It’s no secret that the cloud delivers high availability, reliability, scalability, performance, and affordability that most companies struggle to match with on-prem chip design solutions. However, concerns around security, predictability, cost, and ease-of-use have impacted cloud adoption among IC design teams. To overcome systemic complexity and constraints of power at higher performance while ensuring quick time-to-market, the time has come to look at the cloud pragmatically as an enabler of progress — one that is both undemanding and flexible.
With today’s market conditions, innovative, cloud-oriented chip design solutions are needed to support the sky-high performance and computing demands from both hardware and software. But just how much IC hardware is “made in the cloud”?
Read on to learn about the top ways that the cloud can support an existing on-prem strategy, enable large and small enterprises to manage cost and capacity more effectively, and offer unparalleled security for valuable semiconductor IP.
For many organizations, on-prem development feels safe (like a lukewarm pool). Even though they recognize that it has limitations, moving to a new system such as the cloud looks like an unsettling jump. While this is perfectly understandable, tepid conditions are rarely conducive to innovation.
The onset of the SysMoore era means increasing systemic complexity and hyper-convergent design flows, calling for exponentially more compute and established electronic design automation (EDA) resources. In addition, as design tools and workflows become more automated and integrated with artificial intelligence (AI), design space exploration and verification will inevitably require increased and flexible access to these resources.
Going forward, restrictions and costs of a rigid corporate on-prem system will become much more apparent. Cloud service providers offer a way forward with the scaled high-performance computing (HPC) infrastructure, availability, capacity, and cost-effectiveness to manage these growing workloads comfortably and confidently.
Something to remember when considering cloud adoption for IC design is that it does not have to be a binary choice between an on-prem and a cloud environment. It isn’t a rejection of the tried-and-tested past for an unknown future.
The economics of shifting methodologies is not always tied to making the most of cost. Even if your organization is large, your teams can start small by using the cloud for set tasks to explore opportunities for productivity and innovation without going all-in. Taking a one-workflow-at-a-time approach, or even a one-tool-at-a-time approach, enables you to get a feel for the environment and its benefits as well as potentially freeing up capacity in a constrained on-prem estate.
The bottom line is that not all workloads need to move. With hybrid optionality, the cloud lets you determine the balance that will enable you to work in the most agile way.
On-prem teams typically worry that the potential for unlimited capacity in the cloud will cause costs to spiral. This is a legitimate fear and a risk that must be managed. Yet, on-prem deployments can become unwieldy fast: once you’ve paid for a certain capacity, you can’t simply dial it back, and maintaining capacity on standby is expensive.
With multiple engineering teams sharing on-prem resources, it falls on someone to anticipate demand and set capacity accordingly. Even with a wealth of historical data, forecasting has a high chance of failure. Getting it wrong risks either disrupting the product roadmap or overspending.
Smaller businesses need to manage their finances with extra care, even if their forecasting procedure may be less complex. Instead of channeling resources into building their own architecture, which includes hardware with a limited shelf-life, it makes more sense to adopt a pay-as-you-go—or grow—model that extends to IC design. This setup makes it possible to grow consumption to meet demand and flex capacities within budget constraints, paying only for what you use. The key is not losing sight of that budget and managing cloud capacity accordingly. In short, the beauty of the cloud is that all the costs associated with running a job can be accounted for on a single OPEX line—in other words, a CFO’s dream.
For a long time, the semiconductor industry has hesitated to put valuable IP into a domain that belongs to someone else. The guiding logic being that keeping propriety information in-house protects it. But as cybersecurity threats grow, this no longer holds true.
It’s worth remembering that most businesses already entrust other confidential assets such as legal, commercial, and operational information to the cloud via services such as Salesforce or Microsoft 365. The investment that cloud service providers have made in security dwarfs the level that any individual company is likely to be willing or able to make. So, if safeguarding your IP is a top priority, using the cloud for IC development offers one of the best options out there.
IC design flows are complex and moving them to the cloud from a legacy on-prem model does require data preparation, compute planning, and most likely some re-architecting of the workflows to make them cloud-ready. You will also need to decide where to run the interactive elements of the workflow, such as debug. While the process will be new, the benefits in terms of compute power, scalability, and parallelization more than justify the effort.
There are various options for establishing a working cloud model for IC design. A bring-your-own-license (BYOL) model lets you use your existing on-prem licenses with your chosen cloud provider, while a bring-your-own-cloud (BYOC) model caters to users who have already established cloud capabilities with a provider and whose capacity demands fluctuate.
Another option is a software-as-a-service (SaaS) environment, which absorbs the complexities of the cloud such as setup and running costs. Lastly, a hybrid approach enables you to meet peak demands by bursting capacity into the cloud while running fewer portable workloads on-prem. Choosing the best option depends on the size of your organization, the extent of on-prem investment, and the scale of your workflows.
IC designers currently live in a myriad of constraints—mostly of capacity and time. These limitations compromise the quality of the final product and business competitiveness in terms of time-to-market. At a personal and organizational level, time is a precious resource. Waiting days for lengthy batch compute jobs to finish is frustrating and inefficient, and something that cloud technology can help address by letting you decide how much compute resources and EDA software you want to throw at a problem to get the results you want at a time you need.
Used well, the cloud can help engineers raise productivity, performance, and accuracy and shorten turnaround time. Being free from capacity and availability constraints means you can focus on what you became an engineer to do: innovate.
Ultimately, you should look at the cloud as a platform not just to maintain, but to improve competitiveness by removing the barriers to innovation from limited compute and capacity. As IC design complexity inevitably rises, whether you are a small or large organization with some or no on-prem investment, the chances of you leveraging the cloud to develop IC hardware is not a matter of if but when.
To experience the new frontier of design and verification in the cloud, sign up for a free SaaS or BYOC evaluation of Synopsys Cloud at synopsys.com/cloud.
Catch up on other cloud-related blog posts: