Google Analytics counts our visits, with cookies. No thanks turns it off and removes its cookies; in Europe, the UK, Switzerland and India, or wherever we cannot tell where you are, it stays off unless you accept. Cookie Notice

Advertise
PRICES
BTCETHSOLXRPBNBTONDOTLTCPOLHBARNEARARBUNI
Back to NewsFeeds

Nvidia says co-designed memory stack lifts inference throughput up to 30-fold

NewsFeeds update

Nvidia says rising HBM costs and tighter supply are pushing it to co-design GPUs, CPUs, networking, storage and software for memory efficiency rather than adding HBM capacity. It says Vera Rubin NVL72 reaches up to 30 times the cost-per-throughput of GB300 NVL72 with only about 3% more HBM per GPU.

AIOdailyAINeutral

Read the original on Odaily

More AI updates

  1. Police in Oakland investigate theft of Nvidia AI chips worth $11.4m

    AIPANewsAINeutral

  2. Chamath Palihapitiya says SpaceX-Anthropic compute contract holds 90-day mutual exit clause

    AIChainCatcherAINeutral

  3. Stop Dirty Data Centres says protests across Europe target AI data centre expansion

    AIChainCatcherAINeutral

Every AI update on NewsFeeds

Every update is a short rewrite, by AI in our own words, of a flash from the outlet named on it, credited and linked. Not investment advice.