# FX100 All-Flash NVMe-oF Array: Specs and Benchmarks

> FX100 has the most complete benchmark record in the FX series: it is the device under test in R1–R5 and R9, across AMD MI308X, MetaX N260 and Huawei Ascend 910B. It offers 100 Gb per port and 16M IOPS, and is sized at one array per 8 inference nodes in production.

URL: https://vuke.top/en/products/fx100 · Updated November 12, 2026

FX100 has the most complete benchmark record in the FX series: it is the device under test in R1–R5 and R9, across AMD MI308X, MetaX N260 and Huawei Ascend 910B. It offers 100 Gb per port and 16M IOPS, and is sized at one array per 8 inference nodes in production.

## Specification

| Specification | Value |
| --- | --- |
| Interface generation | PCIe 3.0 |
| Rate per port | 100 Gb |
| IOPS | 16M |
| Flash form factor | U.2 |
| Controllers and expansion | Dual controllers, 2×3 ports (ASIC) |
| Drive configuration | 24 × 7.68 TB Gen4 enterprise U.2 NVMe SSDs |
| Status | In production |
| Reference price | ¥371,200 |

## How the FX100 reference price is calculated

Fully populated reference price = platform ¥160,000 + 24 × ¥8,800 = ¥371,200 for 184.3 TB raw, about ¥2,014 per TB. SSDs are counted at a reference market price that moves with the market; the purchase contract sets the final price.

## Who it is for

- Teams that want to prove the benefit of KV cache tiering in a joint test before scaling out
- MI308X or Ascend 910B inference clusters that already run 100 GbE RoCEv2
- Existing clusters held back by NFS, where model loading and switching take too long

## Related test reports

| Metric | Value | Baseline | Detail | Source |
| --- | --- | --- | --- | --- |
| Higher inference throughput | +29–40% | vs. a single local NVMe drive | Qwen3-Coder-480B-FP8 in TP=8 production topology under a long-context cold-recovery load: +29% at concurrency 8 and +40% at concurrency 16, the operating point with the largest gain; +35–36% for the whole machine in the dual-instance TP4×2 layout. | Measured · [R2](https://vuke.top/en/evidence/R2) [R3](https://vuke.top/en/evidence/R3) |
| Lower time to first token (TTFT) | −26–32% | vs. a single local NVMe drive | Across three concurrency levels of the same 480B TP8 load, TTFT p50 falls from 10.17–35.73 s to 7.53–26.35 s. | Measured · [R2](https://vuke.top/en/evidence/R2) |
| Faster than recomputing without external storage | 8.6–20× | vs. full recompute with no external storage | At concurrency 16, TTFT p50 is 149.5 s when recomputing and 11.85 s with FX100; throughput is 4.1 tok/s versus 74.9 tok/s. | Measured · [R2](https://vuke.top/en/evidence/R2) |
| TTFT gain from the LMCache parallel-read patch | 4.1× | vs. before the patch | One GPU, concurrency 16, cold read from disk (Qwen2.5-32B): TTFT 37.97 s → 9.30 s; read bandwidth 0.98 → 5.23 GB/s, about 5.3×. | Measured · [R1](https://vuke.top/en/evidence/R1) |
| Faster model loading | 6.2–9.3× | vs. NFS, on Huawei Ascend 910B | On Huawei Atlas 910B: DeepSeek-32B serving load 691 s → 112 s (6.2×); DeepSeek-70B 1399 s → 150 s (9.3×). | Measured · [R9](https://vuke.top/en/evidence/R9) |
| Faster training checkpoint saves | 1.9× | vs. the baseline in report R1 | 8-GPU 32B LoRA with 65.6 GB full-model snapshots: 178 s → 94 s; sustained write bandwidth 3.26 → 6.40 GB/s (+96%). | Measured · [R1](https://vuke.top/en/evidence/R1) |
| Effective compute utilization under model switching | 46.7% → 62.8% | at 20 model switches per hour | Modeled from measured load times: at 20 model switches per hour, effective compute utilization rises from 46.7% to 62.8%. | Measured · [R1](https://vuke.top/en/evidence/R1) |
| Concurrent model loading on 8 GPUs | 16% / 30% | vs. the baseline in report R1 | Eight GPUs cold-reading models at once: 16% faster wall-clock overall and 30% faster per model. | Measured · [R1](https://vuke.top/en/evidence/R1) |
| Deviation between two independent runs | ≤5% | same load level, measured twice | Two independent runs at the same load level deviate by no more than 5%; cross-instance hot sharing through the fs:// shared pool was verified in the same campaign. | Measured · [R3](https://vuke.top/en/evidence/R3) |
| HBM residency an external KV tier can stand in for | 128 GB | MetaX N260 single GPU, 14B model, seven-way comparison | Conclusion of a seven-way comparison on a 14B model: an external KV tier can stand in for 128 GB-class HBM residency. That campaign ran on a single MetaX N260 GPU; the method carries across platforms, the numbers are not mixed with MI308X results. | Measured · [R5](https://vuke.top/en/evidence/R5) |
| Arrays to inference nodes | 1 : 8 | production sizing basis | One FX100 serves 8 inference nodes; 100 GbE sustaining KV read-back for 8 nodes is a measured anchor. On the joint-test basis, storage spend stays within roughly 10% of the GPU budget. | Estimated · [R1](https://vuke.top/en/evidence/R1) [R2](https://vuke.top/en/evidence/R2) [R3](https://vuke.top/en/evidence/R3) |

- [R1](https://vuke.top/en/evidence/R1) · FX100 Comprehensive LLM Inference and Training Benchmark (8 × AMD MI308X) (8 × AMD Instinct MI308X · ROCm 7.2)
- [R2](https://vuke.top/en/evidence/R2) · FX100 KV Cache Benchmark (480B, TP8, Long Context) (8 × AMD Instinct MI308X · ROCm 7.2)
- [R3](https://vuke.top/en/evidence/R3) · FX100 KV Cache Benchmark Summary (480B, TP4×2, All Metrics) (8 × AMD Instinct MI308X · ROCm 7.2)
- [R4](https://vuke.top/en/evidence/R4) · FX100 KV Cache Benchmark (480B, Multi-Instance, No. -006) (8 × AMD Instinct MI308X · ROCm 7.2)
- [R5](https://vuke.top/en/evidence/R5) · FX100 KV Cache Benchmark (14B, HBM Efficiency) (Single MetaX N260 GPU)
- [R9](https://vuke.top/en/evidence/R9) · FX100 on Huawei Ascend 910B: Inference and Training Benchmark (vs. NFS Baseline) (8 × Huawei Atlas 910B (Kunpeng-920))