Sign in with
Sign up | Sign in
Your question

Prime95 Fails After 7 Hours In Dual Channel + Random Reboots

Last response: in Memory
Share
December 13, 2006 12:28:36 PM

Okay...since upgrading my hardware with a new motherboard, cpu, ram and video card, I've been getting random reboots. I replaced the motherboard yesterday with another one to rule it out and the random reboots are still occuring.

So I run Prime95 and it seems to test fine in single channel mode but whenever in dual channel mode I get a fatal error:

[Wed Dec 13 21:21:10 2006]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.

Hardware I'm running:

Gigabyte 965P-S3 775 Motherboard
Core 2 Duo 1.86GHZ
1GB OCZ Platinum DDR2-800 Ram (2x512)
ATI Radeon 800XL PCI-E Video Card
250GB Western Digital Sata2 Harddrive
160GB Western Digital IDE Harddrive

I will also point out that I'm running the latest bios revision for the motherboard and all drivers have been updated.

The way I see it, I've ruled out the motherboard... but have a tendency to think that the ram is obviously bad. I bought the ram as a set in the same box so they are the exact same dates, batch etc...

I've changed the ram timings to match what's on OCZ's manufacturing website for these chips which is 4-5-4-15 v2.1 ... re-running the test as we speak...but if anyone has any suggestions at all, I'm all ears.

PJStyles
December 14, 2006 6:37:57 PM

Quote:
Okay...since upgrading my hardware with a new motherboard, cpu, ram and video card, I've been getting random reboots.

You reinstalled Windows, right? Upon installation, Windows writes hardware-specific data into a Hardware Abstraction Layer (HAL) which is specific for the hardware it sees at installation. If you change the MB (to a different model), you'll need to at least do a "repair install" of Windows.

Quote:


So I run Prime95 and it seems to test fine in single channel mode but whenever in dual channel mode I get a fatal error:

How did you switch into single channel mode? Or did you change the DIMM config by inserting/removing/moving a stick? If the latter, does the error correlate with any specific slot in use or not in use, or with a specific DIMM?

Quote:
]

Gigabyte 965P-S3 775 Motherboard
Core 2 Duo 1.86GHZ
1GB OCZ Platinum DDR2-800 Ram (2x512)
ATI Radeon 800XL PCI-E Video Card
250GB Western Digital Sata2 Harddrive
160GB Western Digital IDE Harddrive

What model number power supply do you have?
Are you overclocking?

Quote:
....

I've changed the ram timings to match what's on OCZ's manufacturing website for these chips which is 4-5-4-15 v2.1 ...

How were they set before? "Auto" (e.g. SPD) mode?
Related resources
December 14, 2006 9:21:08 PM

Quote:


[Wed Dec 13 21:21:10 2006]
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.


did you check the stress.txt? :p 
have you tried with a different graphics card?
i had issues where my PC would crash and i'd have to restart, or it'd just restart when it had heavy workload, during games and stuff like video encoding.
found out later when i tried a different grfx card what was wrong.
btw, try running memtest on each stick indevidually, and try changing the slots the RAM runs in.
July 24, 2008 1:41:24 AM

Quote:
You reinstalled Windows, right? Upon installation, Windows writes hardware-specific data into a Hardware Abstraction Layer (HAL) which is specific for the hardware it sees at installation. If you change the MB (to a different model), you'll need to at least do a "repair install" of Windows.


This is false, I've done it many times without any issue except that if the hard drive controller upon which the OS holding drive resides is changed, it may bluescreen.

Otherwise, it plugs and plays it all. The HAL depends on the CPU which didn't change and otherwise the power management should be the same. It can be verifed as correct also by checking Device Manager.

I suppose there might be some case where you could have an incorrect boot.ini file that generates a missing or corrupt hal.dll message, requiring a boot to the recovery console to run "bootcfg /rebuild".
July 24, 2008 1:50:57 AM

Try leaving the bios set to SPD/Auto detect but with the manually raised voltage.

After you updated the bios did you also clear CMOS? That should be done. Enter the bios hardware/health page and note the temperatures and voltages displayed. If you have a multimeter check the PSU voltage more directly.

Run Memtest86+ for several hours. First in dual channel mode, stopping the test if it has any errors. If there are no errors let it run several hours, preferribly overnight.

If there are errors, you are saying that just changing from dual channel to single channel mode, meaning you do not remove any memory nor change bios settings, results in regaining stability? I mean in Memtest86+, let's isolate the memory subsystem instead of throwing CPU loading into the mix as is done with Prime95 and by the way, when testing a dual core CPU you should use Orthos to load both cores.

OCZ has a forum where you can post your system specs, the info on specific modules and be advised, whether they know of an incompatibility or what they will do to replace your memory if necessary. If you still have the option of returning the memory to the seller and getting money back, that can also be worthwhile to consider because if it is an incompatibility between the memory and motherboard you may not have a resolution by only replacing two otherwise ok modules with the same exact thing. If that were the situation you'd be better off with the money to buy something else.
!