SC92 — Dual-Socket 128-Core CFD HPC Node with 256 GB Memory and 22.5 TB Storage for ANSYS Fluent

The SC92 node strikes a powerful balance between high-density parallel compute and massive storage capacity. Powered by two AMD EPYC 9655 processors, this dual-socket configuration delivers 128 physical cores and 256 threads with a fast 4.2 GHz boost clock. Memory sits at a robust 256 GB, while the storage array mirrors our highest-capacity tiers: six 3.75 TB SATA SSDs, providing roughly 22.5 TB of usable space.

Two sockets also mean two sets of memory controllers: 24 memory channels in total, roughly 1.2 TB/s of aggregate bandwidth, and massive L3 cache across the pair. By pairing 128 high-speed cores with bulk SATA storage, SC92 is heavily optimized for aggressive parallel scaling on large parametric sweeps and extended transient campaigns that generate massive amounts of data.

Specifications

Compared to our 48-core and 64-core nodes, SC92 drastically scales up the compute density. With 256 GB of memory spread across 128 cores, it provides exactly 2 GB of RAM per core. This ratio is designed for slicing moderately large meshes (up to ~200 million cells) into smaller, highly efficient partitions to drive down solve times, while retaining the massive disk space required to hold the results.

What this node is built for

SC92 is the high-compute and bulk-capacity tier. It suits:

• Broad parallel scaling — 128 cores provide the computational width to drastically cut turnaround times on meshes that scale efficiently past 64 ranks.

• Long transient campaigns with massive output — 22.5 TB of storage easily accommodates months of time-step data, acoustic saves, or full DOE sweeps without constant file housekeeping.

• Bandwidth-bound solves — 24 memory channels across two sockets feed the 128 cores with the widest memory path available, preventing data starvation.

• Aggressive partitioning — Breaking a 150-million cell mesh across 128 cores gives each rank roughly 1.1 million cells, an optimal sweet spot for peak solver efficiency.

Understanding the dual-socket layout

A two-socket machine is not simply a bigger one-socket machine. Each processor has its own memory controllers and its own local RAM, and a core reaching data attached to the other socket pays a latency penalty crossing the interconnect between them.

For Fluent, this is highly manageable and worth knowing about. Partitioning the mesh so each socket works mostly on locally-held data — and pinning MPI ranks accordingly — keeps that cross-socket traffic low. Run without any NUMA awareness and you can lose some of the bandwidth advantage the second socket provides. We configure this for you as part of the setup, so in practice it's a detail we handle rather than one you have to manage.

Choosing SC92

Choose SC92 when:

• Your case scales efficiently to 128 cores, and you want to aggressively reduce solve times, or

• You are generating tens of terabytes of result data and need a massive storage footprint, or

• Your solve is bandwidth-bound and benefits from the throughput of a 24-channel dual-socket architecture.

If your workflow demands instant write speeds for massive autosaves, an NVMe-equipped node like SC89 or SC91 is a better fit. If your mesh requires more than 256 GB of RAM, you should look toward our 384 GB tiers. SC92’s specific advantage is providing over a hundred cores of compute power directly alongside enterprise-scale storage capacity.

Storage: capacity over peak speed

Six 3.75 TB SATA SSDs present a different trade-off from our NVMe tiers. While per-drive throughput is lower than NVMe, the total capacity is vast, and six drives working in parallel handle CFD write patterns highly effectively. For long transient runs writing thousands of time steps or a large DOE sweep where every case's complete result set needs keeping, raw capacity is usually the constraint that bites first. SC92 is sized to ensure you never have to pause a run to clear disk space.

How parallel solving works on this node

When you run Fluent in parallel, the solver partitions your mesh across all 128 cores (using algorithms such as METIS), each core solving its share of cells and exchanging boundary data every iteration via MPI. On a dual-socket node, most of that communication stays on-board, while each socket's 12 memory channels feed its own 64 cores from local memory. Sensible partitioning and rank placement keep cross-socket traffic to a minimum, ensuring your heavy parallel investment actually translates to shorter solve times.

Why rent instead of buy

How it works — you're in full control

SC92 is yours to drive directly. You connect to the node over a secure remote desktop connection and run everything yourself, exactly as you would on a local workstation:

1. Connect via remote desktop — you get full remote access to the SC92 node.

2. Set up your simulation — load your geometry and case, configure your solver settings, and run ANSYS Fluent with all 128 cores at your command.

3. Run and monitor in real time — start, pause, adjust, and watch the solution as it progresses, with complete control throughout.

4. Keep your results — post-process on the node or download your solution and data files whenever you're ready.

You have full authority over the node for the duration of your rental: we provide the hardware and the support, and you run the simulations your way.

Frequently asked questions

How is SC92 different from SC99?

Both feature 256 GB of RAM and 22.5 TB of bulk SATA storage, but SC92 offers massively upgraded compute. It jumps from 48 cores to 128 cores, drastically improving solve times on well-scaled meshes.

Is 2 GB of RAM per core enough?

Yes, for appropriately sized meshes. If you run a 150-million cell mesh on 128 cores, each core is only responsible for just over 1 million cells. The 2 GB per core ratio is highly optimized for this kind of rapid, partitioned solving. If your case has extreme physics overhead requiring higher RAM per core, look at our 384 GB nodes or single-socket setups.

Is SATA storage a bottleneck?

For steady-state cases or transient runs with infrequent saves, no. The 6-drive array distributes write loads efficiently. However, if you are doing highly frequent saves on massive datasets (e.g., sliding mesh acoustics every few time steps), an NVMe tier will prevent the solver from pausing during disk writes.

How large a mesh can it handle?

Bounded by the 256 GB of RAM, you can comfortably load meshes up to roughly 200–250 million cells for standard single-phase steady-state runs. Send us your case for a precise estimate.

Is my data kept confidential?

Yes.

Get started with SC92

If your cases have scaled beyond the limits of standard 48-core or 64-core workstations, and you need a massive storage footprint to warehouse the results, SC92 delivers the parallel width and disk space to get the job done.

Contact MR CFD to run your ANSYS Fluent case on SC92 — and we'll help you configure the ideal environment for your large-scale simulations.