# Per-shard IOPS and Gen2 storage on Atlas > Source-anchored research for Per-shard IOPS and Gen2 storage on Atlas. Run status: COMPLETED. Parent: [Independent shard scaling and per-shard tiering](https://llms-explorer.com/tree/independent-shard-scaling-and-per-shard-tiering/) · 3 facets · 10 facts · page: https://llms-explorer.com/tree/per-shard-iops-and-gen2-storage-on-atlas/ ## How it works - Gen2 does not make Standard and High Performance IOPS equivalent. Atlas distinguishes General Purpose SSD from Provisioned IOPS SSD and recommends testing latency-sensitive workloads before switching. AWS likewise recommends io2 Block Express when low and consistent latency is required. (high confidence) (verified as of 2026-10-30) — [source](https://www.mongodb.com/docs/atlas/reference/faq/storage.md) - When updating a Gen2 cluster through the Atlas Administration API, retain the _GEN_2 suffix in instanceSize, for example M30_GEN_2. A plain Gen1 tier value can unintentionally change generation and thus the available storage/IOPS configuration. (medium confidence) (verified as of 2026-10-30) — [source](https://www.mongodb.com/docs/atlas/manage-clusters.md) ## Measurements and reference values - The documented Gen2 included baseline is 3,000 IOPS through 1,000 GB, then 3 IOPS per GB: 2,000 GB includes 6,000 IOPS and 4,000 GB includes 12,000. Storage pricing includes this baseline; Extended Standard IOPS above it incur a separate charge. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/customize-storage.md) - Jayabal Rajendran, identifying himself as InsuredMine’s engineering director, reports replacing an M80/2-TB configuration with M50 Gen2/1 TB and 12,000 custom IOPS. His September 26, 2026 account reports I/O wait falling from about 50% to below 3%; it is one workload report, not a controlled benchmark or evidence of per-shard deployment. (low confidence) (verified as of 2026-10-31) — [source](https://medium.com/@devloprjayabal/how-atlas-gen2-cut-our-iops-wait-from-50-to-3-and-halved-our-cluster-size-e458d3f2b4ca) ## Facts and statements - Atlas Core Gen2 M30+ clusters on AWS and Google Cloud support Extended Standard IOPS: provision additional IOPS separately from storage capacity. Documented platform ceilings are 80,000 IOPS on AWS and 160,000 on Google Cloud; these are not promises that every instance tier sustains that rate. (high confidence) (verified as of 2026-10-30) — [source](https://www.mongodb.com/docs/atlas/customize-storage.md) - For AWS Gen2, the MongoDB Terraform provider documents STANDARD as gp3 and supports disk_iops independently of storage size. HIGH_PERFORMANCE uses io2 on Gen2; PROVISIONED is the corresponding Gen1 io2 setting. These names select different storage performance characteristics, not merely different IOPS quantities. (high confidence) (verified as of 2026-10-30) — [source](https://github.com/mongodb/terraform-provider-mongodbatlas/blob/master/docs/resources/advanced_cluster.md) - The Atlas Administration API supports different IOPS per shard. The sharding guide describes AWS Standard IOPS as storage-size-dependent, while Gen2 documentation allows IOPS independent of capacity. The provider reference explicitly includes AWS Gen2 STANDARD (gp3) among clusters supporting independent disk_iops across replication_specs. (medium confidence) (verified as of 2026-10-31) — [source](https://www.mongodb.com/docs/atlas/cluster-sharding.md) - Moving from Gen1 to Gen2 preserves storage capacity and IOPS rate. On AWS, Provision IOPS maps to High Performance IOPS and Standard remains Standard. The retained rate can exceed the Gen2 included baseline, so review the resulting paid IOPS configuration after migration. (high confidence) (verified as of 2026-10-14) — [source](https://www.mongodb.com/docs/atlas/cluster-config/change-cluster-generation.md) - Atlas Core Gen2 storage controls must not be generalized to Atlas Infinite: Infinite uses Gen2 compute but does not support Extended Standard IOPS. Its storage performance follows its own tier-based model. (high confidence) (verified as of 2026-10-30) — [source](https://www.mongodb.com/docs/atlas/manage-clusters.md) - AWS gp3 independently provisions capacity and performance, supplies 3,000 baseline IOPS, and does not use burst credits. This is underlying-storage context for Atlas STANDARD gp3, not an independent specification of Atlas billing or per-shard behavior. (medium confidence) (verified as of 2026-10-30) — [source](https://docs.aws.amazon.com/ebs/latest/userguide/general-purpose.html)