Despite the DDR2's lower voltage at 1.8 volts - which never gets exceeded on a laptop - the mobile machine's space constraints can result in quite a heap of heat gathering around the DIMMs.
On one Intel booth, the demo showed how, at full speed, a 1 GB DDR2-667 SO DIMM in a "thin and light" notebook reached an awfully hot 85 degrees Celsius case temperature, which is even hotter than the FB-DIMMs I burned my finger on a few booths away from there.
Of course, those FB-DIMMs were darn hot despite their huge heat spreaders, which the notebook SO-DIMM lacked.
Such heat in a notebook, if not properly dissipated, could cause harm - not exluding a potential of hot melted plastic dripping onto the user's crotch, if the casing quality is a bit inferior. So, there are two ways out: one is to create a decent airflow over the two SO-DIMMs in the system, which is not easy in small, thin systems, or implement an equivalent of CPU thermal throttling - which Intel perfected with the microwave oven called Prescott - on the memory module.
Assuming the vendors are too lazy for the first option, Intel has helped on the second one - it designed a Thermal Sensor with a closed-loop throttling mechanism on the DIMM which, in the case a certain temperature limit is breached, throttles the DIMM speed down, say 50 per cent for instance in this case - from 667 MH to 333 MHz effective throughput.
Interestingly, on, say, Yonah or Merom-based 667 MHz FSB systems, a dual-channel DDR2-333 DIMM set will already provide the full FSB bandwidth in sync, so the throttle down shouldn't really impact the performance at all - unless, of course, the CPU and the 945GM integrated graphics share the RAM bandwidth. Also, how about using the same to measure and control the heatr on the notebook casing 'skin'? Well yes, Intel included that option, too. µ