This post gives test timings for different configurations in Autopilot. It follows on from a previous one about modern deployment in 2025. The aim is to see how the new Autopilot Device Preparation (Autopilot v2.0) compares with the classic Autopilot (Autopilot v1.0).
Autopilot v2.0 has significant differences in architecture, which Microsoft says will provide a faster and more reliable setup. This is important for the end-user experience. I have run different configurations in a test environment to see how they compare, and these are the results.
Methodology
First is a description of the test methodology. The tests were done as follows:
- Using two identical VMs in Hyper-V
- Each VM has 4 GB of vMem, and 12 vCPUs
- Running on an HP Z2 workstation G9
- With about 650 Mbps download and 65 Mbps upload speed
- A clean ISO of Windows 11 24H2
- One VM registered as an Autopilot device, the other not
- In Intune, security baselines and a policy each for Microsoft 365 Apps, Edge and Device Inventory: same policies for all.
For apps:
- The built-in Microsoft 365 Apps (actually delivered by the Office CSP)
- Company Portal
- Custom win32 app for Adobe Acrobat Reader DC
- Microsoft Visual C++ 2008 Redistributable
- Microsoft Visual C++ 2015-2022 Redistributable
- 7-Zip
- Google Chrome
- Microsoft Visual Studio Code.
The first three apps on the list were used for a partial list of apps. The next five came from the new Enterprise App Catalog.
The Enterprise App Catalog apps were used as an easy way to pad out the deployment. Curiously, however, they cannot be added to a list of apps in either version of Autopilot.
All apps were assigned to both a dynamic device group for Autopilot-registered devices, and a static device group for Device Preparation.
Pre-provisioning cannot be done with a Hyper-V VM, so this was not tested. As there is no pre-provisioning in Autopilot Device Preparation, it would not be possible to make a comparison.
Configurations
- OOBE
- Account not assigned a Device Preparation profile, and device not registered as an Autopilot device
- Represents the least possible time to deploy a device.
- Autopilot with 3 blocking apps
- Device is registered as an Autopilot device
- Profile is configured with 3 blocking apps.
- Autopilot with all apps blocking*
- Device is registered as an Autopilot device
- Profile is configured with all apps blocking.
- Autopilot Device Preparation with 3 reference apps
- User is a member of group assigned the device preparation profile
- Profile is configured with 3 reference apps.
* In this configuration, I had the same intermittent error described here: Autopilot error with Microsoft 365 Apps.
Timing
Timings given are the average of three runs, given in minutes and seconds.
| Configuration | Time | Time minus OOBE | Intune Time | Comment |
| OOBE | 03:05 | 00:00 | 00:00 | |
| Autopilot with 3 blocking apps | 10:15 | 07:10 | 07:38 | No more apps installed on completion |
| Autopilot with all apps blocking | 12:30 | 09:25 | 10:19 | All apps installed |
| Autopilot Device Preparation with 3 reference apps | 13:07 | 10:02 | 09:27 | Apps continue to install after completion |
OOBE represents the time taken for activities that are common across all configurations: authentication, MFA prompts, language and keyboard settings, Windows updates and reboots, Windows Hello for Business. The main difference is that, in Device Preparation, the privacy settings are not suppressed. This adds possibly 5-10 seconds. You can set a policy to “Disable Privacy Experience”, but I did not.
The time minus OOBE represents the amount of variable time, depending on the work. I would expect a production deployment to take a multiple of this, having more apps and a slower network. For example, a time of 09:05 times 2 equals 18:10, plus 03:05 for OOBE equals 21:15 would be good for a user-driven deployment in my experience. A time of 09:05 times 3 equals 27:15, plus 03:05 for OOBE equals 30:20 would not be unusual.
The Intune time is the time reported in Intune by the enrollment monitoring service. From observation, this is the time in the “Setting up for work or school” dialogue.
Interpretation
In this test, the new Autopilot Device Preparation is not faster than classic Autopilot.
Device Preparation with 3 reference apps is significantly slower than Autopilot with 3 blocking apps. Device Preparation with 3 reference apps is still slower than Autopilot with all apps blocking (i.e. installing 8 apps), although less so.
This is surprising. Enrolment Time Grouping, in Autopilot Device Preparation, is supposed to make it faster because Intune only needs to enumerate the apps assigned to the one group. In classic Autopilot, in the ESP, you can see a significant amount of time spent in “Apps (Identifying)”. But this change does not seem to result in a shorter elapsed time for Device Preparation.
I was not able to test Device Preparation with all apps (actually a maximum of 10) referenced. But I can estimate it would be 2-3 minutes slower, because that is the amount of extra time taken in classic Autopilot.
In effect, the reference apps feature of Device Preparation enables you to close the gap on how much slower it is than classic Autopilot.
Curiously, the times reported in Intune give a shorter time for Device preparation with 3 reference apps than for classic Autopilot with all apps blocking, whereas the total elapsed time was longer. At the moment, I do not know if this is because they measure different things, or some other reason outside this calculated time.
Conclusions
The new Autopilot Device Preparation (Autopilot v2.0) service is not faster than classic Autopilot (Autopilot v1.0), despite the changes in architecture.
However, the new Reference Apps feature will make deployment faster in situations where you have more default applications, or large ones, or both. I can see this being quite a common scenario.
With classic Autopilot, if you select to block on only a few apps, Intune does not carry straight on at the end of Autopilot to install the remaining apps. Instead, it halts and waits for a new sync cycle. This limits the usefulness of blocking. With Autopilot Device Preparation, if you select only a few reference apps, Intune does carry straight on afterwards with the remaining apps. I think this makes it a practical choice with an acceptable experience for the user. The more apps you have, the more useful it is.
Speed is not the only factor for deployment, of course. Reliability is also important. But this post is only about the timings.