Docs

Understanding your scores

After running a benchmark, Novabench produces a set of scores that measure your system's performance. This page explains what the scores mean, how the components contribute to the overall rating, and what factors influence your results.

The Novabench Score

The Novabench Score is a single number that represents your system's overall performance. It combines individual scores from each hardware component tested during the benchmark. Higher scores indicate better performance.

Every Novabench Score is comparable across all users, platforms, and hardware configurations in the Novabench database. A score of 2000 on a Windows desktop means the same thing as a score of 2000 on a MacBook or Linux workstation. This cross-platform consistency is what makes the score useful for comparing different systems, whether you are evaluating a potential purchase or checking that your current machine is performing as expected.

The overall score is most useful as a quick summary. For deeper insight, look at the individual component scores to understand which parts of your system are contributing the most (or the least) to the total.

Component scores

The overall Novabench Score is built from individual component scores. Each component tests a specific piece of hardware:

Component

What it measures

CPU

Single-core and multi-core processing speed using integer and floating-point workloads

GPU

Graphics performance through 3D rendering, GPGPU compute, and VRAM throughput

Memory

RAM read and write bandwidth along with access latency

Storage

Sequential and random read/write speeds for your primary drive

NPU

Neural processing unit throughput (appears only when supported hardware is detected)

Each component score reflects the raw performance of that hardware. The component breakdown lets you identify which parts of your system are strong and which are holding overall performance back.

Note

The NPU score only appears on systems with a supported neural processing unit. If your hardware does not include an NPU, the overall score is calculated without it.

CPU score details

The CPU score has two parts: a multi-core score and a single-threaded score. The multi-core score measures how your processor handles parallel workloads spread across all cores. The single-threaded score measures performance on tasks that run on a single core, which is important for applications that do not scale across cores (for example, many games and older productivity software rely heavily on single-threaded performance).

Both integer and floating-point operations are tested, covering the types of computation most applications rely on. Integer operations handle logic, counting, and data manipulation, while floating-point operations handle decimal math used in scientific computing, media encoding, and 3D calculations. Together, these two workload types provide a well-rounded picture of CPU capability.

For a detailed breakdown of how the CPU test works, see CPU benchmark.

GPU score details

The GPU score combines results from 3D rendering, GPGPU compute, and VRAM throughput tests. If you have multiple GPUs, Novabench lets you select which one to test.

On Plus, the GPU results also include bottleneck analysis, which measures the balance between your CPU and GPU during 3D rendering. Bottleneck analysis identifies whether your CPU or GPU is the limiting factor, which can guide upgrade decisions. See GPU benchmark for details.

Memory score details

Memory scores reflect your RAM's throughput and latency. Throughput measures how quickly data moves between the CPU and memory, while latency measures how long the CPU waits for data to arrive. Both metrics matter for real-world performance.

Factors that influence memory scores:

  • Channel configuration: Dual-channel memory (two sticks) provides roughly double the bandwidth of single-channel (one stick). This is one of the most common reasons for a lower-than-expected memory score.
  • Memory speed and timings: Higher-rated memory (for example, DDR5-6000 versus DDR5-4800) delivers more bandwidth. Tighter timings reduce latency.
  • Platform limitations: Some laptop and budget desktop platforms cap memory bandwidth regardless of the installed RAM speed.

See memory benchmark for more details.

Storage score details

Storage scores measure your drive's read and write performance for both sequential and random access patterns. Sequential tests measure throughput for large file transfers, while random tests measure responsiveness for everyday tasks like loading applications and booting the operating system.

NVMe SSDs typically score significantly higher than SATA SSDs, which in turn outperform traditional hard drives. If your system has multiple drives, Novabench tests the primary drive by default. See storage benchmark for details on testing specific drives.

What affects your scores

Several factors influence benchmark results beyond the hardware itself. Understanding these factors helps you get the most accurate and repeatable results.

System environment

  • Background processes: Other running applications compete for CPU, GPU, and memory resources. A web browser with many tabs, antivirus scans, or cloud sync services can measurably reduce scores. Close unnecessary programs before benchmarking.
  • Power settings: Power-saving modes limit processor speed to extend battery life. On laptops, plug in the power adapter and set the power plan to Balanced or High Performance before benchmarking. On macOS, disable Low Power Mode in System Settings.
  • Startup services: Some background services (backup tools, update checkers, indexing services) run silently and consume resources. If scores seem inconsistent, check your task manager or activity monitor for unexpected CPU or memory usage.

Hardware conditions

  • Thermal throttling: If your system overheats during testing, the CPU or GPU reduces its clock speed to cool down. This lowers scores and makes results less repeatable. Ensure adequate airflow, and avoid running benchmarks on soft surfaces that block laptop vents.
  • Driver versions: Outdated GPU drivers can significantly reduce graphics performance. Keep GPU drivers up to date by installing the latest version directly from the vendor.
  • Disk state: Running benchmarks on a nearly full drive can reduce storage scores, because the drive has fewer contiguous blocks available for writes
  • RAM availability: If your system is using most of its RAM before the benchmark starts, memory may swap at considerable impact to performance.

Pro Tip

Run the benchmark multiple times (Novabench supports multi-iteration runs) and compare results. Consistent scores across iterations confirm that your system was running in a stable state. If scores vary significantly, check for background activity or thermal issues.

Reading the results screen

After a benchmark completes, the results screen shows several layers of information.

Score overview

At the top of the results screen, you see the overall Novabench Score along with the date and your system configuration. Below the overall score, each component (CPU, GPU, Memory, Storage, and NPU if applicable) is listed with its individual score.

Comparison data

The results screen includes comparison context for each component:

  • Score histograms and rankings showing where your score falls relative to other systems in the Novabench database
  • Median and average scores for the comparison group, so you can see the typical range

Health indicators

The results screen includes a health checklist for each component. The checklist uses pass or warning indicators to flag potential issues:

  • Score check: flags if your score deviates significantly from the median for identical hardware
  • Temperature check: flags if your system ran hotter than typical for the same hardware during the test
  • Power check: flags unusual power draw compared to similar systems

Sensor data

On Plus, the results screen also displays sensor data collected during the test, including temperature, power draw, and clock speed readings for each component over the duration of the benchmark.

Explain

Explain (Plus) analyzes your benchmark results in context and provides plain-language explanations with actionable recommendations. See Explain for more details.

Next steps