- Nvidia reported up to 1.5x higher performance for selected Cadence Jasper formal-verification and Synopsys VCS simulation workloads. [1] - The tests used equal core counts, but Nvidia did not name the baseline CPU or disclose software versions, workload configurations, run times, power consumption or independent validation. [1] - Vera systems have reached external organizations, but Nvidia’s disclosed EDA deployment is internal; neither Cadence nor Synopsys announced general Vera support or a hardware-roadmap change. [2][3][4]
Nvidia is deploying its 88-core Vera CPU across the electronic design automation workflows used to build its next CPUs and GPUs, claiming up to 1.5x higher performance on selected production-class verification and simulation jobs from Cadence and Synopsys.[1]
The company identified two applications: Cadence Jasper, used for formal verification, and Synopsys VCS, used for functional logic simulation. Nvidia said the comparisons used the same number of CPU cores, but its announcement omitted the baseline processor, system configurations, software versions, individual test results and whether “performance” measured elapsed time or throughput. It provided no power or efficiency figures.[1]
An internal EDA deployment with limited benchmark detail #
CPU performance remains important in EDA because logic simulation and formal verification include latency-sensitive and branch-heavy work that cannot always be shifted efficiently to GPUs. Faster individual runs can shorten debugging cycles, while greater aggregate throughput lets engineering teams execute more regressions and test more design alternatives before tapeout.[1][5]
Nvidia said Vera is being deployed “throughout” its EDA workflows and published an image of a cluster in its Portland data center, but it did not disclose the number of CPUs, servers or engineering teams involved. The company also did not say when customers will be able to run Jasper or VCS on commercially supported Vera systems.[1]
Vera itself is not confined to Nvidia. The company previously delivered early systems to Anthropic, OpenAI, SpaceXAI and Oracle Cloud Infrastructure. That external activity does not establish that Vera-optimized EDA software is generally available.[2]
No disclosed change to vendor hardware roadmaps #
Cadence’s DAC materials confirm a broader Nvidia collaboration across CPUs, GPUs and CUDA-X software, while Synopsys separately detailed its Nvidia-based autonomous engineering and GPU acceleration work. Neither vendor’s announcement supplied the Vera test methodology, confirmed general availability for Jasper or VCS on Vera, or announced a change to its supported-platform roadmap.[3][4]
The work is an extension of existing Arm-server efforts rather than a new architectural commitment for Synopsys: in 2025, the company said it was enabling more than 15 products on Nvidia’s Grace CPU architecture, including functional verification workloads.[6]
Companies mentioned #
Further sources #
[[1] Nvidia, “NVIDIA Harnesses Vera CPU to Speed Up Design of Next-Generation CPUs a… ↗](https://blogs.nvidia.com/blog/vera-cpu-eda/)
[[2] Nvidia, “Vera Arrives: NVIDIA’s First CPU Built for Agents Lands at Top AI Labs… ↗](https://blogs.nvidia.com/blog/vera-cpu-delivery/)
[3] Cadence, “Cadence’s Agentic AI Design Vision Takes Center Stage at DAC,” July 2… ↗
[4] Synopsys, “Synopsys Showcases Comprehensive Autonomous Engineering Workflows fr… ↗
[5] Nvidia Developer Blog, “NVIDIA Vera CPU: Olympus Cores Built for Maximum Single… ↗
[6] Synopsys, “Synopsys Accelerates Chip Design with NVIDIA Grace Blackwell and AI … ↗
The stories that matter, in one email. Free — unsubscribe anytime.