# Enablement for Domestic and Non-NVIDIA Accelerators

> We adapt the inference stack at source level and publish a separate measured report for each platform: 8 × AMD MI308X, 8 × Huawei Ascend 910B and MetaX N260. Figures from different platforms are labeled separately and never merged into one headline.

URL: https://vuke.top/en/solutions/domestic-compute · Updated November 12, 2026

Accelerator enablement

## The problem

Buying domestic or non-NVIDIA accelerators is only the first step. Inference frameworks must be rebuilt, cache-library read paths rewritten and multimodal operators verified one by one. Until someone does that work, the cards are just hardware in a rack.

## Our approach

We adapt the inference stack at source level and publish a separate measured report for each platform: 8 × AMD MI308X [R1](https://vuke.top/en/evidence/R1), 8 × Huawei Ascend 910B [R9](https://vuke.top/en/evidence/R9) and MetaX N260 [R5](https://vuke.top/en/evidence/R5). Figures from different platforms are labeled separately and never merged into one headline.

## Key points and sources

- Source-level builds and patches of vLLM / LMCache on ROCm: the parallel-read patch improves TTFT by 4.1× [R1](https://vuke.top/en/evidence/R1)
- Measured on several platforms: R1–R4 on MI308X, R9 on Ascend 910B, R5 on MetaX N260
- Multimodal works as well: all 7 ComfyUI + LTX-Video 2.3 models run on MI308X / ROCm 7.2 with no operator or kernel errors [R6](https://vuke.top/en/evidence/R6) [R7](https://vuke.top/en/evidence/R7)
- Delivered down to the code: git patches, load client, orchestration and forensic scripts, and the raw data [R8](https://vuke.top/en/evidence/R8)

## Related test reports

- [R1](https://vuke.top/en/evidence/R1) · FX100 Comprehensive LLM Inference and Training Benchmark (8 × AMD MI308X)
- [R5](https://vuke.top/en/evidence/R5) · FX100 KV Cache Benchmark (14B, HBM Efficiency)
- [R6](https://vuke.top/en/evidence/R6) · ComfyUI + LTX-Video 2.3 Full-Model Deployment, Adaptation and Usage Report (V2)
- [R7](https://vuke.top/en/evidence/R7) · ComfyUI + LTX-Video 2.3 Model Adaptation Report (AMD MI308X)
- [R8](https://vuke.top/en/evidence/R8) · FX100 KV Cache Test Code Export and Raw Benchmark Data
- [R9](https://vuke.top/en/evidence/R9) · FX100 on Huawei Ascend 910B: Inference and Training Benchmark (vs. NFS Baseline)

## Related products

- [FX100](https://vuke.top/en/products/fx100): The workhorse with the most complete benchmark record

## 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.

- **G1 · Week 2 · Delivery acceptance**: All materials delivered and inspected; every array passes its power-on self-test with all 24 drives recognized.
- **G2 · Week 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.
- **G3 · Week 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.
- **G4 · Week 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.