Win7 crashes, BSOD, startup failure...the works?

[Athlon64~SPARTA~]

Well all these newer machines require alot more cooling then the old triditional bedge cases from the old days of win 98. If all you have is 2 80mm fans on it, thats not enough cooling and If I were you my eyes would be glued to the temps of the componets. They can over heat and go into thermal runnaway. These newer machines need massive cooling. Especially if you over clock. That increases the heat even more. With these newer machines the cpus and vid cards are real powerfull and if you are not going head to head on multiplayer deathmacthes really theres no need to over clock your machine at all. I run mine at stock becouse its powerfull enough allready. Maybe set up the vid settings for some cod-4 crouch night. But @3.2 gig my cpu is going fast enough allready.

 

Just becouse all the hardware webb sights say over clock, over clock, over clock. WHY? unless you like to burn up your nice machine. You see they don't have to pay for their computers becouse most of all the stuff they get is donated from the mfgers and want the review sight to say their stuff is better so us gamers will buy the stuff and burn it up for no reasion so we can buy more of their stuff keeping the CEOs in rollsroyces and mansions.

Edited January 19, 2010 by Athlon64~SPARTA~

[MedicSN6]

 

For the record, I love Athlons case. If that thing had wings it would be a Christmas Tree/UFO. Looks too cool.

[Ebden~SPARTA~]

I've just completed the first porper LinX test with underclocked RAM.

 

Max temp observed was 58c

 

 

Settings changed:

DRAMv - 1.8

(G)MCH Frequency Latch - 333MHz

System Multiplier - 2.40B

Command Rate - 2 (from 0)

Created=800MHz mem @ 5-5-5-15. Running FSB:DRAM 5:6.

I never adjusted tRC to 24, as indicated in SPD in CPU-Z

 

To confirm if the memory is hosed, I'll bump up the memory to 1066 again and lower the CPU multiplier to 8.5.

Edited January 19, 2010 by Ebden~SPARTA~

[Ebden~SPARTA~]

It may be the memory. Failed test with underclocked CPU and default mem settings, less v

 

 

For this test, memory settings returned to auto, DRAMv set to 2.08. Memory running at 1066

CPU clock set to 2.83MHz, or 8.5x333

Max temp was 56c

 

Should I pop the memory and begin RMA, or are there any alternative tests to confirm?

Edited January 19, 2010 by Ebden~SPARTA~

[MedicSN6]

Run memTest from a boot disc or flash-drive and confirm. That's what I'd do. Also, Vista and 7 should have memory diagnostics in one of the F-key menus.

Edited January 19, 2010 by Medic~SPARTA~

[Athlon64~SPARTA~]

Ya thats probley is whats wrong. One gig of mine toasted, so awaiting return from RMA. Did you try one of the remaining sticks in that bad slot? I think you did.

 

Ya as you can tell I got real tired of boring cases. Blue fans air in, red fans air out. I had some cold catode tubes in there to, but it got to bright to see the monitor. I love it!

 

Edited January 19, 2010 by Athlon64~SPARTA~

[Rooster]

Run memTest from a boot disc or flash-drive and confirm. That's what I'd do. Also, Vista and 7 should have memory diagnostics in one of the F-key menus.

Well, Ebden I think we have it narrowed down. So that is some good news finally! At this point I would decide to either keep your memory underclocked, which will probably work just fine for you, or go ahead and get the RMA. Since DDR2 memory is pretty cheap right now maybe it wouldn't be out of line to order another set? This way once you get your new 4GB set you could then RMA your "bad" memory and still be able to game. If you wanted to test memtest again by all means feel free to, but it's my opinion that you'd be better off to RMA both sticks and start fresh. As you've experienced before memory problems can be pretty tricky to track down.

 

Like others have posted, you may want to look into adding a couple fans to your case if overclocking is still in your future. For my current build I'm using six delta fans on a fan controller which pushes over 500cfm of air. Obviously this can be loud but with a fan controller it can be turned down to silent when I'm not gaming.

 

Good work diagnosing and please keep us updated.

Edited January 19, 2010 by Rooster

[Ebden~SPARTA~]

Since I wouldn't be able to mail the memory for RMA until a few days pass anyway, I can re-run memtest. The confusing part before was that I tested each stick individually without error. Some forums say to use memtest like prime95, and let the tests cycle overnight (or for 24 hrs). How long should I run memtest on each stick to confirm? I could run a single cycle for each stick individually in each slot, eight total tests (2 sticks, 4 slots).

 

As far as overclocking, I only did so to improve performance in ARMA2. Then again, I've been getting that game lock-up whether it was OC'd or not (where I could struggle to task manager to find my CPU at 100%, and memory at 50% or more).

 

Could that last detail still be memory related, becuase it's the CPU usage which had gone all haywire, there was still plenty of memory left for the computer to use.

 

I'm in no position to spend money on new expensive parts, be it memory, cases, and the like. The memory I have costs nearly $100 on newegg.ca, and that's cheap for the category. Add shipping to Newfoundland and it's a big dent. I'm still recovering from an anniversary-christmas-son'sbirthday triad (or should I say triage?)

 

I can probably pull off a fixed-speed fan for $20 or $30 to run from the case's backside, directing more airflow at the CPU cooler fan. Still, I never saw a temp above 70c, so are high temps even an issue?

[Batwing~SPARTA~]

I am still in doubt about the RAM failure.

 

As you said before, you were able to test the RAM sticks one at the time at nominal values with no errors.

Errors are appearing when they are tested in couple on the dual channel config.

 

Clocking down the RAM helps the ram stick, but also helps the North bridge to be stable, which once again lead me to think your problem is not the RAM itself but the NorthBridge on ure MB.

 

Do you have any chance to call a close friend and ask to swap their DDR2 memory to test on your machine? Their RAM is not gonna be compromised, but at least you'll see if they work or not. You should test your friend RAM at nominal avlues on the dual channel config. Either it works or not. If not, close the deal on the MB.

[Ebden~SPARTA~]

I am still in doubt about the RAM failure.

 

As you said before, you were able to test the RAM sticks one at the time at nominal values with no errors.

Errors are appearing when they are tested in couple on the dual channel config.

 

Clocking down the RAM helps the ram stick, but also helps the North bridge to be stable, which once again lead me to think your problem is not the RAM itself but the NorthBridge on ure MB.

 

Do you have any chance to call a close friend and ask to swap their DDR2 memory to test on your machine? Their RAM is not gonna be compromised, but at least you'll see if they work or not. You should test your friend RAM at nominal avlues on the dual channel config. Either it works or not. If not, close the deal on the MB.

 

There's no other way to test the northbridge? What if my BIOS settings are simply wrong, even with default, for my setup? I only had errors in dual channel config immediately after a crash. Once I had pulled the sticks and putzed through a few tests, I could pass a full cycle of memtest86 w/o any errors in dual channel (both 1-3 and 2-4). I'll try those tests again, and I'll also try testing w/Linx using one stick at a time in each slot. Not sure if I can get all the testing done in a day, but I'll do what I can.

[Batwing~SPARTA~]

Not really many ways to test th NB. It is responsible for many things on your MB.

 

However, reading ure specs, there is no reson why ure MB with stock Bios config should not be able to run ure components.

Sometimes MB are not recognizing RAM 1066 right away and they configure it as 800. if u see that at stock settings, don t worry about. Keep it at 800 for testing porpouse.

 

In ure situation is mandatory u have ure system working great at stock setting, if it doesnt u have an hardware failure and there s nothing u can do. Re-test the mem sticks one at the time, at stock, also if they go on 800.

 

After that, u shold have a clear idea about what is working and what is not.

[Ebden~SPARTA~]

Not really many ways to test th NB. It is responsible for many things on your MB.

 

However, reading ure specs, there is no reson why ure MB with stock Bios config should not be able to run ure components.

Sometimes MB are not recognizing RAM 1066 right away and they configure it as 800. if u see that at stock settings, don t worry about. Keep it at 800 for testing porpouse.

 

In ure situation is mandatory u have ure system working great at stock setting, if it doesnt u have an hardware failure and there s nothing u can do. Re-test the mem sticks one at the time, at stock, also if they go on 800.

 

After that, u shold have a clear idea about what is working and what is not.

 

My goal is to have a system that can play the games I want to play. The whiz-bang factor isn't important. If I can simply underclock my memory to 800 and play ARMAII, then awesome! Of course, that isn't the only thing that worries me. As noted in my last post, the game is freezing more frequently now by sucking up 100% of the CPU capacity. Is that memory related, or do I have other issues?

 

What I want to avoid is an unnecessary memory RMA, reinstall, and then repeat of prior problems.

[Batwing~SPARTA~]

Arma 2 as u know is a very intensive application. it will push your PC to the limit and if your harware is not resoponding correctly you can be sure Arma is going to spoil that problem.

 

With that said, the fact that is freezing at 100% CPU usage is not an indication because who knows where your CPU is hanging up. It could be stuck in any of the billions different computation has to do to keep together your machine.

 

The fact important here is that when your PC goes under massive stress, either through the game or using a stress test, your PC fails. This is the buttom line. You need to understand your PC has an hardware failure ongoing, so unfortunately it will fail anytime you will test Arma, due to the massive charge the application push on the components.

 

once again, during test you can try to figure out what is failing only through experiments, and once again, forget about Arma at this moment. Focus on a steady system at stock Bios settings.

 

That means, stress test at stock Bios. If it is failing at stock, there is no way it will work under Arma stress. Doesn t matter if your RAM is 800 or 1066. If your system works at 800 flawlessy you might have a chance that Arma will work. this is not for sure because these stress tess we are using are pushing an overload on specific components at once, while Arma as full application will stretch your PC all components together.

 

The only thing you can do with the tests is being sure (under reasonable doubt) that ure RAM is working fine, at least at stock. Then if the game is still failing, once again, go back to your MB main issue.

[Athlon64~SPARTA~]

I hate to say it but your board may not be compatible with 7, it says its compatible with vista so it should be with 7. Also I dident see your ram on the ram compatible list for the board as well.

Edited January 19, 2010 by Athlon64~SPARTA~

[Ebden~SPARTA~]

I've done a little searching on how to test the northbridge, but no luck so far.

 

It's beginning to sound like this isn't going to happen. Even if I RMA the memory, it isn't on the mobo compatability list.

 

After looking at the memory compatability list again, I could have sworn my memory was on it.

 

Funny story. I originally bought 2G of OCZ Vista Upgrade (budget) memory, which IS on the list. I had to RMA the memory, but as I learned more about the system the OCZ memory itself was not desirable, so I sold it on e-bay and bought the current memory (without re-checking compatability, shit).

 

I could RMA both the memory and the mobo. I can't confirm which is bad, and I'm going to stop testing because they are incompatable anyway. Given the number of BSODs, I've probably killed the northbridge and the memory, so if I'm lucky then a successful RMA on both awaits.

 

Since they're incompatable, no point in reassembling the new components. Which should I sell and which should I keep, the mobo or the memory?

 

I have zippo budget to replace any components, nor do I have any techy pals with extra gear lying around nearby. My option is to sell one or the other, and see what I can do with the amount recieved plus a little extra to boost. Since it was a budget build from the start, I'll be spending about the same on replacement components, whether it's memory or a mobo.

[Batwing~SPARTA~]

Ok, we still have a situation here I do believe is a bit misunderstood.

 

First of all, I guess you had this PC working for a while, then suddenly it started to fail.

Is it correct u have been able to play Arma with this config for while?

 

I am not sure if I got it right, but if this config have been working for a while, this defeat any conceptb of incompatibility.

 

Other important point: doesn t really matter if the MB is 7 compatible, because when u fail testing BIOS starting from an external boot, you are not even touching 7 at all.

 

Then, "incompatible memory" is quite a myth. Especially when u r talking about consumer chips, they are all made in Taiwan in the same chain. Then they put a brand on it. So, when u read the compatible list, is a matter LtestedL which doesn t mean another brand is not gonna work.

 

If I was u however, I would try to get rid of that MB ASAP. Rma It and then if u have to just sell it.

 

When u will rebuild ure system, do not overclock with the case u have, or open the side and stick a 20 cm fan onvthe side.

 

You may risk to fry also new components if u dont pay attention.

[Rooster]

Just underclock your RAM and use it then. I doubt the problem lies within your motherboard for various reasons, one of which is that you pass linx 20 pass. Linx is highly regarded as the most stressful situation your CPU can be made to compute. Intel uses linpack (a linx derivative) for it's binning process for this reason. It's how an e8600 is deemed better than an e8500. So, rest assured that Arma2 or any program out now, will ever come close to dishing the CPU punishment that linx does. Since you're using higher density modules it is quite plausible that one module on a single chip is clocking poorly which is causing the issue. Also, all RAM is not created equal. Some boards require 2.1vdimm RAM (rampage extreme for one) which is not compatible with the lower 1.8vdimm boards. Asus also had a striker extreme board that required the higher 2.1vdimm compatible RAM. So it makes sense that the 2.1vdimm RAM you're using is just not compatible. A member of my overclocking team sent a pair of crucial ballistix back to crucial and they RMA'd no problem. Gskill is a good company and I do have some contacts that might be able to help you with that.

 

The reasons I believe it not to be your motherboard is because for one, motherboards are pretty straight forward devices in how they act, and do some checking each time you boot. When you post, the board will do some basic compatibility tests, then clock the chips and set voltage according to vid. I think if you were having a motherboard issue then the problems you're encountering wouldn't be as random as they are now... I've ran your board and still use it with an ek supreme, PA 120.2 rad, iwaki pump, and dangerden res. with the same CPU as you. Point being that I've had this board under -72c LN2, and run well over the limits designed by gigabyte. It's a tough board capable of handling over 80c on the NB alone before it starts to error. This setup serves me well as my water cooled backup system these days. My GA-EP45UD3 board even survived some bent pins in the socket from when I knocked a VGA pot (copper cooling container) into the socket area.

 

I believe you have a RAM problem, but if you would like to test the board just pump up your FSB and turn down your multi to 8 or 7x. Turning up your FSB is the single best way to stress your board, mainly by forcing the NB strap to a higher speed, lower latency connection. Your board is capable of at least 360 FSB (7x360=2.5GHz so still underclocked), I'd say if it didn't do that then there could be a possibility of a problem with it. This can be tricky because boards have dead zones, where they may run at 360FSB but not 359FSB... But again, your RAM could still limit your FSB/MB test if it's over speed, vdimm, or timings. RAM is far more sensitive to heat, ESD, and all kinds of finicky problems. If you can't afford the RMA/upgrade then just run at 800MHz. The difference in gaming is very nominal if at all present. Besides at the end of the day our trusty linx test proved your RAM to be faulty in the config it was in prior to turning the speed down.

 

R

Edited January 20, 2010 by Rooster

[Athlon64~SPARTA~]

Ide agree with you except the ram has to be compatible with the board, I think that the MGFER has tested what ram will work with their board and there is truth to what they say. Ive had boards the wouldent even boot with incompatible ram in it. And why does asus ask what operating system is being run on a certian boards for compatibility. So?

[Ebden~SPARTA~]

In the last three posts I see a couple things.

• Two guys who know lots more about this than I disagree on the root issue, Mobo NB or RAM.
• Two other guys who know lots more about this than I disagree on if the memory list is for memory TESTED or memory COMPATIBLE

 

In the meantime, I can look into a couple fans at the local compu-ripoff store (being broke and being poor are different things. Consider my income 'fully allocated').

 

One option that I'll look for, it must only be a couple bucks, is a teensy fan like this

 

 

I got the idea from the associated Tomshardware forum post on overclocking, which I used when originally building this system.

 

The response from Gigabyte on an inquiry last week is as follows. It does not make any comment about the memory type as incompatible, which gives me cause to follow Batwing's hypothesis on a list of tested compatible RAM, rather than only compatible RAM. (I never listed G.Skill manuf, but if compatibility was an issue, they would have mentioned it)

 

Dear customer,

You may need to check memory voltage spec and adjust it match in bios. Mother board bios memory voltage setting is base on 1.8v ( default voltage ), please check current model memory voltage spec if it's higher than 1.8v go in bios under M/B intelligent tweaker memory voltage control item adjust memory voltage to match memory voltage needs .

For example: current memory is 2.1v go in bios M/B intelligent tweaker on memory voltage control item change from 1.8v to 2.1v, this will match the current model memory voltage spec, system will run stable. Please also check cpu temp in bios under PC health status, normal cpu temp should be 37-59 C, if over 60C it considers as overheated.

Question - 863122

From : Edward Stern [ stern.ed@gmail.com ]

Sent : 1/11/2010 21:08

Question : I suspect a failed memory slot (DDR2_3). How can I test to confirm a mobo issue? I suspect this slot is failed because memtest86+(v4) tested my memory successfully from slot A0 (DDR2_1), but with both RAM sticks installed I get errors immediately with Memtest. This is the case when I reversed the placement of the memory sticks in slots 1&3. It gives me reason to suspect an issue with RAM slot 3. I currently am running with both sticks not in pairs, running in slots 1&2, but at lower efficiency. Please advise.

 

 

-------------------------------------------------------------------------------

Model Name : GA-EP45-UD3L(rev. 1.0)

--------------------------

M/B Rev : 1.0

BIOS Ver : F4

Serial No. : 4719331SN091100031682

Purchase Dealer : newegg.ca

-------------------------------------------------------------------------------

VGA Brand : Nvidia Model : Palit 9800GT

CPU Brand : Intel Model : E8400 Speed : 3.0G

Operation System : Win 7 64-bit SP : build7600

Memory Brand : Type : DDRII

Memory Size : 2048 Speed : 1066

Power Supply : 450 W

 

Memory is1066(2.1v) 2x2068mb. CPU overclock to 3.6Ghz(v1.168)

[custard~SPARTA~]

There is definately going to be a heat issue in your case even without any overclock, the cheapest solution is take the side off it and use a desktop fan pointed straight at it.

[Cylawyer~SPARTA~]

Which is also the easiest way to test that theory custard. I also agree it seems to be a heat issue and I would take the side off, and put two or three box fans a few feet away from it blowing at full bore right into the case through the open side and start up and give it a test to see how it runs. Costs nothing (if you have the fans).

[Athlon64~SPARTA~]

Run the right o/s and ram thats on list, It might work right, maybe.

[Ebden~SPARTA~]

Which is also the easiest way to test that theory custard. I also agree it seems to be a heat issue and I would take the side off, and put two or three box fans a few feet away from it blowing at full bore right into the case through the open side and start up and give it a test to see how it runs. Costs nothing (if you have the fans).

 

I did that over the summer.

 

What I haven't heard yet is a treatise on acceptable temps. How hot is too hot? How can I monitor the NB temp?

 

EDIT: After days of testing, the current diagnosis is expanded to include heat as a potential contributor. The current state of affairs is as follows:

• My memory may be faulty.
• My northbridge may be faulty.
• High heat may be causing my memory to malfunction.
• High heat may be causing my northbridge to malfunction.

 

Isolating the issue is not possible given the imprecise testing methodology available. I'll install some fans and see what happens.

Edited January 21, 2010 by Ebden~SPARTA~

[Batwing~SPARTA~]

Unfortunately you are getting to a dead end here. is not your fault, but you don t have hardware available to test and get to the final solution.

 

What you have is;

1 MB

1 Set of RAM

 

you don t know which one is failing, but:

 

Continuing testing on a failing MB will continue to report errors also if your RAM is good

Continuing testing on failing RAM "should" give you at least one stick good and the other bad IF the MB is ok and IF your 2 sticks of RAM are not failing simultaneously (very unusual)

 

However, to really get to the final of this difficult situation, you need another known-working MB and another known-working set of RAM, to switch back and forth to see which is which creating issues. There are no other solutions unfortunately and that s why sometimes the only way to achieve the solution is to take the machine to a local PC repair store which has that hardware to test all of this, and having them providing the final virdict.

[MedicSN6]

Send the whole thing to me and just start over. PM me for my address.

 

Medic's idea. Best idea.

Edited January 21, 2010 by Medic~SPARTA~