> Hi Jonathan, thanks for your thoughs. I am still using the same NVRAM, just > with external battery attached, so no Chineese counterfeit. > My hypothesis is: With the battery losing voltage, some bits flip first. They > cause the error message you see and values get set to proper values. But > there are some bytes which must not flip because they determine e.g. the type > of graphics, processor, speed, RAM timing etc. If one of these bits flips > first, than one is lost because the machine does not reach the OpenBoot > firmware because it tries to test non-existing hardware etc. etc.
I wouldn't think that's possible, the small portion of NVRAM used to set parameters is checksummed, so it's unlikely a random combination would also result in a correct checksum. Still, to rule that out, try blanking the NVRAM by writing all zeroes to it. You may be able to do that with something like a TL866+ and a shim socket -- don't just plug the DS1553 right in, as it's got two output pins that may cause a conflict! Thanks, Jonathan