Good morning, everyone! 
I’m updating my earlier benchmark. This time, I used a much more powerful device, the latest version of Kirby, and PHP.
Testing Environment
For this benchmark, I used:
- Hardware: Intel i5-13600K with 32GB RAM
(Previous test: Intel i5-7300U with 8GB RAM) - Software:
- Kirby Starterkit v4.5.0
- PHP 8.3.1
- laragon (as the local environment)
Keep in mind that the hardware upgrade significantly influences the results. However, the numbers still offer valuable insights into Kirby’s performance IMHO.
User Creation Times
Here’s how long it took to create user accounts this time:
| Users | Creation Time |
|---|---|
| 1 | ~0 Seconds |
| 100 | ~4 Seconds |
| 1.000 | ~44 Seconds |
| 10.000 | ~7 Minutes 45 Seconds |
| 50.000 | ~1 Hour 18 Minutes 55 Seconds |
The new setup clearly excels, especially at scale, when compared to the previous benchmark. The structure is the same: each account generates one folder and three files, resulting in 200.000 elements for 50.000 users.
User Search Performance
This test shows how quickly Kirby could find or not find users by different criteria:
= Found
= Not Found
| Users | by UUID* |
by UUID* |
by Email* |
by Email* |
by Attribute* |
by Attribute* |
|---|---|---|---|---|---|---|
| 1 | ~0 | ~0 | ~0 | ~0 | ~0 | ~0 |
| 100 | ~0 | ~0 | ~0 | ~0 | ~0 | ~0 |
| 1,000 | ~0 | ~0 | ~0 | ~0 | ~0 | ~0 |
| 10,000 | ~1 | ~1 | ~1 | ~1 | ~1 ±1 | ~3 |
| 50,000 | ~7 | ~7 | ~10 ±2 | ~13 | ~31 ±3 | ~58 ±1 |
* (in Seconds)
Key Observations
- Consistency at Scale
Search times have become much more consistent compared to the 2021 benchmark. For instance, with 10.000 users, searches now average a constant 1 second, except for unmatched attribute search. - Improved Performance
Even with 50.000 accounts, the response times are much more stable. This suggests that Kirby’s data structures may have improved. - Hardware Matters
The upgraded test machine clearly helps, but the combination of the latest Kirby version and PHP enhancements plays a big role too. - Decreased Performance for Unmatched Attribute Searches
The only area where performance has slightly decreased is in searches for users by attributes when no match is found. While this decline is minor, it’s worth noting for large-scale implementations.
Conclusion
In 2021, I labeled 10,000 accounts as “okay-ish.” With Kirby’s latest version, a current PHP release, and sufficient computing power, I can confidently say it’s now absolutely fine—no flinch. 
Happy New Year! I’m looking forward to more improvements in 2025. Cheers!