I've designed a motherboard sized PCB (.8sqft) with 24 DDR2 DIMM modules running on it.
I'm having trouble figuring out how to calculate the temperature rise. My calcs come out to 128F temp rise over ambient, but similar equipment only rises around 10F above ambient.
Anyone know how to model this? The DIMM modules are 5.2"x1", so each of those can dissipate their own heat (which is majority of the watt usage).
The controlling FPGAs don't use as much power, but if internal termination is used, then some of the current used by the DIMMs will actually heat up the FPGAs.
(Yes, it's yet another of those evil digital coin mining machines.)
Has to be modeled based on 24 cards. We use Micron datasheet for -667 part. 1GA (1066Ms) is not specified. Max operating temp is 85C (185F). Max rise if room is 100F = 85F.
Each card is 5.2sq" but total surface area is 10.4sq". Watts used for continuous read is 1480ma*1.8V=2.7W/(10.4/144)= 37.4W/sqft. Estimated rise above ambient is 128F. However, interleaved reads can be 2766ma/1416 = 1.95 (just say 200%).
Each PCB can use up to 30W (memory cards not counted). Most of that is from termination of reads. On writes it drives 44W, but that's absorbed in the dimms with unknown implications.
PCB Surface area is 0.79 sqft. Watts per sqft is 30W/.79=38 watts/sqft. PCB temp rise is 128F.
My temp rise estimate is from extrapolating the classic "sealed enclosure temperature rise", and assuming a PCB is a sealed enclosure of it's own dimensions. Same for the memory cards, I treated them like a sealed enclosure.
I didn't find any complex methods for evaluating this, in the first few pages of google.