Our Services
Three Services, One Purpose: Measurements You Can Use
Each Substrate Co service is scoped to a specific point in the benchmarking process. Whether you are starting from scratch, running a comparison, or maintaining ongoing measurement practice, there is a clear entry point.
← Back to HomeMethodology
How We Approach Benchmarking
Every engagement starts with your workloads. We do not arrive with a fixed test suite; we build the benchmark plan around the tasks you actually run — specific model architectures, batch sizes, precision requirements, and concurrency patterns.
This approach takes longer than running a standard suite, but the results are meaningfully different. A figure produced from your workload, under your conditions, is a figure you can act on.
Scoping call
Understand workloads, evaluation timeline, hardware candidates
Method plan
Document benchmark procedure, assumptions, success criteria
Execution
Run benchmarks per the agreed method, record all conditions
Delivery
Results with method notes and stated caveats; walkthrough with your team
Service 01
Workload Benchmark Setup
A guided setup session where we help your team define fair, repeatable benchmarks for your own AI workloads. The output is a benchmark plan and reporting outline your team can execute and reference going forward.
This service is suited to engineers who are beginning an evaluation and want a structured method before running tests, or who have run informal tests and need a documented procedure that holds up to internal scrutiny.
What Is Included
- One structured session with your engineering team
- Written benchmark plan documenting workload definitions and assumptions
- Reporting outline for structuring future comparison results
- Follow-up questions addressed within five business days
Price
RM 620
Method Note
This service produces a plan, not results. Hardware access is not required during the session. The plan can be executed by your team using the agreed procedure, or it can feed into a Comparative Benchmark Study engagement.
Method Note
All configurations are tested using the same benchmark plan under identical conditions. Results are presented without declaring a preferred option; interpretation guidance notes where caveats apply.
Service 02
Comparative Benchmark Study
A study that runs your defined benchmarks across candidate hardware configurations and presents the results side by side. We avoid declaring a single winner and let the data inform the choice rather than the narrative.
Suited to teams weighing two or more hardware options — whether on-premise accelerators, cloud instance types, or a mix. Delivered across two phases with a written report, results dataset, and method note.
What Is Included
- Phase 1: dry run on one configuration with method validation
- Phase 2: full benchmark run across remaining candidates
- Written report with side-by-side comparison frame
- Results dataset in structured format for your own analysis
- Method note for every figure in the report
Price
RM 1,510
Service 03
Benchmarking Advisory Retainer
An ongoing arrangement to help your team maintain and refresh its benchmarking practice as workloads and hardware options evolve. We act as an independent advisor, keeping comparisons honest over time rather than letting procedures go stale between cycles.
Designed for organisations that evaluate hardware regularly — whether quarterly, at contract renewal, or in response to new product releases. Runs as a three-month engagement with scheduled reviews and a living method library.
What Is Included
- Three-month engagement with scheduled review sessions
- Living method library updated as workloads or options change
- Written review notes after each assessment cycle
- Independent advisory perspective on emerging hardware options
- Access to the Substrate Co team for ad-hoc questions within scope
Price (3 months)
RM 3,020
Method Note
The retainer is scoped at the start of the engagement. If evaluation needs change materially mid-term, scope adjustments are discussed openly before additional work begins. No charges are added without agreement.
Decision Guide
Choosing the Right Service
| Feature | Benchmark Setup |
Comparative Study |
Advisory Retainer |
|---|---|---|---|
| Benchmark plan / method doc | |||
| Benchmark execution by Substrate Co | — | ||
| Side-by-side comparison report | — | ||
| Results dataset | — | ||
| Living method library | — | — | |
| Ongoing scheduled reviews | — | — | |
| Price (RM) | 620 | 1,510 | 3,020 |
Standards
Protocols Applied Across All Services
Warm-Up Procedures
All benchmarks include a documented warm-up period before measurement begins. Cold-start figures are noted separately and not mixed with steady-state results.
Multiple Runs
Each measurement is taken across multiple runs. Statistical spread is reported alongside point estimates. Single-run figures are not used as final results.
Version Recording
Driver versions, framework versions, and OS configurations are recorded for every benchmark run. Results are anchored to a specific software environment.
Data Handling
Client data and workload artefacts are handled under confidentiality terms. Benchmark data is not shared with third parties or used to inform other client engagements.
Pricing
Clear Rates, Defined Scope
All prices in Malaysian Ringgit (RM). No hidden fees or variable billing.
Service 01
Workload Benchmark Setup
RM 620
One-time engagement
- One session
- Benchmark plan
- Reporting outline
Service 02
Comparative Benchmark Study
RM 1,510
Two-phase engagement
- Full benchmark execution
- Written report
- Results dataset
- Method notes
Service 03
Benchmarking Advisory Retainer
RM 3,020
Three-month engagement
- Ongoing advisory
- Living method library
- Scheduled reviews
- Review notes
Not Sure Which Service Fits?
A brief scoping call is the simplest way to work that out. Describe your workloads and timeline, and we will tell you which engagement makes sense.
Book a Scoping Call