Sources, rights, and provenance
Source categories
- OEM and qualified-system catalogs, configuration records, and manufacturer specifications
- Selected distributor, reseller, and broker listings, quotes, and availability observations where collection and use rights permit
- Documented or reported secondary-market sale observations where provenance and transaction status can be recorded
- Historical Rillor specification and indicative pricing research retained as research material rather than market tape
Rights posture
Every specification, listing, quote, sale record, and contributed observation requires source-level provenance and a recorded acquisition, use, and redistribution posture before commercial access.
Structure, labels, and lineage
Entity resolution and normalization
GPU family, accelerator count, memory, OEM, chassis, CPU, interconnect, cooling, condition, location, seller type, observation type, currency, unit, price basis, and observation time are modeled separately so comparable system grades can be constructed without treating unlike configurations as fungible.
Lineage
Each normalized record is designed to retain its source, capture time, evidence class, seller or venue category, configuration match, transformation history, dataset version, and applicable index-methodology version.
Quality, reconciliation, and versioning
Quality and reconciliation
Quality work covers SKU and OEM resolution, configuration equivalence, sale-status verification, condition and geography adjustments, currency and unit normalization, duplicate handling, outlier review, and contributor diversity. Index outputs require separate methodology, governance, and sufficiency gates.
Freshness and versioning
Refresh and reverification are scoped per source. No continuous or real-time market-data commitment is offered.
Delivery and access discussion
Access state
There is no public raw feed, download, API, live benchmark, settlement-grade index, or service-level commitment.
Delivery discussion
An authenticated API, a scheduled export, or an agreed structured format can be scoped for this dataset. Coverage, schema, authentication, cadence, versions, rights, and support are defined with the engagement.