Keep every GPU independently useful

Dense GPU workers for jobs that do not need one shared memory pool.

Rendering, image generation, transcription, embeddings and scheduled batch work can map well to separate GPUs. That makes independent-worker architecture valuable and explains where older mining platforms may still have a bounded role.

Decision in one minute

Use independent workers where jobs can be queued separately. Use a modern server platform for business-critical support, system memory, remote management and sustained operation.

Three-quarter supplier render of a 4U OEM multi-GPU rack server
Three-quarter supplier render of a 4U OEM multi-GPU rack server
OEM platform reference render. It is not evidence of a completed customer build or final specification. OEM supplier reference image. Written reuse permission pending.
OEM platform reference render. It is not evidence of a completed customer build or final specification.

Workload before hardware

Describe the queue, the quality bar and the operating owner.

01 / Work unit
Size from model, context, concurrent users and peak demand.
02 / Acceptance
Agree quality, latency, citation or output checks before buying.
03 / Operations
Plan updates, monitoring, support and a fallback route.
Diagram showing an approved request, a local service, an approved store and a policy-controlled data path
Diagram showing an approved request, a local service, an approved store and a policy-controlled data path
A private deployment starts with the permitted data path, access policy and logging boundary. Original explanatory plate. It sets out a decision method, not a measured result.
A private deployment starts with the permitted data path, access policy and logging boundary. Original explanatory plate. It sets out a decision method, not a measured result.

A queue is often more useful than tensor parallelism

Separate GPUs can process different frames, images, audio files or embedding batches with little inter-GPU communication.

This can improve utilisation and fault isolation without pretending the GPUs form one giant accelerator.

Demand shape

Peak demand can matter more than the daily average

Work unit
Tokens, frames, files or jobs.
Duration
How long one active job occupies capacity.
Concurrency
How many jobs overlap.
Deadline
Interactive response or queued completion.
A measured queue or user pattern is more useful than an unsupported user-count claim. Decision aid for: A queue is often more useful than tensor parallelism

Trace the work before sizing the capacity.

The queue, memory shape, data path and management route turn a broad workload name into a testable service.

Diagram of an evidence pack containing an asset schedule, burn-in record, health readings, workload test and admin guide
Diagram of an evidence pack containing an asset schedule, burn-in record, health readings, workload test and admin guide
A credible handover records the supplied assets, checks, operating evidence, instructions and unresolved items. Original explanatory plate. It sets out a decision method, not a measured result.
A credible handover records the supplied assets, checks, operating evidence, instructions and unresolved items. Original explanatory plate. It sets out a decision method, not a measured result.
GPU server remote management dashboard with system status, access logs and sensor monitoring panels
GPU server remote management dashboard with system status, access logs and sensor monitoring panels
Supplier screenshot of the platform management interface. The final management features and access policy depend on the ordered system. OEM supplier reference image. Written reuse permission pending.
Supplier screenshot of the platform management interface. The final management features and access policy depend on the ordered system. OEM supplier reference image. Written reuse permission pending.

Legacy chassis have a bounded second life

An old mining chassis may serve independent workers after exact CPU, RAM, storage, PCIe, thermal and electrical testing.

It should not be sold as a modern enterprise LLM server or used for new-GPU purchases without a prototype.

Acceptance bench

Quality and service conditions pass together

Output quality Correctness, support or usable result.
Service quality Latency, throughput and availability.
Control quality Permissions, logs and refusal.
Pass Named reviewer accepts the combined result.
A fast result is not acceptable when it is wrong, unsupported or shown to the wrong user. Decision aid for: Legacy chassis have a bounded second life

Power economics need a measured queue

Electricity is driven by loaded kW and hours, with cooling overhead. Measure wall power and completed useful work rather than estimating from GPU TDP alone.

The calculator uses 24.14p/kWh as a dated editable UK business input.

Operating loop

The workload continues after the first demonstration

  1. Observe Demand, errors and resource state.
  2. Review Quality drift, access and incidents.
  3. Change Versioned model or runtime update.
  4. Retest Focused acceptance before wider use.
The operating owner needs a repeatable route for change, rollback and evidence. Decision aid for: Power economics need a measured queue

Marketplace mode is off by default

Vast.ai, Render or Golem can be evaluated only for genuinely spare capacity, after security, warranty, insurance and workload-priority review.

No marketplace income appears in the base business case.

Questions answered

Straight answers to common questions

Can an old mining rig become a render server?

Potentially for independent workers after exact compatibility, stability, power and thermal testing. The host platform limits still matter.

Can spare GPUs earn money?

Possibly, but acceptance, demand, rate and availability are not guaranteed. Treat it as a measured opt-in pilot, not purchase justification.

Is rendering included in the AI server package?

The hardware can support agreed render or batch profiles, but application licences, queue integration and workload tests require scope.

Primary-source register

Check the live rule or price before relying on it.

Reviewed 26 July 2026. These links support the dated statements on this page; they do not replace legal, tax, security or professional advice.

Put the claim to work

Turn this guidance into a testable requirement.

The brief asks about workload and operating conditions - not just budget.