Building a Workstation with RigSync

A gaming PC has one job. A workstation has as many jobs as there are workloads — and the right build for training a language model looks nothing like the right build for a CAD seat or a render node. That's what makes a workstation hard to spec by hand.

You don't need to know which GPU wins for your workload, how much VRAM your model needs, or whether your platform has the lanes for four cards. You pick what you're building and your budget; RigSync optimizes a complete build around it, then computes that it actually works — compatibility, power, physical fit, and PCIe lane topology. And it surfaces only the options that matter for the workload you chose.

Where to start

The workstation workflow lives in the Custom Builder. Choose a Primary Use Case, and if it's a professional workload, the builder opens a set of options tailored to it. Everything irrelevant stays out of the way — a CAD build doesn't ask about model precision, and an AI build doesn't ask about ISV certification.

Six workstation use cases are available today. Each one starts from the same two choices — the workload and your budget — and adds only the questions that change its build.

AI / ML

For: training or running language and vision models.

You set the number of GPUs, your model size, and its precision — and the builder shows the estimated VRAM the model needs as you go. From there it filters to GPUs that can actually hold your model, pooling memory across multiple cards if you asked for more than one, and ranks what's left by real throughput. The budget goes to the GPUs, with a CPU sized to feed them. If the model can't fit in budget, it says so plainly and tells you what would — a higher budget, more cards, or a lower precision.

3D Rendering

For: GPU render engines and CPU renderers.

You set the number of GPUs, a GPU VRAM target, and memory capacity. The build is GPU-led, because the render engine is the workhorse. The VRAM target sets a floor on the card, and memory capacity scales for large scenes and assets held in RAM.

Video Editing

For: 4K/8K timelines, effects, and export.

You set the number of GPUs, a GPU VRAM target, and memory capacity. It's built around the accelerated playback and effects pipeline, with enough VRAM for your timeline and enough RAM for cached effects and multi-stream projects.

CAD

For: modeling, large assemblies, and simulation.

You set the number of GPUs, memory capacity, and whether to require an ISV-certified GPU. The build pairs a strong CPU for single-threaded modeling with a capable viewport GPU. Turn on ISV-certified and the GPU is restricted to professional cards validated for pro CAD software drivers. Memory scales with assembly and simulation size.

Scientific Computing

For: simulation, HPC, and double-precision math.

You set the number of GPUs, a GPU VRAM target, and memory capacity. This one is CPU-led, leaning on core count and double-precision throughput, with a compute GPU alongside. Large in-memory datasets pull memory up as the budget allows.

Virtualization

For: running many VMs on one host.

You choose whether to include a discrete GPU and set memory capacity. CPU cores and RAM lead here — VMs live in memory, so it scales RAM up hard. Leave the GPU off for a headless host, or include one for passthrough.

The options, explained

A few of these are workstation-specific. Quick reference:

• Number of GPUs — 1 to 4. Multi-GPU unlocks the higher budget tiers and moves you to a platform that can physically host that many cards.
• Model size / precision (AI/ML) — together these set how much VRAM your model needs. A lower precision fits a bigger model on the same card.
• GPU VRAM target — a floor on the card's memory, for scenes, datasets, or timelines that have to fit.
• Memory capacity — the target for system RAM. Leave it on Auto to let the budget-scaled default decide, or set it explicitly.
• ISV-certified GPU (CAD) — restrict the GPU to professional cards certified for pro software drivers.

Everything from the standard Custom Builder is still here: budget, performance goal, CPU and GPU brand, memory type, case size, and noise tolerance.

One practical tip: the Performance Goal slider shifts more of your budget into the core components — GPU, CPU, cooling — as you raise it. On a multi-GPU build, that can be the difference between mid-tier and top-tier cards. If a build lands under your budget, try nudging the goal up.

Multiple cards, and a platform that can hold them

Ask for more than one GPU and the build becomes as much about the platform as the cards. RigSync moves you to a workstation-class CPU and board with the PCIe lanes and physical slots for that many cards, then computes what the arrangement actually does: which slot each card lands in, what bandwidth it gets, and which slots a wide card blocks or a second card disables — down to the per-CPU-family lane behavior, because different CPU generations route their lanes differently.

You can see all of it in the interactive topology visualizer — a live map of every slot on your board that updates as you build. It's a tool, not a diagram: choose where a card goes and the build recomputes around you, or hit Optimize to hand the layout back to the auto-assigner. For a multi-GPU workstation, this is where "four cards fit the budget" becomes "four cards fit the board."

What it handles for you

Behind every workstation build, RigSync automatically:

• Scores each GPU for your workload — gaming, AI, rendering, video, CAD, and scientific are separate answers, not one gaming number
• Sizes the whole machine to the job — memory, platform, and CPU, not just the GPU
• Validates compatibility and topology — sockets, power, clearance, PCIe lanes, and physical slot blocking
• Flags honest tradeoffs — if a model won't fit or a part is a mismatch for the workload, it tells you, with what to do about it

Think of the estimates as realistic starting points. Your exact throughput depends on drivers, framework, and configuration — but the build gives you a sound foundation for the work.

What you control

• Primary use case (AI/ML, 3D Rendering, Video Editing, CAD, Scientific Computing, Virtualization)
• Budget range and performance goal
• Number of GPUs (1–4)
• Workload-specific options (model size/precision, VRAM target, memory capacity, ISV certification, discrete GPU)
• CPU and GPU brand, memory type, case size, noise tolerance

Ready to build one?

Start with the workload, set your budget, and let it put together the complete machine — then open the topology panel to see exactly how it fits.

Build a workstation with the Custom Builder →