Hello,
Since the Keil toolset supports the LPC2478, I thought you might be interested in this problem report sent to NXP:
"Hello, We have an issue with a LPC2478 working at a temperature of 80 degrees and higher. The processor has almost all of its peripherals in use when inserted into the environmental chamber - running an RTX operating system, LCD, UART etc. We see a strange phenomenon: If we use Fcco 288[MHz], and a CPU clock of 72[MHZ], the processor stops working due to a data abort almost immediately after being inserted into the chamber. We tried many other combinations (keeping in mind that we must have USB active at all times, which limits our choices, of course), only to find that driving Fcco at 480[MHz] and the CPU somewhat slower that desired - at 68.57[MHz] "solves" this problem. This issue is worse on machines that have the LCD controller is use so we suspect insufficient head dissipation at 72[MHz], when the processor is loaded. Can you confirm this issue or perhaps conduct further tests?
Kind regards, Tamir Michael"