SiSoftware Sandra inventories a computer and runs focused performance tests against its parts. An information module reads what Windows and the hardware report about a processor, memory controller, graphics device, drive, network interface, or installed component. A benchmark module sends a defined workload to one of those parts and records a measurement. The result is a synthetic score for that operation, not a prediction of frame rate or the speed of every real program.
Read the machine
Information modules separate reported capability from measured speed. SiSoftware Sandra can expose processor topology and instruction support, cache layout, memory configuration, graphics interfaces, drive details, sensors, and operating-system data. This makes a benchmark easier to interpret. A memory score has more meaning when the report also shows channel arrangement and clock information, while a storage result needs the exact device and controller path beside it.
The amount of detail creates its own friction. Similar names can describe different layers of the same subsystem: cache bandwidth is not main-memory bandwidth, and file-system throughput includes a path that a physical-disk test may bypass. Choosing every available module produces a long report but can hide the one measurement that matters. Starting with the suspected component keeps the output tied to a specific question.
Run one workload
Processor tests cover arithmetic, media-oriented operations, cryptography, and communication between cores. Memory modules measure bandwidth and access latency. Storage tests examine transfer behavior, while graphics and general-processing modules use supported compute interfaces to exercise a selected graphics device. SiSoftware Sandra also has combined indices, but those calculated scores mix several component measurements through a defined weighting.
A single run can change when Windows shifts power mode, a background scan starts, cooling reaches a limit, or another program consumes memory bandwidth. Close avoidable workloads, let the machine reach a stable temperature, and repeat the same module under the same settings. A faster result after changing both the power plan and a driver cannot identify which change caused the difference.
Compare like devices
The Official Benchmark Ranker gives SiSoftware Sandra reference measurements drawn from submitted results. The comparison is most useful when the selected reference uses the same device model and the same benchmark implementation. A number from another module may use different units or a different workload even when both labels mention the processor.
Benchmark revisions can also change scheduling, instruction paths, or scoring. SiSoftware Sandra warns that synthetic results should be compared within the same generation of the test. An older published number may look directly comparable while representing another code path. Saving the report with hardware details and test identity preserves the context needed for a later rerun.
Drivers shape results
Graphics compute tests depend on the driver and runtime that expose OpenCL or another supported interface. A missing runtime can remove a device from the candidate list, while a driver update can change compilation and scheduling without any hardware replacement. Selecting an integrated graphics processor instead of the discrete one also measures a different device.
Diagnostic data can point toward a configuration problem, but SiSoftware Sandra does not repair hardware. A low storage result may justify checking health data, cabling, driver mode, or free space. A low memory result may justify checking channel population and firmware settings. The benchmark supplies evidence for that investigation; it does not turn one score into a diagnosis by itself.
