Xinyuanyi (Tianjin) Semiconductor Equipment Co., Ltd.
About Xinyuanyi
Xinyuanyi (Tianjin) Semiconductor Equipment Co., Ltd. works on one problem: getting more tokens out of the accelerators you already own. We add an external all-flash tier beyond GPU memory and host RAM that catches the KV cache overflowing during inference, so it can be read back when needed instead of being recomputed by the GPU.
Updated
What we do
The product is the FX series of all-flash NVMe-oF arrays, developed on Guangdong Luisuan GP-series hardware and delivered with a vLLM + LMCache KV cache tiering stack that includes our own LMCache parallel-read patch.
We only publish numbers we have measured. Nine reports, R1–R9, cover three platforms: AMD Instinct MI308X, Huawei Ascend 910B and MetaX N260. Figures from different platforms are labeled separately and never mixed. The patch, orchestration scripts and raw data are packaged as R8 so that third parties can reproduce the results independently.
How we treat data
- 01
Every number needs a source
Key figures on this site carry a report ID. If we cannot attach an ID to a number, we do not publish it.
- 02
Platforms are reported separately
Results on MI308X, Ascend 910B and MetaX N260 are documented and labeled individually, never merged into a better-looking headline.
- 03
Baselines are stated
A speed-up means nothing without its baseline. Comparing with a local NVMe drive and comparing with full recompute are two different denominators, and we do not mix them.
- 04
Reproduction material is handed over
The patch, scripts and raw data are packaged as R8, so third parties can rerun the tests instead of trusting our charts.
- 05
Limits are disclosed
Short, low-concurrency conversations benefit little. We say so before you start a project.
Test jointly first, decide afterwards
Every engagement starts with a gated joint test of about 10 weeks. Four gates are passed in sequence; if one is missed, the project stops there, so you never pay for a benefit that was not demonstrated.
G1Week 2
Delivery acceptance
All materials delivered and inspected; every array passes its power-on self-test with all 24 drives recognized.
G2Week 4
Single-node baseline
The single-node 8-GPU inference baseline is reproduced within ±10% of the R1 figures; a local stress test of one array reaches 83% line efficiency against the measured anchor.
G3Week 7
Main joint-test gate
KV tiering must deliver both a TTFT reduction of ≥25% and a throughput gain inside the measured +29–40% band; with 8 nodes sharing one array, read-back bandwidth approaches the NIC ceiling.
G4Week 10
Stability and sign-off
72 hours of continuous load without interruption; single-drive and single-link fault injection stays invisible to the workload; a signed joint-test report is issued.
The full costing model is provided as Python source after an NDA, so you can rerun it with your own parameters.
Company details
| Company | Xinyuanyi (Tianjin) Semiconductor Equipment Co., Ltd. |
|---|---|
| Location | Tianjin, China |
| Contact | Wang Limin |
| Phone | +86 131 6181 8898 |
| mingxin@agentmail.toThe inbox is monitored and triaged by AI around the clock |
A note on names: in earlier test-report filenames FX100 appears as AISSD5000, and before that as WS5000 / GP5000; some reports are signed “Mingxin”. They all refer to the same product. This site uses the FX naming throughout (FX100 / FX200 / FX300 / FX400), while report entries keep their original titles so you can cross-check.
Test jointly first, decide afterwards
Test jointly first, decide afterwards
Every engagement starts with a gated joint test of about 10 weeks. Four gates are passed in sequence; if one is missed, the project stops there, so you never pay for a benefit that was not demonstrated.
- Contact
- Wang Limin
- Phone
- +86 131 6181 8898
- mingxin@agentmail.to
The inbox is monitored and triaged by AI around the clock