{
  "id": 12184312,
  "title": "AMD attempts to get ahead of expected RTX Spark launch with Gorgon Halo benchmarks",
  "url": "https://urgent.news/2026/10/05/amd-attempts-to-get-ahead-of-expected-rtx-spark-launch-with-gorgon",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-10-05T15:38:45.000Z",
  "source": {
    "name": "Tom's Hardware",
    "slug": "tom-s-hardware",
    "url": "https://www.tomshardware.com/pc-components/cpus/amd-attempts-to-get-ahead-of-expected-rtx-spark-launch-with-gorgon-halo-ai-benchmarks-company-says-it-has-shipped-over-half-a-million-agentic-pcs-to-date"
  },
  "original_language": "en",
  "account": "Nvidia is poised to unveil RTX Spark devices later this week, following a recent tease and its announced fall release. In anticipation of Microsoft's event on October 7, AMD has presented benchmarks for its new Gorgon Halo chips, which will compete directly with RTX Spark devices. This data comes shortly after the initial Gorgon Halo devices launched, priced at over $7,099.\n\nFor now, there are no gaming or general application performance figures available. However, it is expected that there won't be a significant performance jump compared to the last-gen Strix Halo chips. Gorgon Halo primarily aims to compete with RTX Spark and, to a lesser extent, Apple's larger M-series SoCs, which fall under the agentic PC category. This segment is expanding but remains smaller than the hype surrounding RTX Spark suggests.\n\nAMD highlighted that it has shipped \"10s of millions\" of AI PCs, clarifying that this figure actually represents \"over half a million\" agentic PCs, likely referring to Strix/Gorgon Halo devices. The competition between Gorgon Halo and RTX Spark has largely centered around memory capacity. While RTX Spark devices top out at 128GB of unified memory, Gorgon Halo supports up to 192GB. This higher capacity enables running larger models locally, albeit at a reduced performance level.\n\nAMD provided benchmark results for GLM 5.3 Flash with 320 billion parameters using the Unsloth's UD-IQ4_XS mixed-quantization format, achieving a peak token throughput of 20 tokens per second. Comparatively, the last-gen Ryzen AI Max+ 395 achieved 56 tokens per second. Additionally, AMD shared performance figures for Qwen 3.8 Flash Next, a multimodal MoE model with 125 billion main parameters, 51 billion embedding parameters, and about 4 billion parameters for multi-token prediction, achieving up to 42 tokens per second.",
  "summary": "AMD is getting ahead of an expected RTX Spark launch later this week with a few benchmarks for its flagship Gorgon Halo chip, the Ryzen AI Max+ Pro 495.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}