What is Power BI Premium?

This article describes Power BI Premium operations and considerations, to view a high level description of how Microsoft Fabric works, see Microsoft Fabric concepts and licenses.

Power BI Premium provides enhancements to Power BI, and a comprehensive portfolio of Premium features. The following table lists some of the Premium enhancements.

Enhancement Details
Purchase Premium for individuals in your organization See Power BI Premium Per User (PPU).
Improved metrics Capacity performance depends only on the amount of CPU usage. Metrics can be easily understood using the Microsoft Fabric Capacity Metrics app.
Autoscale An optional feature that prevents slowdowns caused by throttling on overloaded capacities. When enabled, if the load on the capacity exceeds the capacity limits, autoscale automatically adds one v-core at a time for 24-hour periods. Additional v-cores are charged to your Azure subscription on a pay-as-you-go basis.

Capacities and SKUs

Capacity is a dedicated set of resources reserved for exclusive use. It offers dependable, consistent performance for your content.

Each capacity offers a selection of SKUs, and each SKU provides different resource tiers for computing power. The type of SKU you require, depends on the type of solution you wish to deploy.

SKU Capacity Units (CU) Power BI SKU Power BI v-cores
F2 2 N/A N/A
F4 4 N/A N/A
F8 8 EM1/A1 1
F16 16 EM2/A2 2
F32 32 EM3/A3 4
F64 64 P1/A4 8
F128 128 P2/A5 16
F256 256 P3/A6 32
F5121 512 P4/A7 64
F10241 1,024 P5/A8 128
F20481 2,048 N/A N/A

1 These SKUs aren't available in all regions. To request using these SKUs in regions where they're not available, contact your Microsoft account manager.

Subscriptions and licensing

Power BI Premium is a tenant-level Microsoft 365 subscription, available in two SKU (Stock-Keeping Unit) families.

P EM
Range P1-P5 EM1-EM3
Use Enterprise features and embedding Organizational embedding (embed for your organization)
Commitment Monthly or yearly Yearly
Billing Monthly Monthly
Additional information Includes a license to install Power BI Report Server on-premises EM1 and EM2 SKUs are available only through volume licensing plans. You can't purchase them directly.

Purchasing

Power BI Premium subscriptions are purchased by administrators in the Microsoft 365 admin center. Specifically, only global administrators or billing administrators can purchase SKUs. When purchased, the tenant receives a corresponding number of v-cores to assign to capacities, known as v-core pooling. For example, purchasing a P3 SKU provides the tenant with 32 v-cores. To learn more, see How to purchase Power BI Premium.

Workspaces

Workspaces reside within capacities. Each Power BI user has a personal workspace known as My Workspace. Additional workspaces known as workspaces can be created to enable collaboration. By default, workspaces, including personal workspaces, are created in the shared capacity. When you have Premium capacities, both My Workspaces and workspaces can be assigned to Premium capacities.

Capacity administrators automatically have their My workspaces assigned to Premium capacities.

Semantic model SKU limitation

With Power BI Premium and Power BI Embedded, there are memory limits and other constraints for each SKU listed in the table below.

SKU Max memory (GB)1, 2 Max concurrent DirectQuery connections (per semantic model)1 Live connection (per second)1 Max memory per query (GB)1 Model refresh parallelism Direct Lake rows per table (in millions)1, 3 Max Direct Lake model size on OneLake (GB)1, 3
F2 3 5 2 1 1 300 10
F4 3 5 2 1 2 300 10
F8 3 10 3.75 1 5 300 10
F16 5 10 7.5 2 10 300 20
F32 10 10 15 5 20 300 40
F64 25 50 30 10 40 1,500 Unlimited
F128 50 75 60 10 80 3,000 Unlimited
F256 100 100 120 10 160 6,000 Unlimited
F512 200 200 240 20 320 12,000 Unlimited
F1024 400 200 480 40 640 24,000 Unlimited
F2048 400 200 960 40 1,280 24,000 Unlimited

1 The Microsoft Fabric Capacity Metrics app doesn't currently expose these metrics.

2 The Max memory (GB) column represents an upper bound for the semantic model size. However, an amount of memory must be reserved for operations such as refreshes and queries on the semantic model. The maximum semantic model size permitted on a capacity might be smaller than the numbers in this column.

3 These limits apply to Direct Lake tables and models, and are guardrails that affect fallback to DirectQuery. Direct Lake semantic models have additional constraints that are based on SKUs, as listed in fallback.

Semantic model memory usage

Semantic model operations such as queries are subject to individual memory limits. To illustrate the restriction, consider a semantic model with an in-memory footprint of 1 GB, and a user initiating an on-demand refresh while interacting with a report based on the same semantic model. Three separate actions determine the amount of memory attributed to the original semantic model, which may be larger than two times the semantic model size. The total amount of memory used by one Power BI item can't exceed the SKU's Max memory per semantic model allocation.

  • Loading the semantic model - The first action is loading the semantic model into the memory.

  • Refreshing the semantic model - The second action is refreshing the semantic model after it's loaded into the memory. The refresh operation will cause the memory used by the semantic model to double. The required memory doubles because the original copy of data is still available for active queries, while another copy is being processed by the refresh. Once the refresh transaction commits, the memory footprint will reduce.

  • Interacting with the report - The third action is caused by the user's interaction with the report. During the semantic model refresh, report interactions will execute DAX queries. Each DAX query consumes a certain amount of temporary memory required to produce the results. Each query may consume a different amount of memory. The memory used to query the semantic model is added to the memory needed to load the semantic model, and refresh it.

Refreshes

Power BI Premium and Power BI Embedded don't require cumulative memory limits, and therefore concurrent semantic model refreshes don't contribute to resource constraints. However, refreshing individual semantic models is governed by existing capacity memory and CPU limits, and the model refresh parallelism limit for the SKU, as described in Capacities and SKUs.

You can schedule and run as many refreshes as required at any given time, and the Power BI service will run those refreshes at the time scheduled as a best effort.

Monitoring

When monitoring Power BI Premium and Power BI Embedded, you only need to take into consideration one aspect: how much CPU your capacity requires to serve the load at any moment. To monitor your capacity, use the Microsoft Fabric Capacity Metrics app app.

To install the app, see Install the Microsoft Fabric capacity metrics app. You can learn how to use the app in the article Microsoft Fabric Capacity Metrics.

Here's what happens when you exceed your CPU limit per the SKU size you purchased:

Paginated reports

When using Power BI Premium and Power BI Embedded, Power BI paginated reports benefit from the architectural and engineering improvements reflected in Power BI Premium.

  • Memory - There's no memory management for Paginated reports.

  • SKU availability - Paginated reports running on Power BI Premium can run reports across all available embedded and Premium SKUs, including the EM1-EM3 and A1-A3 SKUs. Billing is calculated per CPU hour, across a 24-hour period.

  • Enhanced security and code isolation - Code isolation occurs at a per-user level, rather than at a per-capacity level.

Dataflows Gen1

Each SKU can run a set number of Dataflows Gen1 parallel tasks, as listed in this table.

Fabric SKU Dataflow parallel tasks
F2 2
F4 2
F8 4
F16 8
F32 16
F64 32
F128 64
F256 64
F512 64
F1024 64
F2048 64

To learn about Dataflow Gen2, see Getting from Dataflow Generation 1 to Dataflow Generation 2.

Considerations and limitations

The following known limitations currently apply to Power BI Premium.

  • Rendering visuals - There's a 225-second limitation for rendering Power BI visuals. Visuals that take longer to render, will be timed-out and won't display.

  • Throttling - Throttling can occur in Power BI Premium capacities. Concurrency limits are applied per session. An error message will appear when too many operations are being processed concurrently. To mitigate throttling, you can use autoscale. When autoscale is enabled, if CPU consumption exceeds the additional limits, throttling will still take place. To read more about throttling in Fabric, see The Fabric throttling policy.

  • Client library version - Client applications and tools that connect to and work with semantic models on Premium capacities through the XMLA endpoint require Analysis Services client libraries. Most client applications and tools install the most recent client libraries with regular updates, so manually installing the client libraries isn't usually necessary. Regardless of the client application or tool version, the following minimum client library versions are required.

    Client Library Version
    MSOLAP 15.1.65.22
    AMO 19.12.7.0
    ADOMD 19.12.7.0

    In some cases, manually installing the most recent client libraries may be necessary to reduce potential connection and operation errors. To learn more about verifying existing installed client library versions and manually installing the most recent versions, see Analysis Services client libraries.

  • Semantic models compatibility - Some semantic models are incompatible with the Power BI service modern infrastructure:

    • Semantic models created in Power BI service from CSV files.
    • Semantic models in the admin monitoring workspace and usage metrics models.
    • Semantic models that still use 1103 compatibility level.
    • Push semantic models.
    • Semantic models that use deprecated features such as content packs.