China Hot Tech @ChinaHotTech
← All posts
Chips · GPU · Inference

Iluvatar CoreX Tiangai 300: Domestic GPU Claiming H200-Class E2E Throughput*

天数智芯 Iluvatar CoreX (09903.HK) put its next-gen GPGPU — 天垓300 / Tiangai 300 on the 天枢 (Tianshu) architecture — in front of mainland media around 2026-10-28 in Shanghai. The headline figure, cited from a Shanghai Yidian (上海仪电) third-party run completed earlier (around WAIC July timing), is that under a typical PD-disaggregated (Prefill / Decode) inference layout, normalized per-card end-to-end throughput lands at about ~1.05× NVIDIA H200*. Lab follow-ons add MoE, Agent TTFT, multimodal, and recommendation deltas — all asterisked below. This is a China-market GPU / system story: not Biren 壁砺300, not XHEART (already covered), and not DF1000’s near-memory / no-HBM path. And it is not a Digi-Key or Western-cloud SKU you order tomorrow.

· ~7 min read · China Hot Tech

Cool steel-blue, electric teal, and warm graphite editorial: dense IXL144-style GPU cabinet with token-flow claims for Iluvatar CoreX Tiangai 300 vs H200*

Illustration: steel-blue / teal / graphite — cabinet + PD-disagg token factory. Distinct from MiniCPM lilac/jade, Xiaomi paper/coral, UniWAM, DF1000 sand/copper, X-Era mint/cream, XHEART. Asterisked labels are vendor or third-party-as-cited figures.

For overseas readers the useful questions are narrower than “did China match H200?” Peak FLOPS tables already mislead. What Iluvatar is selling into domestic buyers is usable tokens under a specific deployment — PD separation, long-context Attention/MoE, Agent first-token latency — plus a cabinet density story (IXL144) and a hetero decode story (AFD with partner XPUs). SCMP summarized the company line as matching / beating H200 on live multimodal demos; Chinese trade press (21经济网, 雷峰网, 快科技) carried the Shanghai Yidian 1.05×* figure and the lab deltas. Treat every number as conditioned on unpublished full methodology until someone else re-runs it.

What shipped (and what was shown)

Per company framing and Chinese coverage of the media open day:

Can you use it today?

As a China domestic system / partner deployment — that is the path Iluvatar describes (cards, 8-GPU boxes, IXL144 cabinets, software stack with Agent/Skills tooling, plus TY1200 edge modules). As something an overseas engineer can buy on Digi-Key, AWS, Azure, or GCP tomorrow — no. There is no public Western-cloud SKU, no hobbyist board listing, and export / procurement realities for China GPGPUs still dominate. Watch this for competitive pressure, software-stack maturity, and how Chinese buyers mix GPU + XPU — not as a drop-in H200 replacement in your US/EU cluster.

Four-panel steel-teal editorial: Tiangai 300 card, IXL144 cabinet, AFD GPU+XPU split, TY1200 edge — plus PD-disagg token flow

Card → IXL144 → AFD hetero → TY1200 edge. Asterisks mark vendor / lab / third-party-as-cited benches.

Benches* (third-party-as-cited + lab)

Prefer primary Chinese reporting over secondary blogs. Key cited figures:

What changes for an overseas stack

Caveats

Bottom line: Tiangai 300 is Iluvatar CoreX’s bid that a domestic GPGPU can compete with H200-class end-to-end inference under PD-disagg*, then extend that story into dense cabinets, GPU+XPU decode, and dense-TOPS edge boxes. Overseas readers should track the methodology gaps and the China-only procurement reality — useful competitive intelligence, not a cart-checkout SKU.

Sources