Data corruption

[jzzcutler]

I've built a new system with an ASUS P6x58D mobo; 12GB of Crucial Ballistix RAM; an Intel I7-920 CPU; and 3 Western Digital WD1001FALS. I put the OS on one drive and then split up the other 2 into 4 250GB partitions each (1,2,3,4 on drive 1 and 5,6,7,8 on drive 2). I filled partitions 1, 3, 5, and 7 with 220GB of data consisting of 22 10GB folders. Then I set up xcopy commands to repeatedly/continually copy from the full partitions on each drive to an empty partition on the other drive (1>8; 7>2; 3>6; 5>4). I let this run for several days to test the throughput and to check for overheating. Everything ran fine without any errors or overheating. After I canceled, I ran simultaneous windiffs on each of the source/target folders pairs. Hence there were 88 simultaneous windiffs running. The windiffs came up with a lot of file differences, but when I recompared the files that the windiffs had found to be different they all came up identical. Thus, something in my system corrupts data when reading and comparing it under the stress of 88 concurrent disk readers but 8 concurrent disk reader/writers copying data from one drive to another works flawlessly. I am now trying the 88 windiffs on partitions on the same drive instead of across drives (1-3; 2-4; 5-7; 6-8) to see if just one of the drives is causing the problem. Is there any other way to determine where the data corruption is coming from: mobo, drive cache, RAM, CPU, windiff software? --Thanks in advance for any help.

 

[GhislainG]

I would suspect that the isue is with the RAM. Any errors when running memtest86+ for several passes? It will take a long time with 12 GB of RAM to test.

 

[jhutchey]

I have the P6X58D also and am using 12G of 1600MHz Corsair RAM. I have run into stability issues whenever I attempt to run my memory past 1333MHz. The ASUS documentation states that OC RAM is not supported with all six slots full, so I reduced my RAM multiplier (and tightened the timings to somewhat offset the loss) and my system is now rock solid.

BTW- I ran memtest86+ for days at up to 1600MHz on the RAM and it would not find any issues.