Had to detach one of my systems: null pointer error
26 Jul 2008 12:57:24 UTC
Topic 193779
(moderation:
I have consistenly been getting null pointer error
"0x004c1c42 tried to read memory at 0" from einstein_S5R3_4.26_windows_intelx86.exe. The address of the error can change and 4.26 can fault a few minutes later or run for 7 hours before faulting and can also run to completion without error. I have several other systems running 4.26 and others running 4.38 and this dual-mp system is the only one generating errors. Temp for both cpus, motherboard and disk drive are all normal as reported by speedfan. I assume this is a hardware problem as I dont see any other complaints in the forum.
I set the FSB speed of the MP2600+ down to 200 from 266 and raised all the memory wait states thinking that would help, but it didnt. This system has been running seti and rosetta with all 3 at %33 equal share and neither seti nor rosetta generate any errors.
if this is a hardware problem, I hate to run other apps as they might be reporting bad info. No other apps on this system are reporting any problems.
Is there any type of "burnin" work unit I can run where the output can be validated and I will know for certain if the CPU is malfunctioning? I do not have easy access to this system as it is controlled remotely using VNC but I plan on booting a memory checker program when I get a chance. I have 4 other dual mp systems that are not mis-behaving, two run Ubuntu. I could also switch this system from XP to Ubuntu if that is usefull.
Can I check previous turned in work units even seti and rosetta to see if other the same workunit was passed to someone else and they got an error too? I only paid $75 USD for the mombo & 2cpu's and the previous owner may have overclocked it so possibly it is bad.
Had to detach one of my systems: null pointer error
)
Get and run memtest86+ to see if it is a memory problem.
I recomment running Prime95
)
I recomment running Prime95 stress test for some hours.
See http://en.wikipedia.org/wiki/Prime95#Use_for_stress_testing
On the bottom you find the download link.
cu,
Michael
Thanks! I downloaded both
)
Thanks! I downloaded both memtest86 and prime95.
I verifed the problem was not einstein as rosetta also generate the same type of null pointer error shortly after I detached from einstein.
I had a problem configuring prime95 to set affinity to cpu0 and cpu1 for each of two instances respectively. No matter what I specified, either in the windows xp task manager or in the prime95 cpu config (local0001.ini etc), or both, it was not possible to put one instance on cpu0 and the other on cpu1 and actually get them on different cpu's. The only way I could get 100% utilization of both cpu's was to set affinity in the task manager to let the OS decide which cpu to use.
Anyway, in the process of booting up memtest86 I discovered that ECC was not enabled in the CMOS. I ran memtest86 for an hour with no problem, then reattached to the bam! manager but with ECC enabled in the CMOS. I am going to let it run overnight and see if enableing the ECC fixes the problem. I think the CPU's have been overclocked by the previous owner. When I first powered up the motherboard, I noticed the clock was set to manual and the freq was up near 150. The normal setting is auto and 133 on a A7M266-D mombo. The heatsinks are very top heavy and one cpu is about 6c higher then the other so I am going to lay the system flat so the heat sinks are not hanging at an angle like they currently are.
RE: Anyway, in the process
)
Memory: ECC should only be enabled if your memory is ECC enabled memory. Else leave it off.
CPU: Here the option should be enabled. It will detect and correct single-bit errors in data stored in the CPU's Level 2 cache.
Please note that the presence of this feature in the BIOS does not necessarily mean that your processor's L2 cache actually supports ECC checking. Many processors do not ship with ECC-capable L2 cache. In such cases, you can still enable this feature in the BIOS but it will have no effect.