![]() |
|
Don´t Expect Cheaper SSDs Anytime Soon: HBF Would Cost Just $0.64 Per GB But Consume 6.5x More Wafers Than A Typical TLC - Printable Version +- T0rrentinvites - #1 To Buy, Trade, Sell Or Find Free Tracker Invites! (https://t0rrentinvites.com) +-- Forum: Other Discussions (https://t0rrentinvites.com/forumdisplay.php?fid=30) +--- Forum: General PC Chat And Help (https://t0rrentinvites.com/forumdisplay.php?fid=33) +--- Thread: Don´t Expect Cheaper SSDs Anytime Soon: HBF Would Cost Just $0.64 Per GB But Consume 6.5x More Wafers Than A Typical TLC (/showthread.php?tid=22066) |
Don´t Expect Cheaper SSDs Anytime Soon: HBF Would Cost Just $0.64 Per GB But Consume 6.5x More Wafers Than A Typical TLC - Portugal.FC - 11-03-2026 ![]() If High Bandwidth Flash or HBF proves to be a commercial success, it would likely preclude cheaper SSDs for the foreseeable future, especially as HBF presumably entails just one-tenth of the cost of a comparable HBM4 stack, leading to a Jevons paradox-driven incremental demand.
An 8-Hi stack of HBF will cost $0.64 per GB vs. the $6.78 per GB cost for an 8-Hi HBM4 stack For the benefit of those who might not be aware, just as HBM stacks DRAM, HBF stacks NAND die on top of each other, with Through Silicon Vias (TSVs) connecting the dies together and a controller logic die bonded to this NAND array. RELATED STORY October 2026 Steam Hardware Survey: 16 GB RAM Reclaims Top Spot, Intel Gains CPU Share, Linux Approaches 4% The problem with using NAND cells, however, is their turtle-like speeds. An SRAM offers read speeds of just around a nanosecond vs. ~100 nanoseconds for DRAM and a whopping ~100 microseconds for NAND. This means that NAND offers 1,000x slower read speeds than a DRAM-based HBM. HBF, however, leverages the power of parallelism to increase bandwidth by orders of magnitude. Basically, the logic die schedules thousands of parallel reads of NAND cells simultaneously. So, while each individual read of a given NAND cell is around 1,000x slower, thousands of parallel reads can deliver a cumulative bandwidth of 0.3TB/s to 3TB/s. Even so, an HBF can't counter NAND's atrocious write speeds and the attendant fragile write endurance. At any rate, SanDisk is currently developing an HBF standard with SK hynix, one that would offer 512GB of storage along with a bandwidth of 0.4TB/s to 3TB/s, with a commercial rollout expected by 2028/2029. This brings us to the core of today's topic. The analysts at Bernstein have now estimated that a typical 8-Hi HBF stack will cost just $0.64 per GB, which is just around one-tenth the cost of a comparable HBM4 stack that entails a price of around $6.78 per GB. Given this substantial cost differential, HBF can feasibly displace HBM in several use cases, such as storing model weights, or when dealing with shared prefix KV caching - where thousands of distinct user prompts often query the exact same massive system prompt, codebase, or multi-hundred-page documentation context; the KV cache for this foundational "prefix" data is computed once and then written permanently to HBF. However, each 8-Hi HBF stack uses 6.5x more wafers than a typical TLC NAND, which would likely squeeze the available supply of NAND wafers once HBF enters commercialization phase, precluding cheaper SSDs. |