We upgraded to Valkey 9.0. Here is what happened.

Valkey 9.0 is the latest release. Naturally we assumed it would be the fastest. It wasn't.

On our AMD c7a instance at concurrency 50, Valkey 8.1 pushed 34,207 jobs per second. We swapped in 9.0 expecting at least the same. We got 30,483. That is an 11% regression on the exact same hardware, same workload, same config.

Valkey 7.2-7%

31,672 jobs/s

Valkey 8.1baseline

34,207 jobs/s

Valkey 9.0-11%

30,483 jobs/s

This was AMD specific. On Graviton all three versions landed within 1% of each other. On Intel the differences were small too. Something about 9.0 interacts poorly with AMD's architecture under high concurrency.

If you are running BullMQ on AMD and considering an upgrade, 8.1 is where you want to be. Not 9.0, not yet.