Long time enthusiast asking for help.

A

Atreides82

n00b
Joined
Jun 9, 2005
Messages
53
I've been on computers since I can last remember. I've been through upgrade after upgrade, some better than others. This latest one however is getting really old. I've had a busy year, finally starting to settle down and get back into gaming. I upgraded last November and it really hasn't ever been 100% satisfying. Specs are as follows.

Intel Core 2 Duo E6850 Conroe 3.0GHz LGA 775
CORSAIR XMS2 2GB (2 x 1GB) 240-Pin DDR2 SDRAM DDR2 1066
Thermaltake TR2 W0070RUC 430W ATX Power Supply
ASUS P5N32-E SLI Plus
EVGA Nvidia 8800GT SSC Edition.
Audigy 2 zs
WD 74G Raptor
Lite on DVDRW

Where do I begin to explain things..

I guess the main problem that has always been here is locking up in-game. Back when Crysis came out it would lock up sometimes 15 minutes into it, sometimes 2 hours. As of today Assassin's Creed and TF2 lock up at random intervals. Sometimes it's a frozen screen with repeating sound. Sometimes it's a blue screen with "Hardware Malfunction, the system has haulted".

I have gotten this same blue screen out of game too, while watching a youtube video. In the morning sometimes I will come in and try to use it and when I turn on the screen, nothing but power save, dead keyboard.

My BIOS is up to date, My video card BIOS is up to date, I have the latest nForce drivers, Video drivers and Audigy drivers. I have installed rivatuner to make sure the card is not overheating, it's not. I run my power off of a 1500VA UPS, so I don't think it's bad power. I have ran my memory through memtest86 overnight with no errors. I've tried using the onboard sound with no change. I reinstalled DirectX. I defaulted my BIOS. I noticed that my power supply only has a 18A +12v rail, so I have put another power supply with a 17A rail to only supply power to the graphics card. None of these things has fixed the problem. I even RMA'd my motherboard because I thought it had to be it, it wasn't.

Sometimes I think I have fixed the problem because I game for 2 hours. I switch games and it happens immediately and I'm back to square one. To say the least it's been an infuriating experience as many of you have probably felt. So here I am, reaching out to fellow [H] members for some support. Something I don't do often :)

Please, Tell me there is something I overlooked. I don't really have access to other components that I can swap in and out, just a friend with a 7900gt which I will do this week.

Thanks!

edit: forgot to mention I'm Running XP Pro SP2
 
Have you tried reinstalling XP? Different motherboard (w/diff chipset would help)? CPU change?
 
A bad OS install is always a possibillity.

That power supply isn't exactly known for its quality. That's what I'm betting it is, but its hard to check for. I bring this up fist because try it with a different PSU if you can. edit: you cant load most PSUs with only 1 rail, you usually need at least a small load on each rail. I've also heard theories about how mix and matching different 12V rails can be bad on a components lifespan, but that guy mighta been full of it.

The fist thing I'd do is check memory. make a bootable CD of memtest86 and let it run on each stick of memory (IE have the other(s) stick(s) out of the system) for 10 hours or so.

Give Orthos a run as your using the computer. People seem to think that a basic repetitive work load is the be all end all for computer stability measurements. Its not. While the processor is at its TDP you want to start pushing weird odd calls down the pipe to make sure it can respond to all of them properly. Orthos doesn't do that, you have to do it yourself.

Look for any bulging caps, If you notice that the light on the aluminum top of a liquid cap is deflected in an odd way it means somethings probably gone wrong inside that cap and it needs to be replaced (asap).

The Nforce 680i NB silicon (identical to the 780i NB silicon), while performing like an F1, dies like an F1. Its simply too high strung for a day-to-day consumers use. The result is one of the highest failure rates of any motherboard chipset ever, in the history of the industry. Erratic problems like the ones your reporting could be due to a motherboard problem. Odds are your still covered by that Asus warranty, but your likely to get one that will burn out just as quickly. The permanent solution is to move to a new motherboard.
 
i would definately say motherboard or PSU, my mate did run a system that has lesser specs than yours on a 430W CM PSU and he was fine, so it may jsut be the motherboard dying...
 
I'll find another HDD laying around and do another Windows install right off the bat. Never seen an install throw these inconsistent symptoms, but hey, it's Windows so who knows.

MrWizard6600: I've ran memtest overnight with both sticks in. Would it make a difference to run it separately on each?

450: Don't have another Mobo/CPU combo with a PCI-e slot. Or even another LGA775 CPU.

I'm leaning tword motherboard, 680i didn't stay around long for a reason. My first experience with an nvidia chipset isn't a good one.
 
Atreides82 said:
MrWizard6600: I've ran memtest overnight with both sticks in. Would it make a difference to run it separately on each?
Click to expand...

Yes. You do get the occasional false negatives when both RAM sticks are tested at the same time. So test each stick of RAM separately.

But I'm leaning towards the mobo or PSU. Neither of those two components are trustworthy.
 
MrWizard6600: I started running the Orthos Blend test, stepped away and when I came back I had this red STOP error.

Test 3, 4000 Lucas-Lehmer iterations of M19374367 using 1024K FFT length.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 5 minutes 6 seconds - 1 errors, 0 warnings.
Execution halted.

Sooo. I guess I will run memtest86 tonight on one stick, and start the other tomorrow morning.
 
Danny Bui said:
Yes. You do get the occasional false negatives when both RAM sticks are tested at the same time. So test each stick of RAM separately.

But I'm leaning towards the mobo or PSU. Neither of those two components are trustworthy.
Click to expand...
Going to respectfully disagree here WRT memory testing methodology. I've had more than one system on my bench that would run sticks one a time for days error free but would generate errors when run together. Turns out in each case that yes, it was the memory that was the issue.

You'll see issues of voltage drop and timing marginality with both sticks installed that you won't see with one stick at a time. Since what is important is the stability of the entire memory subsystem, I always test the memory with all the memory installed. If I get errors, I'll test one stick at a time to try to find the bad stick. But, IMHO what matters is the performance of the whole system, as it's going to be run. Got to wonder if those "false" errors you've seen with both sticks installed previously are false or not..??

However, an overnight run of Memtest86+ that yields no errors is usually pretty conclusive. It's not perfect by any stretch of the imagination, but it's pretty darn good. I always like to recommend the Ultimate Boot CD for testing because the price is right (free), it boots into its own clean OS environment, and it's got a ton of test tools on it. I'd run every diagnostic that it has that is applicable to your system; memory, hard drive diagnostics, and one of the CPU "burn in" tests.

If everything passes, it's time to look elsewhere. The Tt isn't a bad supply, but I wonder that it's a little light for the system. An investment in a solid supply in the 6-800W range would be my next step. Even if you end up replacing the MoBo or other hardware, a new supply won't be a bad upgrade to the system overall.

Are the memory timings and voltages set properly in the BIOS? I've seen more than one system where the BIOS didn't configure things correctly. Is the CPU identified properly in the BIOS? You could sure clear the CMOS and see if that makes a difference.

Also, how are the temps looking?
 
BlindedByScience: Currently running Memtest on the 2nd stick today while I'm at work so we'll see what that shows, I'm guessing nothing. I know the power supply is probably a little light for this system, What I have done to rule that out was use a 2nd power supply to supply power to just the graphics card and hdd. Wouldn't this be a valid solution to system under-voltage? The BIOS is running defaults, memory timing set to auto which CPU-Z shows 5,5,5,18, 400Mhz. The memory is rated 5,5,5,15, 533Mhz but I haven't tweaked it to get that extra mile out of it. I figure running slower than it is rated couldn't hurt, and it just rules out another thing that could be wrong.
Temperature is fine on the CPU, never really above 40-45, 72-77 on the graphics card using Rivatuner. Both are under load.
 
I had the same issue with my 570sli amd platform. North bridge was running too hot and over time began to fail
 
Atreides82 said:
MrWizard6600: I started running the Orthos Blend test, stepped away and when I came back I had this red STOP error.

Test 3, 4000 Lucas-Lehmer iterations of M19374367 using 1024K FFT length.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 5 minutes 6 seconds - 1 errors, 0 warnings.
Execution halted.

Sooo. I guess I will run memtest86 tonight on one stick, and start the other tomorrow morning.
Click to expand...
Boy, this is a bit of a puzzle. As the post above mine said, it could be an overheating North Bridge....that would do it. Yes, using two power supplies "should" eliminate that but the supply you have should be upgraded sooner than later. More than simple voltage drop, the DC rails can sometimes get noisy when under load and that can sure be a problem

If the NB isn't overheating and a new supply doesn't take care of it, I'm leaning towards the motherboard as the source of your trouble. Only way to know for sure at this point is to substitute parts. If you have a buddy that would let you try your CPU and memory in a different motherboard, you could get some info there.
 
BlindedByScience: Thanks for the input. Another interesting observation is when the memory is clocked at 800Mhz, it takes anywhere from 30 minutes to 6 hours to fail that test. If I crank it to it's advertised rating of 1066Mhz, it takes a dump almost immediately.
 
You must log in or register to reply here.
Back
Top