AI News Feed
Market watch
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
LG UltraFine Evo 32U990A-S Hits Record-Low $1,099 on Amazon, TechRadar ReportsNewegg Labor Day Deal: 32GB Kingston Fury Beast DDR4-3200 Kit Nearly Half Price at $179.99 with CodeGalaxy Z Fold 8 inner screen corners reported with excessive giveHuawei unveils Mate XT 2 tri-fold with privacy-screen edition as Apple prepares folding iPhoneWall Street Analysts Recommend Nvidia, Uber and Marvell for Long-Term GrowthLenovo, AMD and Huawei Put Local Large-Model AI at Center of New Devices
Products & Applications

$16,000 Quad A9 Mega Mini PC Cluster Runs DeepSeek V4 Flash with 512GB RAM

Geekom has deployed DeepSeek V4 Flash across a $16,000 cluster of four A9 Mega mini PCs linked by USB4, delivering 512GB of memory for local AI inference. The vendor-reported performance of 14.61 tokens per second awaits independent verification.

AI News Feed · 2026-08-22 10:14

According to a TechRadar report, the four-node configuration replaces traditional data center servers with desktop-sized units. Each A9 Mega is built around AMD's Ryzen AI Max+ 395 processor, which combines 16 Zen 5 CPU cores with Radeon 8060S graphics and unified memory in one chassis. The systems run Ubuntu with ROCm and DwarfStar software to distribute the optimized model, and an OpenAI-compatible API connects applications and AI agents to the cluster.

Geekom said the local setup keeps prompts, documents, source code, and credentials within an organization's own infrastructure rather than routing them through a public cloud. This enables private knowledge assistants capable of searching contracts, manuals, and internal reports without external exposure. For agent-based systems, the cluster can process tools, policies, memory, and logs before taking action, with a technical path that has handled contexts up to 250K tokens.

Performance testing reported by Geekom and cited by TechRadar shows about 14.61 tokens per second at single concurrency and a P95 time to first token of roughly 0.42 seconds. The company positions the cluster as offering greater capacity for long prompts, not just faster short-response generation. Users can begin with one or two A9 Mega units and expand to four nodes as workloads increase; each unit works standalone or can contribute to distributed inference.

Each A9 Mega supports up to 128GB of LPDDR5x memory, 120W sustained performance, dual M.2 PCIe Gen4x4 storage slots, and up to 8TB of RAID-configured storage. It also delivers 126 total TOPS of combined processing output, Wi-Fi 7, Bluetooth 5.4, dual 2.5G Ethernet, and four simultaneous 8K display outputs. Geekom lists the A9 Mega at $3,999 with 128GB RAM and a 2TB SSD, making the four-system cluster $16,000.

The performance values are vendor-supplied benchmarks without third-party testing, TechRadar noted, so real-world reliability for sustained enterprise use has not been independently confirmed.