The main contribution of this dissertation is the demonstration of the impact of a new emerging technology called 3D-integration technology on conventional high-performance microprocessors. 3D-integration technology stacks active devices in the vertical dimension in addition to the conventional horizontal dimension. The additional degree of connectivity in the vertical dimension enables circuit designers to replace long horizontal wires with short vertical interconnects, thus reducing delay, power consumption, and area. To adapt planar microarchitectures to 3D-integrated designs, we study several building blocks that together comprise a substantial portion of a processor s total transistor count. In particular, we focus our attention on three basic circuit classes: static random access memory (SRAM) circuits, associative/CAM logic circuits, and data processing in conventional high-performance processors. We propose 2-die-stacked and 4-die-stacked 3D-integrated circuits to deal with the constraints of the conventional planar technology. We propose high-performance 3D-integrated microprocessors and evaluate the impact on performance, power, and temperature. We demonstrate two different approaches to improve performance: clock speed (3D-integrated processors with identical microarchitectural configurations as the corresponding planar processor run at a higher clock frequency), and IPC (3D-integrated processors accommodate larger-sized modules than the planar processors for the same frequency). We demonstrate the simultaneous benefits of the 3D-integration and highlight the power density and thermal issues related to the 3D-integration technology. Next, we propose microarchitectural techniques based on significance partitioning and data-width locality to effectively address the challenges of power density and temperature. We demonstrate that our microarchitecture-level techniques can effectively control the power density issues in the 3D-integrated processors. The 3D-integrated processors provide a significant performance benefit over the planar processors while simultaneously reducing the total power. The simultaneous benefits in multiple objectives make 3D-integration a highly desirable technology for use in building future microprocessors. One of the key contributions of this dissertation is the temperature analysis that shows that the worst-case temperatures on the 3D-integrated processors can be effectively controlled using microarchitecture level techniques. The 3D-integration technology may extend the applicability of Moore s law for a few more technology generations.

Ce travail a montré que l’apport de la technologie d’intégration 3D, permet de surmonter les limites imposées par la technologie monolithique sur les performances électriques (« coupling » et consommation) et sur l’implémentation physique (aire du pixel) des imageurs. Grâce à l’analyse approfondie sur la technologie d’intégration 3D, nous avons pu voir que les technologies d’intégration 3D les plus adaptées pour l’intégration des circuits dans le pixel sont : 3D wafer level et 3D construction séquentielle. La technologie choisie pour cette étude, est la technologie d'intégration 3D wafer level. Cela nous a permis de connecter 2 wafers par thermocompression et d’avoir une interconnexion par pixel entre wafers. L’étude de l’architecture CAN dans le pixel a montré qu’il existe deux limites dans le pixel : l’espace de construction et le couplage entre la partie analogique et numérique « digital coupling ». Son implémentation dans la technologie 3D autorise l’augmentation de 100% l’aire de construction et la réduction du « digital coupling » de 70%. Il a été implémenté un outil de calcul des éléments parasites des structures 3D. L’étude des imageurs rapides, a permis d’étendre l’utilisation de cette technologie. L’imageur rapide type « burst » a été étudié principalement. Cet imageur permet de dissocier la partie d’acquisition des images de la sortie. La limite principale, dans la technologie monolithique, est la taille des colonnes (pixels vers mémoires). Pour une haute cadence d’acquisition des images, il faut une grande consommation de courant. Son implémentation dans la technologie 3D a autorisé à mettre les mémoires au-dessous des pixels. Les études effectuées pour ce changement (réduction de la colonne à une interconnexion entre wafers), ont réduit la consommation totale de 90% et augmenté le temps d’acquisition des images de 184%, en comparaison à son pair monolithique. / This work has shown that the contribution of 3D integration technology allows to overcome the limitations imposed by monolithic technology on the electrical performances (coupling and consumption) and on the physical implementation (area of the pixel) of imagers. An in-depth analysis of the 3D integration technology has shown that the most suitable 3D integration technologies for the integration of the circuits at the pixel level are: 3D wafer level and 3D sequential construction. The technology chosen for this study is the 3D wafer level integration technology. This allows us to connect 2 wafers by thermocompression bonding and to have an interconnection or “bonding point” par pixel between wafers. The study of the architecture CAN at the pixel level showed that there are two limits in the pixel: the construction area and the coupling between the analog and digital part «digital coupling». Its implementation in 3D technology allows the construction area to be increased by 100% and the digital coupling reduced by 70%. It has been implemented a tool for computing the parasitic elements of 3D structures. The study of high speed imagers has allowed the use of this technology to be extended. The "burst" imager was mainly studied. This kind of imager’s architecture can dissociate the image acquisition from the output part. The main limit, in monolithic technology, is the size of the columns (pixels to memories). For a high rate of image acquisition, a high current consumption is required. Its implementation in 3D technology allowed to put the memories below the pixels. The studies carried out for this change (reduction of the column to an interconnection between wafers) reduced the total consumption by 90% and increased the acquisition time of the images by 184%, compared to its monolithic peer.