Driven by demand and ever more efficient technological advancements, as the form factors of computing platforms get smaller and component densities get higher, system power consumption and thermal issues have become more challenging than ever before, especially in the case of laptops, slim desktops, and blade servers. Naturally, each system has a total cooling capacity per its specific design, and each major component has a cooling limit or power consumption allowance for balancing the system performance in that design. When the total system cooling capacity is smaller than the sum of each component’s Thermal Design Power (TDP), as seen in many small-form factor systems, all the components simply cannot simultaneously operate at their TDP level since the system cannot cool them. In such a system, it is also unacceptable to allow one component to free-run without power consumption and thermal restrictions as that would leave the performance of other components to suffer. Balanced cooling limits for major components should thus be achieved at the design level. Having briefly laid the background on cooling limits, we now focus on system memory (specifically in laptops), which is a major power-consuming component of platforms; specifically we discuss its cooling limits and enhancement in performance through better power/thermal management (memory bandwidth recovery using new throttling techniques).

To better understand the variation in memory cooling limits from system to system, a glimpse at three popular categories of laptops is helpful. The first of these categories is the Thin & Light (T&L) laptop, which is the mainstream, more conventional model. These laptops have a Z-height of around 1.1"-1.2" and a memory cooling limit of 4-5 watts. The second category is the Mini-Note PCs, which have a smaller form factor than the T&L and a Z-height of around 0.9". They have a smaller cooling fan and a memory cooling limit of 2-2.5 watts. Finally there is the Sub-Note PCs, which have a form factor even smaller than the Mini- Note PCs. These laptops do not have a cooling fan and their cooling limit is around 1 watt. These data clearly show that mobile platforms have limited cooling capabilities, and as the form factors continue to get smaller, the cooling budget also shrinks considerably. With increased memory speeds and capacities, we are now reaching the point where the memory thermals are starting to exceed the cooling capabilities of mobile systems. When the cooling budget is exceeded, that means the system is no longer able to cool the memory subsection, and the DRAM case temperatures begin to exceed their maximum case temperature specification of 85°C. Our lab data, taken on multiple notebook systems while analyzing multiple Small Outline-Dual In-line Memory Modules (SO-DIMMs)1, show that some 1 GB and greater capacity SO-DIMMs are exceeding their maximum specified case temperature when running a realistic workload at an ambient temperature of 35°C. Thus the memory bus needs to be throttled2 to ensure that the DRAM devices operate within their thermal limits, reducing the risk of memory corruption and system instability.

Platforms built on Intel® Centrino® Duo mobile technology implement two memory throttling techniques to address this issue.