Skip to main content
All comparisons/ScramDB vs SQL Server

ScramDB vs SQL Server

SQL Server can do analytics, in the edition you pay per core for. Columnstore memory is capped at roughly 1.4 GB on Express and 128 GB on Standard, batch-mode parallelism is capped at one on Express and Web and two on Standard, and the uncapped behaviour plus SIMD aggregates arrive only with Enterprise. ScramDB is columnar and parallel across every core in the one edition that exists, and it holds the transactions on that same copy.

Columnar analytics
Included
Columnstore memory and batch-mode parallelism are capped below Enterprise edition.
Licensing
No core count
Standard and Enterprise are licensed per core, with a four core minimum per server.
Concurrency
Readers never block writers
SQL Server is lock based by default; row versioning is opt-in through snapshot isolation.
Free ceiling
64 GiB
Express caps a database at 50 GB, with about 1.4 GB of columnstore memory.
At a glance

Side by side

ScramDBSQL Server
Analytical executionColumnar and parallel across every core, in the one edition that existsColumnstore memory capped below Enterprise; batch-mode parallelism capped at 1 on Express and Web, 2 on Standard
LicensingCommunity free up to 64 GiB stored. Enterprise for clustering and beyondPer core, minimum four cores per server, on Standard and Enterprise
Free edition ceiling64 GiB of stored data, every CPU core you haveExpress: 50 GB per database, 4 sockets or 32 cores, 1.4 GB of columnstore memory
Concurrency modelReaders never block writersLock based by default; row versioning is opt-in via snapshot isolation
Wire protocolPostgreSQL wire protocol, every standard driver and BI toolTDS, with its own driver stack
BranchingFork the database instantly, or open it as it was at any past timestampBackup and restore
AI and agentsA semantic layer inside the engine that agents connect to directly, bound by the caller’s own permissionsNative vector type in 2025; the vector index is in public preview
PlatformsOne static binary on Linux, x86-64 or arm64Windows, Linux and containers

The edition matrix is the product

Everything that makes SQL Server quick at analytics is metered by edition. Express gets columnstore with about 1.4 GB of memory for it and batch-mode parallelism of one. Standard raises the memory and caps parallelism at two. Uncapped columnstore memory, full parallelism and SIMD aggregates arrive with Enterprise, and Enterprise is licensed by the core with a four core minimum.

ScramDB has one engine. Every core of the machine, the whole columnar path, the row-level security and the agent interface, with no feature flag tied to a purchase order.

Readers that never block

SQL Server defaults to lock-based concurrency, so a long reporting query and a busy write path contend until someone enables snapshot isolation and accepts the version store that comes with it. In ScramDB the reporting query sees a consistent picture of the data and never blocks the writes.

When SQL Server is the right answer

A Windows and .NET estate with SSIS, SSRS and Active Directory woven through it, where the tooling and the operational knowledge are worth more than the licence line.

The measurement, as of 5 September 2026. A full analytical workload over 100 million rows of real web analytics data, run on a 16 vCPU AWS instance. Ours is a single ScramDB node in its stock configuration, with the GPU off and nothing tuned for the test. Rival figures are those systems’ own published results for the same instance type. These are the numbers as of that date, not a permanent scoreboard: the engine is tuned continuously and it has got faster through every round of this work, so a later run should read better than this one.

Run it yourself in one line

One static binary, no dependencies. It picks its own config, listens on the postgres port 5432 and runs in the background.

$curl -fsSL https://scramdb.com/install | bash

Quick start · How the engine works · Talk to us

A full analytical workload over 100 million rows of real web analytics data, run on a 16 vCPU AWS instance, measured 5 September 2026. Ours is a single ScramDB node in its stock configuration, with the GPU off and nothing tuned for the test. Rival figures are those systems’ own published results for the same instance type. These are the numbers as of that date, not a permanent scoreboard: the engine is tuned continuously and it has got faster through every round of this work, so a later run should read better than this one.