SC109 — 32-Core CFD HPC Node with 256 MB Cache and 12-Channel Memory for ANSYS Fluent

The SC109 node is the top parallel tier in the lineup. It runs a single AMD EPYC 9355P — 32 cores and 64 threads at a 4.2 GHz all-core boost — with 256 MB of L3 cache and twelve DDR5 memory channels behind it. Paired with 128 GB of RAM and 2 × 1.88 TB NVMe SSD, it's the node to choose when your case scales well and you want the widest parallel spread we offer on a single socket, with the bandwidth and cache to actually use it.

The 256 MB cache figure deserves a note: it comes from AMD spreading 32 cores across eight chiplets — four cores per chiplet, each keeping its full 32 MB of L3. That's 8 MB of cache per core, an unusually high ratio, and each chiplet gets a wide interconnect link back to the memory controllers. For a memory-hungry workload like CFD, that layout matters as much as the core count.

Specifications

Three things work together here. Thirty-two cores give Fluent the widest parallel spread available on one socket. Twelve memory channels — roughly 600 GB/s per socket — keep those cores supplied rather than queuing for data. And 256 MB of L3, at 8 MB per core, gives each mesh partition substantial room to stay on-chip. The 4.2 GHz figure is the all-core boost, the clock held with every rank loaded, not a single-core peak.

What this node is built for

SC109 is the maximum-throughput tier: core count, bandwidth, and cache all at their highest. It suits:

• Cases that scale well across many cores — 32 MPI ranks split a mesh finer than any other node here, and with real bandwidth and cache behind them, that converts into shorter wall-clock time rather than stalled cores.

• Bandwidth-heavy physics — multiphase, combustion, many transported scalars, coupled solvers: cases that move a lot of data per iteration.

• Large parameter sweeps and design studies — more cores means more work per hour, so you get through more design iterations per day.

• Long sustained runs — an all-core clock held under full load matters more across a multi-day transient than a peak that only appears on one core.

• Cases with large per-partition working sets — 8 MB of L3 per core is a high ratio, so each partition keeps more of its active data on-chip.

• Transient, LES, and DES runs writing heavy data — NVMe storage keeps large data files and frequent autosaves from slowing down the solve.

Choosing SC109

SC109 leads on core count, cache, and bandwidth simultaneously — it doesn't trade one away for another. Choose it when:

• Your case scales efficiently past 24 cores, with enough cells per partition to keep each rank busy relative to the boundary data it exchanges, and

• Your solve is limited by throughput, bandwidth, or working-set size rather than by single-thread speed.

The caveat is scaling. Thirty-two ranks is a fine partition count, and a case that doesn't have the cell count to fill them will spend more time on inter-rank communication than on solving. If your case doesn't scale past a handful of cores, or is gated by serial sections and UDF execution, a high-clock 8-core tier will finish it sooner. And if memory capacity is your constraint rather than throughput, our high-memory nodes are the right tool. As a rough guide, Fluent needs about 1 GB of RAM per million cells for a typical steady, single-phase case, so with 128 GB, SC109 handles meshes well beyond what a 32 GB node can hold — and at 32 cores, a mesh in that range gives each rank plenty to work on.

Tell us your cell count, physics, and how your case scales with core count, and we'll point you to the right tier.

Fast NVMe storage

The two 1.88 TB NVMe SSDs give substantially higher read/write throughput than SATA SSDs, plus nearly 3.76 TB of combined capacity — which matters most on the highest-throughput node in the lineup, where a fast solve shouldn't be held up waiting on disk. That means faster loading of large case and data files, quicker autosaves, and room to keep a full DOE sweep's output on disk without constantly clearing space.

How parallel solving works on this node

When you run Fluent in parallel, the solver partitions your mesh across all 32 cores (using algorithms such as METIS), each core solving its share of cells and exchanging boundary data each iteration via MPI. On a single node, that communication runs through fast on-board channels rather than a network, keeping parallel overhead low. Between exchanges, each core pulls cell data from memory — and this is where a 32-core node either delivers or doesn't. With twelve memory channels, 8 MB of L3 per core, and a wide interconnect from each chiplet to the memory controllers, SC109's cores keep working rather than stalling as the partition count rises.

Why rent instead of buy

For engineers whose cases scale well and whose bottleneck is throughput rather than single-thread speed, SC109 delivers the cores, the bandwidth, and the cache together — and you can still drop to a high-clock tier when a case calls for it instead.

How it works — you're in full control

SC109 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 SC109 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 SC109 different from our 24-core nodes? A third more cores, twice the cache, and the same 12-channel bandwidth as our widest-bandwidth tier. If your case scales efficiently past 24 ranks, SC109 finishes it sooner; if it doesn't, the extra cores mostly sit idle and a 24-core tier is the better value.

Will 32 cores always be faster than 16 or 24? No. Fluent scales well only as long as each core has enough cells to stay busy relative to the data it exchanges with its neighbors. On a small or moderate case, splitting it 32 ways can leave cores spending more time communicating than computing — and a faster, lower-core node would finish first. This is the tier where right-sizing matters most, and we'll help you check it before you book.

Why is the cache so large for 32 cores? AMD spreads these 32 cores across eight chiplets — four cores each — while each chiplet keeps its full 32 MB of L3. The result is 8 MB per core, a much higher ratio than on denser parts, which benefits cases with large per-partition working sets.

Is 4.2 GHz slow compared to a 5.7 GHz node? On a single thread, yes. But 5.7 GHz figures are single-core peaks that come down under a full parallel load. SC109's 4.2 GHz is the all-core number, held with all 32 cores working — which is what a real parallel solve sees.

How large a mesh can it handle? With 128 GB of RAM, plan for meshes well into the tens to low hundreds of millions of cells for simpler steady single-phase cases, fewer for transient or multiphase physics (about 1 GB of RAM per million cells is a good starting estimate). Send us your case for a precise answer.

What if my case needs more memory than cores? For large, memory-bound simulations we offer higher-memory nodes — we'll recommend the right tier so you get the capacity, or the throughput, your case actually needs.

Is my data kept confidential? Yes.

Get started with SC109

If your cases scale well across many cores and you want the widest parallel spread, the largest cache, and the fullest memory bandwidth in the lineup, SC109 is built for exactly that.

Contact MR CFD to run your ANSYS Fluent case on SC109 — and we'll help you choose between throughput and clock speed for your simulation.