Internal Fluid Flow
Hyperscaler vs. Corvid HPC
August 12th, 2026
Simcenter Star CCM+ v.19.02.009
Hyperscaler vs. Corvid HPC Results
-
We benchmarked a 40-million-cell case using Siemens STAR-CCM+ to compare Rescale’s cloud-based HPC platform against Corvid HPC’s bare-metal platform. Because large simulations require fast, constant communication between processors, typical cloud environments often drag down performance with hidden layers of virtualization. Our architecture fixes this by removing those virtual roadblocks entirely, replacing them with dedicated AMD EPYC Genoa processors and direct, high-speed network connections. By reducing the infrastructure and virtualization overhead that can limit performance, we enabled the simulation to scale efficiently across 384 bare-metal cores. This reduced simulation time by more than 3x compared with the Rescale baseline while also lowering the cost of each run. Engineers can use these gains to complete the same workload for less or run more design iterations using the STAR-CCM+ software they already have.
-
Getting rid of virtual roadblocks lets your software run at its true speed. By moving the exact same STAR-CCM+ simulation to Corvid HPC’s dedicated hardware—powered by AMD EPYC Genoa processors—the engineering team completely removed the hidden delays found in standard cloud setups.
When we ran the simulation across 384 bare-metal cores, we cut the runtime down from 3.94 hours to just 0.97 hours. That is a 75.38% reduction in time, allowing the simulation to run over 3x faster than the standard cloud baseline.
The biggest win for the business is that getting these massive speed improvements doesn't punish your budget. The team finished these simulations in under an hour for just $25.32 per job, compared to the $67.47 spent for a single, much slower run in the cloud. That means engineering teams can choose how to apply the performance gains based on what matters most to their business today.
What is more valuable to your business today: Reducing cloud compute costs by 62%, or using your existing software footprint to run more simulations and develop better designs?
-
Moving sensitive aerospace and defense projects to a remote environment means data security has to be flawless. We make this transition easy by providing an ITAR-compliant platform right out of the box. While standard cloud providers often host your data on shared servers, Corvid HPC locks your workloads down on your own isolated, dedicated hardware. This ensures you meet strict federal rules, like DFARS and NIST SP 800-171, and keeps your environment completely CMMC ready without the usual compliance headaches.
To keep things user-friendly, our platform lets you access your software visually right through a secure web browser. Your team can set up, run, and review massive simulations entirely within our protected environment. This completely eliminates the security risk of downloading huge, sensitive files back to local workstations. With Corvid HPC, you get the massive computing speed you need while keeping your data completely secure and compliant.
| Platform | Core Count | Node Architecture | Wall-Clock Time | % Time Reduction | Case Cost (USD)* |
|---|---|---|---|---|---|
| Hyperscaler (Rescale) | 120 Cores |
Standard Virtualized | 3.94 hours |
Baseline | $67.47 |
| Corvid HPC OD3 | 96 Cores |
AMD EPYC Genoa | 3.80 hours |
3.55% | $27.36 |
| Corvid HPC OD3 | 192 Cores |
AMD EPYC Genoa | 1.88 hours |
52.28% | $27.07 |
| Corvid HPC OD3 | 384 Cores |
AMD EPYC Genoa | 0.97 hours |
75.38% | $25.32 |
Simcenter STAR-CCM+
Performance & Cost Comparison
-
To understand the real-world impact of these cost efficiencies, let's consider an engineering budget of $24,000 dedicated to this specific STAR-CCM+ workload.
On the Hyperscaler (Rescale): At $67.47 per run, a $24,000 budget allows for approximately 355 simulation cases.
On Corvid HPC (Weekday Standard Rate): At $25.32 per run, the exact same $24,000 budget allows for approximately 947 simulation cases.
Comparing the real costs for this effort may look like the following where the $24k in compute costs is amplifying the $230k existing investment in an engineer and software:
$55k for software + $175k FTE Engineer + $24k compute = $254k
That’s ~$716 per run Hyperscaler vs $268 per run for Corvid HPC
By eliminating the cloud tax, engineering teams don't just run jobs 3x faster—they gain the ability to run nearly three times as many design iterations on the exact same budget, significantly expanding their capacity for design exploration and validation.