Skip to content
DuoBolt

Performance Benchmarks: DuoBolt vs Czkawka, dupeGuru & Gemini 2

DuoBolt delivers exceptional performance in real-world duplicate file detection scenarios. These benchmarks compare DuoBolt against three leading alternatives: Czkawka, dupeGuru, and Gemini 2.

All tests were performed on Apple Silicon M1 Pro hardware running macOS Tahoe 26.1 with identical configurations, cleared caches, and back-to-back execution to ensure fair comparisons.


Up to 21× faster

Against dupeGuru on local files. Against Czkawka the gap is 1.0× to 2.3× — the tables below give the figure for each dataset, including the one where it ties.

1.2–1.4× on NAS

On network-attached storage with warmed caches, DuoBolt maintains a 1.2–1.4× performance advantage over competitors.

Handles 1+ TB datasets

DuoBolt successfully scans terabyte-scale datasets while maintaining speed and accuracy, even when competitors time out.


Every number on this page is ours. That is worth saying plainly, and it is worth pointing at the one assessment of DuoBolt’s speed that is not.

Softpedia’s Catalin Chelariu tested DuoBolt 2.0.2 in August 2026 and rated it 4.0 out of 5. On the scan itself:

It’s very fast, surprisingly so.

The same review calls DuoBolt “quite expensive for a duplicate cleaner.” We link it whole rather than quoting only the half that flatters us — read it at Softpedia.

Note what that review did not test: no NAS, no ~/Library, no RAW, no APFS clone accounting. It is a verdict on ordinary folders on a Mac, which is the workload the tables below are least dramatic about. The gap it confirms is the one we measured at 1s versus 5.5s, not the one at 64 seconds versus a timeout.


The following table shows measured scan times (in seconds) across five real-world datasets. Lower times are better. Timeout indicates the tool exceeded 15 minutes without completing. Failed indicates permission errors or crashes.

Tool Time Speedup vs DuoBolt
DuoBolt 1s
Czkawka 1s 1.0×
dupeGuru 11s 11.0×
Gemini 2 5.5s 5.5×

Dataset: 13.17 GB of Sketch and Figma export files Result: DuoBolt and Czkawka tie on small, uniform datasets. DuoBolt pulls ahead significantly as complexity increases.


Tool Time Speedup vs DuoBolt
DuoBolt 3s
Czkawka 4s 1.3×
dupeGuru 63s 21.0×
Gemini 2 17s 5.7×

Dataset: 37.92 GB across three code directories (21.13 GB, 13.61 GB, and 3.18 GB web apps) Result: DuoBolt’s multi-core parallelism shines on diverse file structures, completing in 3 seconds while dupeGuru takes over a minute.


Tool Time Speedup vs DuoBolt
DuoBolt 24s
Czkawka 55s 2.3×
dupeGuru Timeout
Gemini 2 Failed

Dataset: Codebase plus macOS user Library (~/Library) directory Notes:

  • dupeGuru timed out before completion
  • Gemini 2 lacked permission for ~/Library

Result: DuoBolt handles complex permission scenarios and large file counts where competitors fail or time out.


Tool Time Speedup vs DuoBolt
DuoBolt 64s
Czkawka 85s 1.3×
dupeGuru 142s 2.2×
Gemini 2 Timeout

Dataset: 1.01 TB Synology NAS (DS920+) snapshot with cold, spun-down disks Breakdown: 605 GB video, 246 GB photos, 162 GB music Result: DuoBolt’s streaming I/O and BLAKE3 parallelism maintain speed even on slow, cold network drives.


Tool Time Speedup vs DuoBolt
DuoBolt 20s
Czkawka 24s 1.2×
dupeGuru 59s 3.0×
Gemini 2 Timeout

Dataset: 1.01 TB Synology NAS snapshot with warmed cache Breakdown: 605 GB video, 246 GB photos, 162 GB music Result: With caches warm, DuoBolt completes in 20 seconds while dupeGuru still requires nearly a minute.


All benchmarks follow strict testing protocols to ensure fair, reproducible results:

  • DuoBolt version: 1.4. These figures predate DuoBolt 2.0; a re-run on the current version is pending, and until it lands these numbers stand as the slower, older ones rather than as our best.
  • Hardware: Apple Silicon M1 Pro processor
  • OS: macOS 26.1 (Tahoe)
  • NAS: Synology DS920+ with 3 HDD drives installed
  • Network: Gigabit Ethernet connection
  • All caches cleared between runs
  • Identical include/exclude rules applied across all tools
  • Back-to-back execution on the same machine
  • No other intensive processes running during tests
  • Tools configured for maximum performance (default settings where applicable)
  • Gemini 2 ran in duplicates-only mode (similarity detection disabled for fair comparison)
  • All tools used their default hash algorithms
  • dupeGuru and Gemini 2 timed out on NAS datasets (15-minute timeout limit)
  • Gemini 2 encountered permission errors on ~/Library directory
  • Scan times measured from start to completion
  • Times rounded to nearest second for readability
  • Timeout threshold: 15 minutes (900 seconds)
  • Only successful, complete scans included in timing comparisons

DuoBolt’s performance advantage comes from several architectural decisions:

DuoBolt keeps all CPU cores busy with per-root parallelism and BLAKE3 worker hashing, instead of processing files sequentially. This approach scales efficiently across modern multi-core processors.

Chunked, streaming reads overlap disk/NAS I/O with hashing operations, hiding latency on slower disks or remote drives. This design ensures CPUs never sit idle waiting for data.

BLAKE3 is a modern, cryptographically secure hash function optimized for speed and parallelism. It significantly outperforms older algorithms like SHA-256 or MD5 while maintaining collision resistance — see the full BLAKE3 in DuoBolt page for the algorithm comparison and how DuoBolt applies it.

The default head+tail prehash filters candidates efficiently before full hashing, reducing unnecessary work on large datasets. Full-content hashing still occurs for matched candidates, ensuring accuracy remains perfect.

macOS and Windows share the same high-performance engine, so both platforms inherit these optimizations without compromising accuracy or features.


Core Concepts

Learn how DuoBolt’s engine works under the hood, including hashing, prehashing, and parallelism strategies.

Read Core Concepts →

Similar photos

How DuoBolt groups photos that look alike, and what AI Curation measures before it recommends anything.

Read How grouping works →

FAQ

Common questions about performance, accuracy, and optimization.

Read FAQ →