Your BullMQ workers may be costing you 2.5x more than they should

Everyone picks Intel by default. We did too, until we ran the numbers.

We benchmarked BullMQ on three AWS instance families — Intel, AMD, and Graviton — processing 50,000 jobs per run across three Valkey versions. Same workload, same config, five runs each.

The cheapest instance delivered the most throughput. Here is jobs per dollar across the three platforms:

Intel c6i.xlarge ($0.170/hr)108.8M jobs/$

5,140 jobs/s

AMD c7a.xlarge ($0.153/hr)197.8M jobs/$ (1.8x)

8,409 jobs/s

Graviton c8g.xlarge ($0.136/hr)268.8M jobs/$ (2.5x)

10,154 jobs/s

Graviton is 20% cheaper per hour than Intel and processes nearly double the jobs per second. Combined, that is 2.5 times more jobs for every dollar you spend.

The kicker: Graviton was also the most stable across all three Valkey versions we tested. At concurrency 50, all versions landed within 1% of each other. On Intel and AMD, version choice caused 5–12% swings.

If you are running BullMQ on AWS and haven't tried Graviton, you are leaving money on the table.