The issue has been around to some extent since we started using the 6251 card. The problem got much worse when we compressed the system from a 32U rack to an 18U.
The difference in behavior is really in the motherboards. It appears to be a corruption of memory. The first computer has integrated video and the usual crash mode is screen of a given color.. often with vertical stripes, like one row of display repeated across the entire monitor. The second computer has an external graphics card, so is not running system memory in most cases, but apparently has ECC RAM so catches the memory corruption as a parity error.
I do not know if it is a timing issue, bus contention, bad data, or min high /max low violations. Regardless there is definitely a thermal component to the problem. It is too bad that it starts occurring at what should still be operational temperatures.
