.NET 8 vs .NET 10 EF Core Performance Benchmarks Across SQLite, InMemory, and Azure SQL

🚀

Over the past few weeks, I’ve been benchmarking Entity Framework Core to understand what .NET 10 actually improves in real workloads.

Not release notes. Not marketing slides. Actual code. Actual queries. Actual numbers.

This article compares .NET 8 vs .NET 10 using the same EF Core codebase across three very different database providers:

🗂 SQLite (file-based)

⚡ InMemory (no I/O, runtime-focused)

☁️ Azure SQL (real networked SQL Server)

The goal is simple: understand where .NET 10 helps, where it doesn’t, and why.

🧪 Benchmark Setup Runtime & Tooling

BenchmarkDotNet for measurement

Same codebase, same queries, same dataset

Only runtime and provider change

Sample Data Shape Blogs : 1,000 Posts : 10,000 (1-to-many with Blogs) Users : 1,000 Products : 100

This dataset intentionally mixes:

Simple queries

Tracking vs AsNoTracking

Inserts, updates, deletes

Complex LINQ projections

Raw SQL paths

🗂 SQLite Provider Results

Benchmark	.NET 8 Mean	.NET 10 Mean	Improvement	.NET 8 Allocated	.NET 10 Allocated
Ef.AsNoTracking	102,674.46 ns	75,025.46 ns	[+] +26.9%	42281 B	45873 B
Ef.BulkInsert	31,672,144.12 ns	29,318,670.42 ns	[+] +7.4%	11538680 B	10309430 B
Ef.BulkUpdate	123,968,292.43 ns	128,406,550.20 ns	-3.6%	59492305 B	57383714 B
Ef.Caching	36.54 ns	29.27 ns	[+] +19.9%	64 B	0 B
Ef.ColdStart	104,321.34 ns	102,538.32 ns	+1.7%	59567 B	62868 B
Ef.ComplexQuery	114,037.51 ns	173,775.50 ns	[-] -52.4%	42016 B	54736 B
Ef.Delete	1,654,737.89 ns	1,727,686.80 ns	-4.4%	79423 B	82730 B
Ef.Insert	873,156.78 ns	895,244.05 ns	-2.5%	66383 B	68131 B
Ef.LinqQuery	67,529.55 ns	64,983.62 ns	+3.8%	35168 B	37696 B
Ef.QueryTop100	54,568.43 ns	57,398.64 ns	[-] -5.2%	34513 B	37569 B
Ef.RawSql	65,609.71 ns	61,233.70 ns	[+] +6.7%	38185 B	40384 B
Ef.SimpleQuery	36.12 ns	24.89 ns	[+] +31.1%	88 B	56 B
Ef.Update	104,024.48 ns	101,779.97 ns	+2.2%	60140 B	63236 B

SQLite is lightweight and fast, but its query planner and execution model can behave very differently from server-based databases.

Key Observations

✅ Clear wins in common paths

SimpleQuery: +31.1%

AsNoTracking: +26.9%

Caching: +19.9%

RawSql: +6.7%

BulkInsert: +7.4%

⚠️ Noticeable regressions

ComplexQuery: -52.4%

QueryTop100: -5.2%

What This Means

.NET 10 improves hot paths and low-level execution, but complex LINQ translations on SQLite can amplify regressions. This isn’t unique to .NET 10. SQLite has always been sensitive to query shape.

⚡ InMemory Provider Results

Benchmark	.NET 8 Mean	.NET 10 Mean	Improvement	.NET 8 Allocated	.NET 10 Allocated
AsNoTracking	39,156.38 ns	38,308.54 ns	+2.2%	103378 B	95986 B
BulkInsert	3,644,025.27 ns	3,777,690.15 ns	-3.7%	4039985 B	3351774 B
BulkUpdate	7,193,296.40 ns	7,514,367.05 ns	-4.5%	52364986 B	44947268 B
Caching	36.39 ns	30.45 ns	[+] +16.3%	64 B	0 B
ColdStart	61,243.14 ns	59,822.17 ns	+2.3%	103376 B	95476 B
ComplexQuery	1,333,247.11 ns	456,938.87 ns	[+] +65.7%	1175145 B	985288 B
Delete	41,059.23 ns	35,994.73 ns	[+] +12.3%	34234 B	33132 B
Insert	33,996.85 ns	31,247.18 ns	[+] +8.1%	30794 B	30396 B
LinqQuery	51,215.60 ns	48,373.73 ns	[+] +5.5%	108488 B	100304 B
QueryTop100	39,525.17 ns	37,632.59 ns	+4.8%	98810 B	90882 B
RawSql	NA	NA	N/A	NA	NA
SimpleQuery	36.86 ns	25.55 ns	[+] +30.7%	88 B	56 B
Update	62,905.66 ns	63,223.70 ns	-0.5%	95992 B	87882 B

This provider removes disk and network entirely. What you measure here is mostly:

EF Core internal behavior

LINQ execution

Runtime improvements

And this is where .NET 10 really stands out.

Highlights

🔥 Big improvements

ComplexQuery: +65.7%

SimpleQuery: +30.7%

Caching: +16.3%

Delete: +12.3%

Insert: +8.1%

⚖️ Minor regressions

BulkInsert and BulkUpdate are slightly slower

What This Means

When I/O is removed, .NET 10’s runtime improvements are very visible. If your app relies heavily on complex LINQ queries or projections, this is the strongest signal that upgrading can help.

☁️ Azure SQL Provider Results

Benchmark	.NET 8 Mean	.NET 10 Mean	Improvement	.NET 8 Allocated	.NET 10 Allocated
AsNoTracking	222,421.48 ns	202,049.18 ns	[+] +9.2%	54011 B	42650 B
BulkInsert	18,761,982.45 ns	18,077,035.58 ns	+3.7%	9728528 B	8354063 B
BulkUpdate	NA	96,384,991.95 ns	N/A	NA	25233403 B
Caching	37.60 ns	30.04 ns	[+] +20.1%	64 B	0 B
ColdStart	246,371.30 ns	239,864.02 ns	+2.6%	90743 B	90542 B
ComplexQuery	1,597,342.25 ns	1,699,335.67 ns	[-] -6.4%	62658 B	60482 B
Delete	2,809,270.73 ns	3,322,021.47 ns	[-] -18.3%	111285 B	108805 B
Insert	375,608.11 ns	356,960.98 ns	+5.0%	98763 B	97103 B
LinqQuery	186,734.97 ns	189,260.08 ns	-1.4%	33657 B	33985 B
QueryTop100	211,699.48 ns	201,654.46 ns	+4.7%	36722 B	34346 B
RawSql	193,049.71 ns	195,063.53 ns	-1.0%	37402 B	36929 B
SimpleQuery	38.85 ns	25.15 ns	[+] +35.3%	88 B	56 B
Update	260,622.89 ns	246,550.18 ns	[+] +5.4%	91013 B	90647 B

Azure SQL introduces:

Network latency

Server-side execution plans

Real-world variability

This makes results more realistic and more nuanced.

Improvements

✅ Consistent gains

SimpleQuery: +35.3%

Caching: +20.1%

AsNoTracking: +9.2%

Update: +5.4%

Insert: +5.0%

⚠️ Regressions

Delete: -18.3%

ComplexQuery: -6.4%

RawSql and LinqQuery mostly flat

What This Means

Once network and server execution dominate, runtime improvements matter less. .NET 10 still helps, but it won’t magically overcome latency or poor query plans.

🧠 Patterns Across All Providers

After running all three providers, a few patterns are very clear:

1️⃣ Simple queries consistently get faster 2️⃣ Caching improvements are real and repeatable 3️⃣ Complex queries are provider-dependent 4️⃣ InMemory exaggerates wins 5️⃣ Azure SQL dampens them

There is no single “.NET 10 is X% faster” number that makes sense.

📉 Why Some Regressions Exist

A regression doesn’t automatically mean “worse code.”

Possible reasons include:

Different query plans

Changed execution order

Provider-specific behavior

Runtime tradeoffs favoring common cases

This is exactly why benchmarking your own workload matters more than reading generic numbers.

🔁 Why I’m Sharing This

These results reflect:

My machine

My dataset

My environment

Change any of those and the numbers will change too.

That’s why I’m sharing the benchmarks and encouraging others to run them:

Different CPUs

Different cloud regions

Different data sizes

Different providers

Benchmarks are a starting point, not a verdict.

🤝 Want to Compare Results?

If you’re running EF Core on:

SQLite

InMemory

SQL Server or Azure SQL

I’d love to see what you get on your setup.

Different environments tell a better story than any single benchmark ever could.

🏁 Final Thoughts

.NET 10 brings real, measurable improvements to EF Core, especially in:

Simple queries

Caching

LINQ-heavy workloads

But databases are complex systems. Runtime upgrades help, but they don’t replace good schema design, indexing, and query discipline.

Measure first. Upgrade second.

If you want the repo or want to discuss results, feel free to reach out 👋