Synopsys and Ansys power the future of innovation—connecting silicon to systems.
EDA
System
IP
SolvNetPlus
Training & Education
Faster verification for increasingly complex designs
Verification is one of the most compute-intensive and schedule-critical stages of modern chip development. As AI accelerators, advanced CPUs, GPUs, chiplets, and complex systems-on-chip increase verification demand, engineering teams need to complete regression testing sooner, identify issues earlier, and use compute capacity more efficiently.
New Synopsys testing shows that Synopsys VCS® workloads on AWS Graviton5-powered Amazon EC2 M9g instances can achieve up to 43% faster end-to-end turnaround time than on Graviton4-powered M8g instances. Across 11 representative workloads, Graviton5 improved compilation and simulation performance in every evaluated case, with gains of up to 61% and 49%, respectively. The strongest end-to-end workload completed 43% faster.
| Synopsys VCS Results on AWS Graviton5 vs. Graviton4 | |||
| Up to 43% | Up to 61% | Up to 49% | 500+ |
| faster turnaround time | faster compilation | faster simulation | Qualified test cases |
Gains like this can shorten individual verification jobs and increase regression throughput, giving engineering teams more opportunities to find issues earlier while providing IT organizations with another qualified architecture for heterogeneous HPC environments.
A decade of deep engineering collaboration
Amazon describes Graviton5 as its most powerful and efficient CPU to date, delivering up to 25% better compute performance than the previous generation. The processor includes 192 cores, a five-times-larger L3 cache, DDR5-8800 memory, and up to 33% lower inter-core communication latency.
Synopsys and AWS have collaborated since the inception of Annapurna Labs to enable AWS custom silicon development. Synopsys EDA solutions help AWS design and verify Graviton, Nitro, and Trainium chips. In turn, each new generation of AWS infrastructure gives the broader semiconductor ecosystem additional cloud compute options for demanding engineering workloads.
Proven readiness for production-representative verification
Performance gains matter only when functional correctness is maintained. Before conducting any performance benchmarking, Synopsys first qualified VCS on AWS Graviton5 using more than 500 production-representative functional tests. The validation achieved a 100% pass rate within the defined qualification scope and produced equivalent quality of results across the evaluated Graviton5 and Graviton4 environments.
The suite covered Verilog, System Verilog, VHDL, mixed-HDL, UVM, OVM/VMM, assertions, functional coverage, UPF, DPI, PLI, interactive debug, and parallel compile and simulation configurations. It included IP-level through large SoC-scale workloads, providing confidence for customers to evaluate AArch64-based CPU platforms for VCS verification.
| Functional Qualification Scope | Performance Scope | Comparison Control |
| 500+ test cases | 11 representative workloads | Matched 96-vCPU instances |
| 100% pass rate | CPU, GPU, AI, MCU, SoC, GLS, DFT, UPF | Same VCS build, inputs, scripts, and settings |
With functional qualification established, Synopsys then evaluated performance using a controlled set of representative verification workloads to understand the impact of the Graviton5 architecture on compile, simulation, and overall turnaround time.
Putting Graviton5 to the test
Synopsys benchmarked 11 representative workloads on Amazon EC2 m9g.24xlarge instances powered by Graviton5 and m8g.24xlarge instances powered by Graviton4. Both configurations used 96 vCPUs. The same Synopsys VCS 2026.03-SP1-1 build, design inputs, scripts, and run settings were maintained, making the processor generation the primary hardware variable.
| Metric | Observed improvement | What it measures |
| Turnaround time | Up to 43% faster | Compile time plus simulation runtime |
| Compilation | 17% to 61% faster | Elapsed wall-clock compile time |
| Simulation | 6% to 49% faster | Elapsed wall-clock simulation runtime |
| Simulation peak memory | Effectively unchanged | No change on 10 of 11 workloads; up to ~2% lower on one |
| Compile peak memory | ~3% lower geometric mean | Workload-dependent range from ~25% lower to ~16% higher |
The results varied by workload, as expected. The strongest end-to-end result was 43% faster turnaround time. The findings also complement the public AWS announcement in which joint partner Arm reported up to 40% faster Synopsys VCS runtimes on Graviton5 relative to previous generations, providing an additional ecosystem proof point.
From faster jobs to greater engineering throughput
Verification is highly iterative. Engineers modify a design, compile it, run simulations, analyze results, resolve issues, and repeat the cycle throughout a semiconductor program. Faster jobs do not translate one-for-one into a shorter product schedule, but they can increase regression throughput and provide feedback sooner. Here’s what these results mean for customers:
For engineering users
- Complete compile and simulation jobs sooner, reducing wait time during iterative debug cycles.
- Run more regressions within existing windows and reinvest time savings in additional coverage.
- Adopt an AArch64-based option that demonstrates equivalent results across the tested qualification scope.
For IT and HPC leaders
- Add a qualified architecture to a heterogeneous compute strategy for VCS workloads.
- Use elastic Amazon EC2 capacity for nightly regressions, milestone testing, and pre-tape-out peaks.
- Plan simulation capacity with effectively unchanged peak-memory behavior across the tested suite.
- Evaluate infrastructure efficiency using workload-specific performance, pricing, licensing, storage, and operational requirements. The benchmark demonstrates performance, not a universal cost-savings result.
For engineering executives
- Increase verification throughput and help teams surface design issues earlier.
- Reduce adoption risk through broad functional qualification and controlled generational testing.
- Expand infrastructure choice as verification demand grows across AI and advanced semiconductor programs.
Actual results may vary based on design, workload, software version, instance configuration, storage, and operating environment.
Optimizing EDA engineering software and infrastructure together
The results demonstrate the value of co-optimizing EDA software and cloud infrastructure. Graviton5 contributes advances in processor architecture, cache, memory performance, and inter-core communication. Synopsys contributes decades of expertise in simulation and verification technology for complex semiconductor designs. Together, Synopsys VCS on AWS Graviton5 gives engineering teams greater infrastructure flexibility while accelerating cloud-based verification with qualified functional equivalence, higher workload throughput, and up to 43% faster verification turnaround times.
Ready to see the results on your own workloads? Contact us to evaluate VCS on AWS Graviton5 and find the configuration that best fits your verification environment.
Continue Reading
