Nvidia Brings Day-One DiffusionGemma Support to RTX and DGX Nvidia为RTX与DGX首日适配DiffusionGemma
Nvidia has added launch-day support for Google DeepMind's DiffusionGemma open model across its RTX and DGX platforms, with DGX Spark reported at 150 tokens per second. The release shows how quickly accelerator vendors are turning model-specific software optimization into a hardware adoption tool. Nvidia已在RTX和DGX平台首日支持Google DeepMind的开源模型DiffusionGemma,报道称DGX Spark可达到每秒150个token。此次适配显示,加速器厂商正把针对具体模型的软件优化转化为推动硬件采用的重要手段。