XC100 AI workstation, domestic (Stellar)

规格

• Series tagline: Liquid Cooling · Silent Operation供应商自述
• Badge: CN供应商自述
• Variant 1 label: Zhaoxing Solution供应商自述
• CPU (Zhaoxing): Zhaoxin KX-7000供应商自述
• GPU (Zhaoxing): Moore Threads MTT S80供应商自述
• RAM (Zhaoxing): 32GB DDR4 (Expandable to 64GB)供应商自述
• I/O (Zhaoxing): 4*USB3.0 Type-A 2*USB2.0 Type-A供应商自述
• Video Output (Zhaoxing): • 1×DP 1.4 • 1×HDMI®2.0 • 1×VGA供应商自述
• Network I/O (Zhaoxing): • 3×RJ45 2.5G供应商自述
• Variant 2 label: Hygon Solution供应商自述
• CPU (Hygon): Hygon 3450G供应商自述
• GPU (Hygon): MetaX N260供应商自述
• RAM (Hygon): 64GB DDR5供应商自述
• I/O (Hygon): 4*USB3.0 Type-A供应商自述
• Video Output (Hygon): • 1×HDMI®2.0 • 1×VGA供应商自述
• Network I/O (Hygon): • WiFi 6 + BT 5.2 • 4×RJ45 2.5G供应商自述
• ROM: PCIe 4.0 + Dual M.2 2280 (both)供应商自述
• GPU compute: MuXi N260: 280 TOPS (INT8) | Moore Threads MTT S80: 14.7 TFLOPS (FP32)供应商自述
• GPU compatibility: Compatible with MetaX N260 And Moore Threads MTT S80; NVIDIA Pro 6000 / Pro 5000供应商自述
• Headline: Up to 96GB Total VRAM | 4,000 TOPS (Theoretical Peak)供应商自述
• Chassis Volume: 11L 320*136*250mm供应商自述
• Real-World Benchmark: XC100 Mortex N260 Qwen3.6-35B Performance: Model | Inference Framework | Quantization | Single-Concurrency Peak | Multi-Concurrency Peak | Max Context Length | ITL Latency ;; Qwen3.6-35B-A3B-W8A8 / vLLM / W8A8 / 21.60 / 611 / 256K / 19~936ms供应商自述
• Peak throughput under extreme context length (Single-concurrency, output=64). Unit: tok/s: Model | 128 | 256 | 512 | 1k | 2k | 4k | 8k | 16k | 32k | 64k | 128k | 256k ;; Qwen3.6-35B-A3B-W8A8 / 21.60 / 21.58 / 21.32 / 20.72 / 19.99 / 17.93 / 14.70 / 10.28 / 5.78 / 2.54 / 0.89 / 0.28供应商自述
• Multi-Stream Throughput (Single-Stream Ref: input=512, output=64, tok/s): Model | Concur=1 | Concur=2 | Concur=4 | Concur=8 | Concur=16 | Concur=32 | Concur=64 | Stream Peak ;; Qwen3.6-35B-A3B-W8A8 / 21.32 / 86.23 / 140.24 / 189.97 / 278.64 / 346.56 / 376.10 / 376.10供应商自述

型号

XC100

最低起订量

随报价确认

销售单位

unit

这些信息是供应商自述，我们尚未测试。

运输与订购

付款时选择海运或空运。运费、进口关税和税费之后按代理实际成本收取，不计入商品价格。