{"id":"hbf-read-tier-contract","level":"software","name":"HBF: a read-oriented tier and its evidence boundary","shortName":"HBF: a read-oriented tier and its evidence boundary","role":"Distinguish an announced HBF interface/capacity specification from validated workload behavior.","description":"High Bandwidth Flash proposes a NAND-based tier near the processor, alongside the existing memory hierarchy. SK hynix’s August 2026 announcement describes up to 512 GB and bandwidth grades of approximately 0.4–3.0 TB/s, with UCIe connectivity. These are reported specification boundaries, not proof of sustained throughput on a deployed inference system. A capacity and read-bandwidth headline leaves access latency, request granularity, concurrency, write service and host software to be examined. NAND-based capacity can be attractive for read-heavy state, but it does not acquire DRAM semantics merely by sharing a package or a fast link.","mechanism":["Identify the source date and distinguish technical specification, sample, qualified component and shipping system.","Map the processor link, HBF controller and NAND array as separate interfaces; use the same transfer direction and byte convention.","Request latency, granularity, concurrency and read/write conditions before treating the headline rate as usable.","Keep mutable working state and recovery requirements in the placement contract rather than assuming every tensor can move to flash.","Measure application-level correctness, latency and energy on a named integration before claiming a deployment benefit."],"science":[{"title":"Minimum bulk-read transfer time","equation":"ttransfer ≥ V / B","explanation":"Moving V payload bytes cannot take less than V/B at usable bandwidth B, even before initial access, scheduling or contention.","assumptions":"Single direction with compatible decimal units; B is a sustained payload assumption, not automatically an advertised interface maximum."}],"specs":[{"label":"Evidence snapshot","value":"Announced · 4 August 2026","context":"Reported specification: up to 512 GB; grades approximately 0.4–3.0 TB/s."},{"label":"Interface","value":"UCIe in the announcement","context":"A common framework is not a drop-in compatibility or shipped-system certificate."},{"label":"Still configuration-dependent","value":"Latency, granularity, read/write behavior","context":"The headline grades do not settle end-to-end service performance."}],"interfaces":[{"name":"Processor/HBF interface","direction":"in","signal":"Requests, completion, error and transfer semantics"},{"name":"NAND-backed capacity","direction":"out","signal":"Data with specified access and persistence behavior"}],"tradeoff":"More near-compute capacity may reduce remote movement, while page access, writes and integration requirements can move the performance bottleneck.","whyAI":"Large read-mostly parameter pools are a possible use case; they require a software and system plan that accounts for access behavior.","prerequisites":["memory-interface-choice","die-link-compatibility"],"related":["hbf-latency-prefetch","memory-working-set-placement","memory-system-qualification"],"misconception":"An interface/specification announcement and a measured shipping-system result are different evidence states.","check":{"question":"What does the announced HBF bandwidth range establish?","options":["A reported specification range that still needs workload validation","Guaranteed HBM-equivalent random-access latency","Proof that every GPU can attach HBF today"],"answer":0,"explanation":"An interface/specification announcement and a measured shipping-system result are different evidence states."},"engineeringExample":"Synthetic payload-only calculation: reading 100 GB at an assumed sustained 1 TB/s requires at least 0.1 s. No setup, concurrency loss or other work is included. The assumed usable rate is not a measured HBF benchmark.","evidenceNotes":"SK hynix and Sandisk identify the August 2026 milestone as specification release. The atlas has not verified a volume-shipping deployment from these announcements. Missing access metrics remain unknown rather than receiving SSD or HBM defaults.","sources":[{"title":"SK hynix — first HBF specifications at FMS 2026","url":"https://news.skhynix.com/en/hbf-at-fms-2026/"},{"title":"Sandisk — release of first OCP HBF specification","url":"https://investor.sandisk.com/node/8126/pdf"},{"title":"UCIe — introduction Q&A recap","url":"https://www.uciexpress.org/post/introduction-to-ucie-webinar-q-a-recap"}],"learningObjective":"Distinguish an announced HBF interface/capacity specification from validated workload behavior.","workstream":"W11","visualFamily":"memory-route","scale":"Processor interface to NAND-backed capacity; logical topology, not a disclosed product floorplan","kind":"system","labId":"inference-placement","branch":"Memory","subbranch":"HBF interface, latency and access contract","scope":"Place read-only weights and mutable KV across specified tiers and inspect transfer paths; no HBF shipping-system benchmark or universal memory semantics are implied.","investigationId":"execution","labScope":"Place read-only weights and mutable KV across specified tiers and inspect transfer paths; no HBF shipping-system benchmark or universal memory semantics are implied."}