inari@piefed.zip to Technology@lemmy.worldEnglish · edit-222 days agoDeepSeek ditches Nvidia for Huawei chips in V4 launchcybernews.comexternal-linkmessage-square81fedilinkarrow-up11arrow-down10
arrow-up11arrow-down1external-linkDeepSeek ditches Nvidia for Huawei chips in V4 launchcybernews.cominari@piefed.zip to Technology@lemmy.worldEnglish · edit-222 days agomessage-square81fedilink
minus-squaregens@programming.devlinkfedilinkEnglisharrow-up0·edit-222 days agoLLMs are limited by memory bandwidth much more then calculating power. You need HBM. Dedicated accelerators only lower power usage.
LLMs are limited by memory bandwidth much more then calculating power. You need HBM. Dedicated accelerators only lower power usage.