SC101 — 192-Core CFD HPC Node for High-Volume ANSYS Fluent Batch Work

The SC101 node carries the highest core count available: a single AMD EPYC 9965 with 192 cores and 384 threads, 384 MB of L3 cache, and twelve DDR5 memory channels. Paired with 128 GB of RAM and 2 × 1.88 TB NVMe SSD, it is a batch machine — built to work through a large volume of simulations, not to accelerate any single one.

That distinction is the whole story of this node, so it's stated up front rather than buried below.

Specifications

Two specification notes matter for CFD. This chip uses AMD's Zen 5c cores — a density-optimized variant that fits more cores per socket at lower clock speeds. Its 3.35 GHz all-core boost is the lowest in our lineup, roughly 20% below the 4.1–4.2 GHz of our 48- to 128-core tiers, and well under half the 5.7 GHz peak of the high-clock nodes. The trade is deliberate: SC101 buys core count with clock speed.

Sizing note: this is a volume node

SC101 divides 128 GB of RAM and roughly 576 GB/s of memory bandwidth across 192 cores. That works out to under 700 MB and about 3 GB/s per core — the tightest ratios in the lineup by a clear margin.

For a single Fluent case, SC101 is not the right node. A mesh sized to fit in 128 GB, split 192 ways, leaves each rank with too few cells to stay efficient; inter-rank communication would dominate the solve, and the low per-core clock compounds it. On a single case, most of our smaller tiers — including the 8-core high-clock nodes — will finish sooner. That isn't a caveat, it's what the hardware is.

What SC101 is for is volume. Twelve cases on 16 cores each. Six on 32. A 40-run DOE sweep processed in batches. A whole team's workload running concurrently without anyone queuing. Measured in cases completed per day rather than iterations per second, no other node here comes close.

What this node is built for

• Large DOE sweeps and design-of-experiment studies — dozens of variants, run in concurrent batches rather than sequentially.

• Mesh-independence and sensitivity studies — many related cases at once instead of one after another.

• Shared team infrastructure — several engineers each running their own moderate cases simultaneously, each with their own core allocation.

• Deadline batch work — turning a month of sequential runs into a few days of parallel ones.

• Heavy concurrent I/O — NVMe storage keeps up with many jobs writing data files and autosaves at the same time.

Choosing SC101

Choose SC101 when you have many cases, each of moderate size, and total throughput is what matters.

Do not choose it for a single case, however large. Fluent scales well only while each rank has enough cells to stay busy relative to the boundary data it exchanges with neighbours, and 192 ranks is far past that threshold for any mesh fitting in 128 GB. Combined with the lower Zen 5c clock, a single case will almost always run faster on a smaller tier — our 24-, 32-, or 64-core nodes for cases that scale, or an 8-core high-clock node for cases that don't.

Tell us how many cases you're running and how large each one is. If SC101 isn't the right answer, we'll say so and point you at the node that is.

Fast NVMe storage

The two 1.88 TB NVMe SSDs give far higher read/write throughput than SATA SSDs, plus nearly 3.76 TB of combined capacity. On a node running a dozen jobs at once, that's not a convenience — several cases writing data files and autosaves simultaneously would bottleneck on slower storage, and the capacity holds a full sweep's output without constant clearing.

How parallel solving works on this node

When you run Fluent in parallel, the solver partitions your mesh across whichever cores you assign it (using algorithms such as METIS), each core solving its share of cells and exchanging boundary data every iteration via MPI. On SC101 the sensible pattern is a modest core allocation per case — 16 or 32 — with many cases running side by side. Each case then has enough cells per rank to scale properly, while the node as a whole stays fully loaded. Spreading one case across all 192 cores is possible but rarely productive, for the reasons above.

Why rent instead of buy

Batch work arrives in bursts. Renting a 192-core node for the weeks when a sweep needs to run, then dropping back to a smaller tier, matches that pattern far better than owning a machine that sits underused between projects.

How it works — you're in full control

SC101 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 SC101 node.

2. Set up your simulations — load your geometry and cases, configure solver settings, and allocate cores across as many concurrent runs as you need.

3. Run and monitor in real time — start, pause, adjust, and watch your solutions as they progress, 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

Will 192 cores make my case solve faster? For a single case, almost certainly not. This node has the lowest per-core clock in the lineup and the least memory and bandwidth per core. Its value is running many cases at once. For one case, ask us — a smaller tier will very likely be faster.

Why is the clock lower than other nodes? This chip uses AMD's Zen 5c cores, a density-optimized design that fits 192 cores in one socket by running them at lower frequencies. The all-core boost is 3.35 GHz, against 4.1–4.2 GHz on our 48- to 128-core nodes.

How does it compare to the 128-core node? Half again as many cores at a noticeably lower clock, with the same memory and bandwidth spread thinner. SC101 completes more concurrent jobs; the 128-core node is faster on any individual one. Choose by how your work is shaped, not by core count alone.

Is 128 GB enough for 192 cores? For many moderate concurrent cases, yes — that's the intended use. For anything large, memory binds long before core count does, and our high-memory nodes are the right tier.

Can I really run a dozen cases at once? Yes. Allocate cores per case and run them side by side. That's the workflow this node is built around, and we'll help you set it up.

Is my data kept confidential? Yes.

Get started with SC101

If you have a large batch of simulations to get through — a sweep, a study, or a team's full workload — SC101 completes more of them per day than anything else in the lineup.

Contact MR CFD with your case list, and we'll confirm whether SC101 fits your workload or point you to a faster tier for what you're actually running.