SC90 — Single-Socket 32-Core High-Frequency CFD HPC Node with 256 GB Memory and 2 TB NVMe Storage for ANSYS Fluent

The SC90 node is our high-frequency, single-socket powerhouse. By combining 32 highly clocked cores with a massive memory pool, it shifts the focus from sheer core volume to raw per-core performance. A single AMD EPYC 9355 processor delivers 32 cores and 64 threads with an aggressive 4.4 GHz boost clock. Memory is remarkably generous at 256 GB, and storage is provided by two 1 TB NVMe SSDs — giving you roughly 2 TB of high-speed usable capacity.

Because this is a single-socket configuration, all 12 memory channels wire directly into one processor. There is no cross-socket latency and zero NUMA overhead. For CFD cases burdened by complex physics, moving meshes, or serial bottlenecks, SC90’s combination of high clock speed and huge memory-per-core ratio (8 GB per core) is incredibly potent.

Specifications

Compared to our high-core-count or dual-socket nodes, SC90 offers fewer total cores but clocks them significantly higher. Per core, it provides an exceptional 8 GB of RAM. This is one of the highest ratios in our lineup, making this node ideal for memory-heavy physics models rather than just purely massive mesh cell counts.

What this node is built for

SC90 is the high-frequency and single-thread-performance tier. It suits:

• Compute-bound and high-physics cases — Combustion, heavy multiphase, discrete phase models (DPM), and reacting flows require far more math (and memory) per cell. High clock speeds chew through this overhead much faster than simply throwing more slow cores at it.

• Tasks with serial bottlenecks — Meshing in Fluent Meshing or running heavy User Defined Functions (UDFs) that don't scale perfectly across dozens of ranks benefit heavily from a 4.4 GHz boost clock.

• Models that don't scale past 32 cores — Some smaller or highly complex cases lose parallel efficiency quickly. SC90 gives you 32 incredibly fast lanes instead of 64 slower ones.

• Memory-heavy meshes up to 256 GB — If you have a 100+ million cell mesh but want to avoid the complexity of dual-socket NUMA domains, this node holds it all on one piece of silicon.

The single-socket advantage

A single-socket machine like the SC90 offers a distinct advantage in structural simplicity. Because there is only one processor, there is only one memory domain. Every core has equal, direct access to the entire 256 GB of RAM across 12 blazing-fast DDR5 channels.

For Fluent, this means no cross-socket latency penalties (NUMA overhead) when a core needs data residing on the other side of the motherboard. While we expertly configure dual-socket nodes to minimize that traffic, SC90 eliminates the variable entirely. It is brute-force, high-speed compute in its purest form.

Choosing SC90

Choose SC90 when:

• Your solve is gated by per-core clock speed or complex physics calculations, or

• You are performing meshing tasks that benefit from fast single-thread performance, or

• Your case requires a massive 8 GB of RAM per core to run without memory exhaustion, or

• You prefer the zero-overhead simplicity of a single-socket architecture.

If your case scales beautifully to 64 cores and you need peak memory bandwidth (24 channels), a dual-socket node like the SC89 will solve faster. If you need 22 TB of storage to hold months of data, the SC99 is required. But for pure per-core muscle and massive memory depth without NUMA complexity, SC90 is the sharpest tool in the lineup.

Storage: high-speed NVMe

Two 1 TB NVMe SSDs provide around 2 TB of incredibly fast storage. While not the massive bulk capacity of our SATA tiers, NVMe drives are built for extreme I/O throughput. When saving large data files or executing transient autosaves, the NVMe drives allow Fluent to write gigabytes of data in seconds. Your solver gets right back to calculating physics rather than hanging on disk operations.

How parallel solving works on this node

When you run Fluent in parallel, the solver partitions your mesh across all 32 cores. Because they all live on the same silicon, MPI communication between ranks happens at the maximum internal bus speed. The single processor's 12 memory channels feed all 32 cores directly from local memory. This frictionless environment is exactly why single-socket, high-clock nodes are universally praised for reliable CFD scaling.

Why rent instead of buy

How it works — you're in full control

SC90 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 SC90 node.

2. Set up your simulation — load your geometry and case, configure your solver settings, and run ANSYS Fluent with all 32 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 SC90 different from the SC89 and SC99 nodes?

SC90 is a single-socket node with 32 cores, whereas SC89 and SC99 are dual-socket nodes (64 and 48 cores, respectively). SC90 counters its lower core count with a much higher clock speed (4.4 GHz) and zero cross-socket latency, making it the best choice for compute-bound physics and serial tasks.

Is 2 TB of storage enough?

For most standard steady-state simulations, meshing, and short transient runs, 2 TB is plenty. If you are doing long parameter sweeps or thousands of transient saves that need to be kept on the drive natively, you may need a higher-capacity storage tier.

Will I miss the second processor?

Not if your case is better suited to high frequency over massive parallelism. Many complex reacting flow or DPM cases scale poorly past 30-40 cores anyway, making 32 exceptionally fast cores much more effective than 64 slower ones.

How large a mesh can it handle?

With 256 GB of RAM, you can comfortably hold meshes into the hundreds of millions of cells. However, with an 8 GB per core ratio, this node really shines on models where cell count is moderate (e.g., 20 to 50 million) but the physics per cell are incredibly demanding.

Is my data kept confidential?

Yes.

Get started with SC90

If your simulations are bogged down by complex physics, serial bottlenecks, or if you simply need the high-frequency punch of a 4.4 GHz processor backed by a massive memory pool, SC90 is built to accelerate your workflow.

Contact MR CFD to run your ANSYS Fluent case on SC90 — and we'll help you leverage peak per-core performance for your simulations.