Originally Posted by LightningManGTS
Originally Posted by Anty
Exactly - GND plane is not made from superconductor
and everybody who works with electronics know how many problems wrong grounding can make.
1usmus - you should measure voltage using GND from given slot as reference - not some random global point.
Very true, but at the same time it should be the same level offshoot across all 4 slots voltage wise unless a2 and b2 are on separate ground planes, if nothing else 1usmus readings are +/- 0.01~.0050 off actual
Comparisons of measurements against an arbitrary ground point should represent valid differences, relative to that ground point, but not necessarily of the voltages actually seen by the cards if we are to assume that the traces have some resistance, particularly if the resistance changes from one slot to the next if they are all at different positions on the same power bus.
In my view, given the VRM quantization level of 22 mV, more or less, if stability seems impacted by differences of 5 to 10 mV, then one's system is not very stable. Stability needs to be assured over the range of temperatures and voltages that can occur during normal operation of the board. I would also assert that small differences in voltage are inevitable, and do not represent some egregious electrical engineering design error. Circuits are normally designed with margin for variation in components and operating conditions, at some baseline, but the margin may be pretty thin once one gets significantly past the baseline, such as into the over 3200 MT/s regime. Even if voltages across the DRAM cards were identical, the components on the cards are likely to not be identical and have different stability margins at a particular voltage and frequency.