A friend made this meme about me following a recent conversation where I pointed out that they could save 20% on their AWS Lambdas with this simple trick. Lambda pricing used to be determined by the amount of memory you have configured for the Lambda and the number of ms that it spends executing, but now, if your workload can be compiled for an arm64 architecture, you can deploy it on Graviton2 for an additional 20% savings.
| Memory (MB) | Price per 1ms (x86) | Price per 1ms (arm64) | Difference |
|---|---|---|---|
| 128 | 0.0000000021 | 0.0000000017 | 19.05% |
| 512 | 0.0000000083 | 0.0000000067 | 19.28% |
| 1024 | 0.0000000167 | 0.0000000133 | 20.34% |
| 1536 | 0.0000000250 | 0.0000000200 | 20.00% |
| 2048 | 0.0000000333 | 0.0000000267 | 19.82% |
| 3072 | 0.0000000500 | 0.0000000400 | 20.00% |
| 4096 | 0.0000000667 | 0.0000000533 | 20.09% |
| 5120 | 0.0000000833 | 0.0000000667 | 19.93% |
| 6144 | 0.0000001000 | 0.0000000800 | 20.00% |
| 7168 | 0.0000001167 | 0.0000000933 | 20.05% |
| 8192 | 0.0000001333 | 0.0000001067 | 19.95% |
| 9216 | 0.0000001500 | 0.0000001200 | 20.00% |
| 10240 | 0.0000001667 | 0.0000001333 | 20.04% |
| 19.89% |
There are other advantages to also run your workload on Graviton. Graviton CPUs are considered almost 20% more performant than their x86 rivals. As a result, most workloads result in 30-40% savings overall.