I don't need to have a reason to RMA an Intel product, they are very good about that with partners. However this issue seems to only effect RIS preloads; as such machines that fail the RIS load can still be imaged via PE for example. The timing issue, if indeed that's what it is, is what I am trying to flesh out more before reporting it back to my intel tech sales rep inside. Failing that I can work with quadsupport to flesh it out further, but it takes time, and I am a bit short of that these days. This isn't dramatically different then say BOXD845GVSRL which had a somewhat similar issue with RIS however in that case just under 5% of the last batch revision would refuse to get DHCP during RIS. Again an issue only during RIS preloads, no where else was it a factor. Very strange none the less.... That is a very good question on editing those timings, I will have to dig into the ROM options to see. However bottom line when identically configured hardware exists, they should behave identically... These do not, hence a design of manufacturing flaw exists in the 945G series.