007 9:29:36 AM

Manually setting it didnt change much for me. But to be honest Ive done some research and it may no longer be an issue. It seems the performance boost from dual channel vs single channel on my system is negligible. Im going to stick to single channel, increase the voltages slightly and OC the thing instead. Just need to figure out if the SC4 crash is a system problem or game problem.
November 27, 2007 8:03:05 PM

Is there any way to see what voltage the "auto" option picks for the northbridge?
!