Cloud Hosting

In the past decade, the move towards cloud computing occurred primarily in sectors like finance, retail, and healthcare, with the emergence of leading public cloud vendors such as Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform, and others accelerating the trend. However, the chip design industry has been slower to adopt cloud computing.

In the current highly competitive environment where the cost of semiconductor design and manufacturing has soared, innovative cloud-enabled solutions are what the semiconductor industry needs to address quality requirements, time-to-market schedules, and high-cost challenges. Electronic design automation (EDA) solutions are essential components of semiconductor design and verification and play a significant role in the industrys move to the cloud. We will take a closer look at how using the public cloud for semiconductor design and verification is helping drive semiconductor innovation.

Fig. 1: Cloud computing technology is helping accelerate silicon innovation.

Chip design complexity has dramatically increased, driven by the explosion of big data and semiconductor industry market pressures from expanding applications including high-performance computing, Internet of Things (IoT), automotive, artificial intelligence (AI), and 5G mobile communications. The design complexity is compounded further by using advanced technology nodes for high-performance, low power and area IPs, the building block of SoCs (System on a Chip).

Thorough verification of these IPs is imperative. However, it requires significant compute capacity and resources (figure 2) to handle the corresponding heavy workload and achieve robust IP design through the various IP design and verification stages. As a result, on-premises compute capacity has undeniably become a bottleneck for chip design teams to design and verify their IP designs and bring their SoC to market faster.Cloud computing has become a viable option for the design teams to use for designing and verifying modern SoCs Siemens EDA has demonstrated and provided ahighly scalable cloud computing capacity for AMS verification workloads using Microsoft Azure cloud.

What are the key market drivers and benefits influencing chip design teams, and how can cloud-based EDA solutions help drive innovation?

Fig. 2: IP design and verification workflow.

With chips becoming more complex and larger, the amount of compute resources required is significant to accommodate the compute-intensive workloads needed for thorough verification of every stage of the IP design and verification flow. Then there is the burst computer resource demand at the design tape-out stage. In addition, concurrent IP development projects within a company compete for the same compute resources, causing further delays. As a result, compute resources are becoming a significant bottleneck, impacting time-to-market targets. Furthermore, on-premises data centers cannot quickly adapt to varying design and verification workloads. In contrast, cloud computing can help increase throughput with access to significantly more and flexible compute resources. Also, the ability to quickly scale up and or down elasticity based on the verification needs makes the cloud even more attractive for chip designers.

Like highly parallelizable verification tasks, specific workloads are more suitable for cloud computing, providing significant throughput improvement. For example, consider an advanced node library characterization flow, a highly parallelizable task. Built-in machine learning enabled technologies may significantly reduce runtimes required for library characterization. However, a significant amount of SPICE simulation is needed to characterize each library revision. Today, characterizing an entire standard cell library requires hundreds of millions to billions of SPICE simulations to generate the timing.lib files used by synthesis, place & route, and signoff timing analysis tools. Because of this, library characterization tasks take up to several days to weeks to complete. Siemens EDA and AWS have collaborated to providea scalable, secure and cost-effective cloud-based library characterization flowthat enables users to leverage cloud computing resources to accelerate library characterization.

To meet specifications and adapt to the challenges of smaller, more advanced technology nodes without compromising the quality of results power, performance, and area (PPA) and yield, designers need to perform extensive verification to account for all potential design variations. This includes performing extensive process, voltage, and temperature (PVT) and Monte Carlo SPICE simulation at a granular level. These are highly compute-intensive workloads demanding significantly more compute resources than on-premises data centers can accommodate. Hence, design teams often must make difficult trade-offs by adding design margins to meet time-to-market (TTM). Siemens EDA enables customers to migrate to the cloudand meet and exceed quality of results with higher chip yieldwithin production run times.

Another equally compelling benefit to moving to the cloud is overall lower cost. Companies like a startup may choose to go all on the cloud. On the other hand, enterprise semiconductor companies may use a hybrid cloud model to augment their on-premises data center. With cloud service vendors providing flexible pricing models like pay-as-you-go pricing, design teams can scale up and down the compute resources on a need basis at each stage of the IP design development cycle. Cloud computing resources help reduce capital expenditures and lower the cost required year-round to upgrade and maintain a data center that is typically architected and sized for peak usage, such as sign-off verification.

In the past decade, the move of EDA to the cloud has been hindered by concerns about the security of semiconductor intellectual property (IP) and data, which at the time outweighed the significant benefits gained by computing in the cloud. But chip design complexity continues to drive the need for massively parallel computing. This huge catalyst has accelerated the development of advanced cloud security and fault-tolerant solutions by cloud service providers. Siemens EDA is working with cloud service providers and foundries, ensuring the solutions support the cloud security protocols and infrastructure. Consequently, leading foundries have adopted the cloud and acknowledged the security of the cloud infrastructure by having their IP and process design kits (PDKs) in the cloud helping diminish the security concerns.

Siemens EDA is partnered with cloud service vendors to innovate and provide semiconductor companies a seamless transition to the cloud by delivering optimized cloud-ready AMS verification solutions and best practices. The cloud-ready solutions provide scalability, flexibility, and security, enabling the chip design team to meet their PPA metric and TTM goals. In addition, proven best practices equip design teams with cloud reference architectures to choose from, including the network configurations, type of virtual machines, storage, and data transfer mechanisms to guide them on their cloud journey.

The semiconductor industry is under relentless market pressure to produce innovative, high-performing chips and meet TTM targets. In tandem, innovation of cloud-ready and cloud-optimized EDA solutions will be crucial to support the high demand for compute resources to reduce IP design development cycle times.

The rest is here:
EDA In The Cloud Is Driving Semiconductor Innovation - SemiEngineering