Iluvatar CoreX BI-V100 (Tiangai 100) product page - vendor prints memory as '32 GB DRAM HBM2', a wording no other vendor uses and which must not be copied to the MR-V100

VERBATIM 2026-10-05 via web_extract. Recorded because data/accelerators/iluvatar-bi-v100-mr-v100 collapses the BI-V100 and the MR-V100 into one record and stored memory_type 'hbm2e' for the pair - which is right for the MR-V100 and wrong for the BI-V100. THE 产品规格 TABLE, verbatim, in full: 软件生态 | 兼容主流通用计算软件框架 支持主流深度学习开发框架 散热方式 | 被动散热 尺寸 | 全长全高双槽 PCIe 卡，兼容主流服务器 功耗 | 板级功耗 250W 接口 | PCIe Gen4.0 x 16 lane 共享 64 GB/s 主控双向带宽 共享 64 GB/s 片间互联带宽 内存 | 32 GB DRAM HBM2 架构 | 通用GPU 型号 | 天垓100加速卡（BI -V100） THE MEMORY STRING IS '32 GB DRAM HBM2'. Read literally that is 'DRAM' and 'HBM2' in one field. It is not a typo to be silently normalised: the pairing is Iluvatar's own way of distinguishing this card from its own inference line, which prints plain '32 GB HBM2E' with no 'DRAM'. Whatever the intent, the two strings differ, and the record must carry the difference rather than pick the convenient one. It is also the only place in this whole slice where a vendor prints the word 'DRAM' next to an HBM part, so a reader who has never seen it may assume a transcription error and 'fix' it. NOTHING COMPUTE-RELATED APPEARS ON THIS PAGE. From 产品特点: '提供业界领先的高算力和高能效比，在百余个算法模型的测试平均性能可媲美主流产品' - industry-leading compute and energy efficiency, average performance across 100+ benchmark models comparable to mainstream products. '支持200余种人工智能模型'. '天垓100的核心IP、系统架构、指令集、核心算子、软件栈均由天数科学家团队开发完成'. No TFLOPS, no TOPS, no precision-with-number anywhere. The absence is genuine and is now confirmed by two independent methods (this page, and the sibling MR-V100 page which is equally silent). THE INTERCONNECT STRING IS STILL AMBIGUOUS IN THE VENDOR'S OWN WORDS and this record's existing note about it is confirmed: the 接口 row prints '共享 64 GB/s 主控双向带宽' and '共享 64 GB/s 片间互联带宽' - the same figure 64 GB/s for both host bidirectional bandwidth and device-to-device interconnect bandwidth, with '共享' (shared) doing ambiguous work. It is not stated whether these are one shared resource or two parallel ones. Do not resolve it. SPARSITY: no footnote, no '2:4', no sparsity wording. Moot given nothing numeric is printed.