News for hackers. Stuff that matters.

A comment on DeepSeek could represent Nvidia CEO Jensen Huang's worst nightmare. Parent comment

Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by thefounder on Tuesday January 28, 2025 @08:33AM (#42852089)
It looks like that’s not the case anymore. That’s what this is all about.
[ Parent | on HN ]
Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by senko on Tuesday January 28, 2025 @08:47AM (#42852223)
It still is. The DeepSeek R1 is 670B parameters, requiring more than 300GB of video ram (or unified memory), even if using 4-bit quantized models.
[ Parent | on HN ]
Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by amazingamazing on Tuesday January 28, 2025 @08:55AM (#42852310)
How much video RAM was needed for similar performance a year ago?
[ Parent | on HN ]
Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by senko on Tuesday January 28, 2025 @09:21AM (#42852584)
Roughly the same, as our collective goalposts have shifted, and are still shifting.

What was amazing output last year is today's slop.

NV is in no immediate danger; medium/long term, to anyone knowing what "75% gross margin" means it's no secret that there will be serious threats. You don't need DeepSeek for that particular realization.

[ Parent | on HN ]
Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by amazingamazing on Tuesday January 28, 2025 @09:29AM (#42852691)
Which open source model last January gave you similar results with the same amount of video ram?
[ Parent | on HN ]
Re:DeepSeek could represent Nvidia CEO Jensen Huang's worst ... by talldayo on Tuesday January 28, 2025 @09:48PM (#42860931)
Nomic Falcon, Mixtral, YALM, Bloom, Grok-1 and Jamba, just to name a few of the foundation models.
[ Parent | on HN ]