A one-year-old server running simulation and verification will not be able to compete with the current generation of compute developed this year. This phenomenon, coupled with the fact that businesses may need to “guess” their required memory, accelerators, and other hardware months in advance for on-premises hardware, makes the cloud a more appealing option.
During the design cycle, each task may require different compute capabilities, storage access, and memory. As the environment is built and progresses, additional computer resources may be needed. Instant availability of expanded memory capacity for timing analysis is only possible through the cloud.
For static timing analysis and physical verification tasks, access to multicores with high memory-per-core machines and throughput is advantageous. With the AI features EDA vendors have, multiple place and route runs could be performed for PPA optimization; however, this can only be accomplished through parallel high-performance computers. Furthermore, the specific servers used for this task differ from the servers best-suited for verification runs. Having HPC for specific tasks allows for enhanced optimization.
Ultimately, access to more compute resources allows project design and verification processes to be completed faster. The cloud essentially removes compute limitations and optimizes performance.