Need help with random crashes and system freezes.

[palouse]

I need help identifying cause of random system shutdowns and lockups. No BSOD received. With the OC indicated in the sig file, or stock/default settings, system goes from fully functional, to shutdown without warning when under load. No system event errors, no popup error message, nothing I can see in any log, ever. When system set with no OC, instead of shutdown, system will sometimes freeze (screen still displaying as normal, but mouse, keyboard frozen,) depending upon load. If game is playing (sound coming from speakers), you can hear the last sound repeating itself in a 60-cycle stutter.

 

Within the last 4 weeks, I installed a X2 3800+ with Freezer64 Pro heat sink (HS). I replaced a Venice 3000+ & stock HS, that ran rock-solid for 20 months at 2600MHz. Rock-solid defined as no application problems, and no error for 24Hr Prime95 and Memtest sessions. I had no issue with the Venice setup, other than I wanted to upgrade to the features of the X2.

 

The Venice ran with BIOS 6/23-3. At X2 install, I decided to continue using the 6/23 and RAM settings, with adjustment for CPU voltage. I installed the AMD CPU driver, XP hot fix, and AMD Dual Core Optimizer as per instructions found on other forums, example. I set BIOS to run at stock settings, and ran various tests, including 12 hour Memtest but no Prime95.

 

Next, I tried the OC in my sig. Memtest passed, but the system would shutdown after a few hours with dual Prime95. Crashes and freezes, so I backed down to stock settings, and started looking at other BIOS settings.

 

Testing with dual Prime95 at stock, the system will eventually shutdown. I had my son try his new Vanguard game (3D) at stock settings, any other applications disabled or off. The system would freeze anywhere from 5 to 90 minutes of starting the game. I have loaded default settings, changed only CPC and Bank Interleave so that I can boot with my 4x512 RAM, all other settings stock. It still fails within 2 hours of using dual Prime95.

 

The following describes some of the steps I have gone through.

 

RTFM: I have about 24 free-time hours into searching and reading forum posts on the issues, outside of time to make system changes and testing.

 

Heat:

I noticed initially that temps were a bit high, relative to suggestions found on the web. I am using MBM5 v5.3.7.0 with NF4 .ini files, and CoreTemp v.94. CPUZ to verify settings, v1.39

OC, dual Prime95:

MBM5: PWMIC 60-68 Celsius, CPU 48-55

CoreTemp: CPU0 up to 50 to 57, CPU1 56 to 65

Stock, dual Prime95:

MBM5: PWMIC 53-57 Celsius, CPU 45-48

CoreTemp: CPU0 up to 48 to 52, CPU1 54 to 61

No load:

MBM5: PWMIC 38 Celsius, CPU 35

CoreTemp: CPU0 up 31, CPU1 27

 

The ArcticFreezer HS does not blow air onto the chipset/PWMIC sensor area. I took off the side panel, and set up an 8” ‘personal fan’ to run full speed directly over the PWMIC sensor (upper right corner of CPU slot). With OC and dual Prime95, the PWMIC temps dropped about 10 degrees, CPU temps about 5. I had my son try his game with 8” fan (no Prime95) and system still shutdown. I had him try fan and game at stock settings. System still froze. Even though I have discounted heat as an issue, I ordered AS5 to use to reseat the CPU, and will continue testing with the 8” fan.

 

BIOS:

As soon as I started looking at the BIOS, I realized it was corrupted. It was not corrupted before the X2 install. The Commands Per Cycle selection in DRAM section was gone! I tried resetting CMOS multiple times, using exact procedures found on this website, including 24 hour reset. CMOS could not be fully reset. Upon flash of the 6/23-3 BIOS using TMod’s CD and 8 hour CMOS reset, CPC command was again available. All testing still produced crash or freeze.

 

I flashed to the 04/06/06 BIOS, given no stop to crashing with 6/23. I was extra careful not to overwrite CMOS with any 6/23 saved setting. I hand-keyed settings back into the new BIOS. The only issue I found was that I originally had the Advanced Chipset Features “APIC SRAT” disabled, and 04/06/06 would not boot XP until set to enabled.

 

I have tried many tweaks, including vCore to 1.5v and down to 1.31v.

 

Current CMOS settings are ‘default’, other than CPC and Bank Interleave.

 

OS:

No, I have not reinstalled Windows. Although often suggested, none of the many forum posts I have read with similar issues that tried reinstall saw any change in behavior. I have dealt with a lot of Windows issues over the years, and have never seen where Windows exits without some sort of error message. AMD driver and optimizer, and MS hotfix are properly installed. OS is otherwise configured properly.

 

Hardware:

Power supply gone bad? Audio card? Video? Hard drive error? Unlikely that something would have died during the time it took to swap CPUs. I have reseated all cables, and ensured that all 4 power supply connectors are in place on the UltraD. RAM is passing Memtest, and has been reseated multiple times. I am careful with static discharge. Fans are running from hard drive connectors, other than CPU fan. I have installed latest ATI 7.2 card and WDM drivers, I don’t use CCC.

 

Voltages at load have little variance, as read from MBM5:

3.3 rail: 3.20v

5v rail: 4.91v

12v rail: 12.01

 

Is it the CPU? Ideas?

[Kingfisher]

What power supply and what video card?

Is it overclocked?

What kind of cooling does it have?

What type/brand/maker of RAM are you using?

[AnUnknownSource]

You say you had no problems until you replaced the CPU? try putting the old one back in, and see how it goes... otherwise the info kingfisher wants would be helpful...

[palouse]

You say you had no problems until you replaced the CPU? try putting the old one back in, and see how it goes... otherwise the info kingfisher wants would be helpful...

sig is in the public profile... I'll click "Show your signature". The sig shows the OC, system currently running at 'stock' except for CPC and BnkInt.

 

Given that a re-seat of the HS is on the agenda and I now have some AS5, trying the old CPU will not hurt. Again, what are the odds that something (other than the CPU) broke between Venice and X2? I was hoping for some thoughts on other options, first.

 

-The PC Health setting for shutdown temp on CPU is disabled. Is there another BIOS option that could be shutting down the power?

-I have MBM5 set to log temps every 5 seconds. I've seen 'hiccups' in the MBM5 readings, where PWMIC or chipset temps spike from about 58 degrees to 85 for one 5 second interval, and never both measures together. No CPU spikes. Out of 1000 readings, there might be 4 or 5 like that, each. Is that a MBM5 glitch, or a sensor problem? I.e., is a spike causing an internal shut down mechanism to fire in the CPU?

-I read where recommended voltage (1.35) plus 10% was 'max'. Should I try more vCore, more than 1.49v?

-etc.

[Kingfisher]

Again, what are the odds that something (other than the CPU) broke between Venice and X2? I was hoping for some thoughts on other options, first.

 

Current CMOS settings are ‘default’, other than CPC and Bank Interleave.

 

The cause of your problem isn't necessarily something is broken but rather wrong memory timings perhaps. Changing your cpu without changing RAM timings, especially your secondary timings couldbe your issue. Running with default settings isn't recommended. Each cpu's memory controller is different.

When you say RAM is passing memtest how long and which tests?

How many errors did you get?

[palouse]

When you say RAM is passing memtest how long and which tests? How many errors did you get?

12 hrs test 5, about 12hrs test 6, about 24 hrs general blended tests. No errors. Turn it on, leave it for the day...

[Kingfisher]

Try using 2 sticks of RAM and see if you experience the same trouble.

passing memtest is great but it still does not rule out unstable RAM. Have you tried stressing only RAM in Prime 95 and/or orthos? If you are stable there your cpu may need more voltage.

You have to play around and experiment. There is no black and white but plenty of grey.

[palouse]

Thank you for the replies.

 

After last post, I went back to original BIOS settings, at stock speeds. I increased voltage on CPU, 1.46v, and LDT, 1.27vjust to try something. System seemed to be a bit more stable, at least failure seemed to take longer. Decided to try one at a time, start with LDT. LDT to 1.36v, and CPU to 1.31v, AND FSB250 & mem divider at 166, my previous OC. LDT multi is 4X. System now 8 hours Prime95 stable. Temps are way down, too.

 

In every post that I have read, LDT voltage makes little or no difference, and vLDT is usually set to minimum, 1.2v. Whatever. It seems to be working. I'll play with this for a while.