Skip to content

Platform feature · Core engine

Workload Optimization: faster, for less.

Run your existing program unchanged through the ORIQX compute engine—splitting work across CPUs, GPUs, TPUs, NPUs, and QPUs in parallel.

How it works

01

Submit unchanged

Keep existing applications, solvers, and scripts intact with zero manual rewrite.

02

Split compute blocks

Separate sequential logic, sparse math, dense kernels, and data movement into placeable blocks.

03

Place on fitting chips

Match each block to the CPU, GPU, TPU, NPU, or QPU that meets your goal.

04

Run & sharpen

Execute blocks in parallel, return unified results, and calibrate future predictions.

What you get

  • Same workload finishes faster because blocks run concurrently on fitting processors.
  • Lower compute bill because lightweight stages stop idling expensive accelerators.
  • Higher utilization across self-managed clusters and cloud accounts, or fully managed capacity.
  • Clean operation beneath existing queues and agent orchestrators.

When to use it & who it is for

Core capability for engineering, research, ML platform, data, and AI teams running mixed heavy compute in any cluster or cloud.

Engine scope

Supported chips

CPUs, GPUs, TPUs, NPUs, and QPUs (more classes added as needed)

Ways to buy

Fully managed or self-managed (data centre, HPC cluster, or your cloud accounts)

Code impact

Zero codebase rewrite; works beneath existing schedulers

01 · Decompose & Place

Any workload to the chip that fits

Splits monolithic programs into compute blocks and executes each part in parallel across CPUs, GPUs, TPUs, NPUs, and QPUs.

From any workload to any chip.

ANY WORKLOADYOUR INFRASTRUCTUREanyworkloadanychip

02 · Speed & Cost

Same machines, shorter run, lower spend

Cutting cross-node stalls and keeping every chip lane busy lowers elapsed wall time and total compute bill together.

Same workloads. Finished faster, for less.

SimulationC++OptimizationCAgentic AIRustAI trainingPythonAI inferenceJuliaData analyticsGoSAME CODEYOUR WORKLOADSCPUGPUTIMECOSTSTATUS QUOCPUGPUTIMECOSTWITH ORIQX

Optimization · Multi-stage pipeline

Example

Run coupled stages in parallel without rewriting pipeline code

Workload input

Pipeline: preprocessing, dense linear algebra, FFTs, and post-run aggregation.

ORIQX action

ORIQX maps every stage across parallel CPU and GPU lanes and eliminates idle gaps.

With ORIQX parallel placement
Selected
Elapsed
Active silicon
Serialized single-chip queue
Elapsed
Active silicon
Same codebase, higher hardware utilization.Workload Optimization