Abstract-This paper proposes accurate architectural-level interconnect wirelength and fan-out models. Existing models are based on Rent's rule and fail to capture the impact of microarchitectural and implementation parameters. Hence, significant deviation is observed when validated against implementation data, i.e., up to 79% (22%) in total wirelength (average fanout). Our proposed models both enable architectural-level prediction of interconnect wirelength and fanout, and show significant accuracy improvement vs. existing models with respect to layout data.I. INTRODUCTION Clock frequency, power consumption, and chip size are largely determined by the wiring requirements of a VLSI system. Hence, wirelength estimation has always been instrumental in determining physically achievable design implementations [4]. Likewise, interconnect fanout can be used in interconnect modeling to enable more accurate delay and power calculations.Based on different design stages and available information at each stage, we categorize wirelength estimation approaches into: (1) analytical, (2) netlist-based, and (3) placement-based [1], [6]. Analytical methods are based on Rent's rule, a famous wellestablished empirical relationship, and assume that the netlist is not known. However, the input may consist of the major factors, the total number of gates, and average number of pins per gate. In netlistbased methods, the total number of gates, total number of nets, total number of pins, and fanout distributions are easily obtained. Since we know exact values of major parameters, wirelength and fanout values can be predicted more accurately. Such estimates are typified by the "wireload models" used in RTL synthesis optimizations. After the placement stage, all the pins in the netlist are placed and the available routing resources, including number of routing layers, etc. are known and better estimation accuracy is achieved.In this work, we propose interconnect wirelength and fanout models that have the flexibility of the analytical models with the accuracy of placement-based models. Our contributions are as follows.• We propose accurate closed-form models to estimate interconnect wirelength and fanout. Our models are derived from implementation data.• We consider microarchitectural and implementation parameters in our models to allow more accurate estimation of interconnect wirelength and fanout.• Our models show significant accuracy improvement vs. the existing models with respect to implementation data. The remainder of this paper is organized as follows. In Section II we review two of the recent works in interconnect wirelength and fanout estimations [3], [9]. Section III describes our implementation flow and testcases, and presents the details of our design of experiments. Section IV describes our modeling methodology and presents our proposed average wirelength and fanout models for one of our testcases. In Section V we validate our proposed models and compare them with the existing models. Finally, Section VI concludes the paper.