Maybe, but CPUs are coming out with LLM tuned chips, and I can run a basic model on an i5, 8gb ram, and no dedicated card. It’s not super powerful, but for most users, it’s more than enough for what they use the big models for. Also, if it does take a discrete card to get that needed boost in performance, then at least consumers would be able to get GPUs again.
I think as hardware improve and is further designed around LLM efficiency, and local models are tuned for specific uses and being able to run on lesser hardware, it will make Nvidia obsolete for large swaths of the population. A good GPU will still be necessary for high performance, graphic/physics intense gaming, but that’s a really small subset of all users.
Hopefully Nvidia just shits and has to grovel back to the consumer to get there marketshare back when all of the DCs go tits up.
I can run models on my 3yo midrange smartphone. Gemma-4-E2B totally runs on there. Bonsai-8B too. But neither is really good for most tasks.
Heck, “a basic model” even runs locally on the 8MB RAM of an ESP32-S3, but it’s utterly worthless at anything.
If you get a bit more into self-hosting AI it quickly becomes obvious that for any actually useful real world tasks you need at least 24GB you can dedicate to the LLM alone, and if this is fully GPU vRAM, the performance is way, way higher than on CPU, even with an NPU.
Maybe, but CPUs are coming out with LLM tuned chips, and I can run a basic model on an i5, 8gb ram, and no dedicated card. It’s not super powerful, but for most users, it’s more than enough for what they use the big models for. Also, if it does take a discrete card to get that needed boost in performance, then at least consumers would be able to get GPUs again.
I think as hardware improve and is further designed around LLM efficiency, and local models are tuned for specific uses and being able to run on lesser hardware, it will make Nvidia obsolete for large swaths of the population. A good GPU will still be necessary for high performance, graphic/physics intense gaming, but that’s a really small subset of all users.
Hopefully Nvidia just shits and has to grovel back to the consumer to get there marketshare back when all of the DCs go tits up.
I can run models on my 3yo midrange smartphone. Gemma-4-E2B totally runs on there. Bonsai-8B too. But neither is really good for most tasks.
Heck, “a basic model” even runs locally on the 8MB RAM of an ESP32-S3, but it’s utterly worthless at anything.
If you get a bit more into self-hosting AI it quickly becomes obvious that for any actually useful real world tasks you need at least 24GB you can dedicate to the LLM alone, and if this is fully GPU vRAM, the performance is way, way higher than on CPU, even with an NPU.