SC91 — Dual-Socket 192-Core CFD HPC Node with 384 GB Memory and 22.5 TB NVMe Storage for ANSYS Fluent

The SC91 node represents the pinnacle of extreme parallel compute and uncompromising I/O performance in our lineup. Driven by two AMD EPYC 9655 processors, it deploys an unprecedented 192 physical cores and 384 threads, backed by a blistering 4.5 GHz boost clock. Memory stands at an expansive 384 GB, and the storage array completely rewrites the rules: six 3.75 TB NVMe SSDs, delivering roughly 22.5 TB of ultra-high-speed usable capacity.

This dual-socket configuration brings 24 total memory channels and immense compute density. By combining the massive core count of flagship silicon, the huge storage bulk usually reserved for slower SATA arrays, and the peak throughput of pure NVMe, SC91 is engineered for simulations that demand absolute maximum parallel scaling and generate colossal amounts of transient data.

Specifications

Compared to our other dual-socket nodes, SC91 triples the core count of the SC89 and quadruples that of the SC99, while maintaining a staggering 4.5 GHz clock speed. With 384 GB of RAM across 192 cores, it provides exactly 2 GB of memory per core. This is a highly focused ratio designed for slicing moderately massive meshes (up to ~300 million cells) into incredibly small, fast-solving partitions to achieve record turnaround times.

What this node is built for

SC91 is the ultimate high-parallelism and uncompromised-storage tier. It suits:

• Extreme parallel scaling — Cases that scale efficiently to 192 ranks will see dramatic reductions in solve times, turning days of computation into hours.

• Massive transient datasets — With 22.5 TB of pure NVMe storage, you can dump thousands of massive time-step files at lightning speed without stalling the solver or running out of disk space.

• Rapid design iteration — When time-to-solution is the absolute most critical metric, 192 cores running at 4.5 GHz chew through steady-state cases faster than anything else in the fleet.

• Large-scale DOEs — The ability to run massive parametric sweeps rapidly and store the entirety of the resulting data natively on one machine.

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 SC91

Choose SC91 when:

• Your primary goal is minimizing turnaround time on meshes that scale well up to 192 cores, or

• You have a heavy transient workflow that requires both the extreme speed of NVMe and the massive 22+ TB capacity usually restricted to SATA drives, or

• You need the compute density of nearly 200 cores alongside a very high 4.5 GHz boost clock.

If your case requires massive memory-per-core (e.g., heavily complex physics that need 6–8 GB per rank), you may hit a memory bottleneck here and should look to the SC89 or SC90. SC91’s advantage is sheer parallel compute volume paired with the most powerful storage array available.

Storage: capacity and peak speed combined

Historically, CFD engineers had to choose: the massive capacity of SATA (SC99) or the sheer speed of NVMe (SC89). SC91 eliminates the compromise. Six 3.75 TB NVMe SSDs provide 22.5 TB of space with extraordinary sequential write speeds. For large transient acoustics (CAA), sliding mesh, or combustion runs where the solver generates gigabytes of data every few iterations, this array allows Fluent to write autosaves instantly, virtually eliminating I/O bottlenecks and letting all 192 cores stay focused on the math.

How parallel solving works on this node

When you run Fluent in parallel, the solver partitions your mesh across 192 cores using algorithms like METIS. Each core receives a highly optimized, smaller chunk of the total mesh. On this dual-socket node, 24 total memory channels ensure those 192 cores are fed continuously. Sensible partitioning and rank placement keep cross-socket traffic to a minimum, ensuring that the unprecedented core count translates directly into faster iterations.

Why rent instead of buy

How it works — you're in full control

SC91 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 SC91 node.

2. Set up your simulation — load your geometry and case, configure your solver settings, and run ANSYS Fluent with all 192 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 SC91 different from the SC89 and SC99 nodes? SC91 is our most aggressive compute and storage node. It offers three times the core count of SC89 (192 vs 64) and combines the massive 22 TB storage capacity of SC99 with the elite NVMe speed of SC89. It is built for absolute maximum performance on highly parallel workflows.

Is 2 GB of RAM per core enough? Yes, for the right workflows. If your mesh is 150 million cells, running it on 192 cores means each core only handles about 780,000 cells—well within the 2 GB per core limit. It is designed to aggressively slice meshes for maximum speed. For cases with extreme physics overhead that demand higher memory-per-core, nodes like SC90 (8 GB/core) remain the better choice.

What is the benefit of a 6-drive NVMe array? Zero I/O compromise. You get the capacity to hold months of full-scale transient data, coupled with the PCIe bandwidth to write that data to disk in seconds instead of minutes.

How large a mesh can it handle? The node is bounded by its 384 GB of RAM, meaning you can comfortably hold meshes into the 250–300+ million cell range for standard steady-state runs (assuming ~1 GB per million cells).

Is my data kept confidential? Yes.

Get started with SC91

If your cases demand the absolute limits of parallel scaling, extreme clock speeds, and an I/O array that refuses to bottleneck your transient saves, SC91 is the ultimate environment for your simulation.

Contact MR CFD to run your ANSYS Fluent case on SC91 — and we'll help you unleash 192 cores on your most demanding workflows.