---
title: "Discussion on \"Fabric : Not All Delta Tables Are Created Equally\" | Hashnode"
description: "Discussion on \"Fabric : Not All Delta Tables Are Created Equally\". It matters which engine is used to create the Delta table, especially for optimal Direct Lake performance."
url: "https://hashnode.com/posts/fabric-not-all-delta-tables-are-created-equally/649a68da170c64d6eb66d7e1"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Fabric : Not All Delta Tables Are Created Equally](https://fabric.guru/fabric-not-all-delta-tables-are-created-equally?utm_source=hashnode&utm_medium=feed)

It matters which engine is used to create the Delta table, especially for optimal Direct Lake performance.

Jun 27, 2023 · 6 min read · fabric.guru  
Updated Jul 5, 2023  
Read the article

-   [#microsoftfabric](https://hashnode.com/tag/microsoftfabric)
-   [#performance-optimization](https://hashnode.com/tag/performance-optimization)
-   [#delta-table](https://hashnode.com/tag/delta-table)
-   [#direct-lake](https://hashnode.com/tag/direct-lake)

## Written by

[Sandeep Pawar](https://hashnode.com/@PawarBI)

@PawarBI

  
[View profile](https://hashnode.com/@PawarBI)

Building Data Analytics and Data Science solutions using Microsoft Fabric | Principal PM, Microsoft Fabric CAT

Principal Program Manager, Microsoft Fabric CAT helping users and organizations build scalable, insightful, secure solutions. Blogs, opinions are my own and do not represent my employer.

## Responses 2

Markdown supported · ⌘↵ to post

[Lamin Jobe](https://hashnode.com/@Oddity)

[Aug 30, 2023](https://hashnode.com/posts/fabric-not-all-delta-tables-are-created-equally/649a68da170c64d6eb66d7e1/comment/64ef0534a1860c5c6d0a8fdd)

Excellent article!

[Aravind Yarram](https://hashnode.com/@Yaravind)

[Jun 28, 2023](https://hashnode.com/posts/fabric-not-all-delta-tables-are-created-equally/649a68da170c64d6eb66d7e1/comment/649b91ea3f5eb98a5001df80)

Seems like DFg2 is using single thread to write. That explains single file and the time taken. Also, I expected ~600 Mil rows, i.e. all rows in an Row Group (RG) where there is only a single RG. Am I missing something?