Local AI · Stretching a Single RTX 4090 24GB Form Over Function
Joined May 2024
- Tweets469
- Following458
- Followers173
- Likes4.1K
Ghost In The Stack retweeted
Dear @nvidia, a friendly ask from a guy running 4 of your 3090s and 1 of 3080 Ti.
The 30 series is 6 years old. You dont make these cards anymore. Maybe its time to hand them over to the community completely.
My post about unlocking the 3090 got 270K views in a day, and the replies are the real story. People asking about the 4090, 3080, 3060 and 2080 Ti. People asking how to do it on Windows. Others listing what the community already patches on these cards: the NVENC session limit, P2P between GPUs.
Thousands of people are digging through your old cards with kernel modules, 1 setting at a time, because they love this hardware and want everything it can give.
Publish what is limited on consumer Ampere and Turing. Ship an official driver switch for older generations. Or just open it up.
I understand the business side. Some people might keep their old cards longer instead of buying new ones. But it would be a huge gesture. You talk about local AI on stage, you ship open models, you sell DGX Spark for home. This would show the care goes all the way down to a used 3090 under someone’s desk.
Give them to us. We will do the rest: benchmarks, guides, models tuned for these cards.
@NVIDIAAI
NVIDIA sold millions of 3090s with 50% of the tensor speed switched off. Not missing. Switched off.
1 kernel module flips it to full, and my BF16 tensor pipe went from 82 to 155 TFLOPS. Nothing else changed on the cards, and the outputs match the stock run byte for byte.
All so the gaming card wouldnt eat into workstation sales. Every 3090 owner ran at 50% for 6 years, so a pro card could look worth the price. Fine, its their business model.
And LLMs are the least interesting place for this. Decode waits on memory, and my engine runs its experts in int8. Still +10% on prefill.
The models that live on BF16 tensor math are everything else. Qwen Image 2.1 and FLUX 3 Klein, where every diffusion step is big matmuls. MiniMax H3, which takes 4 to 7 minutes on my cards. Music and speech models. And training, every LoRA anyone ever trained on a 3090 went through exactly the path that was cut.
Thats what I measure next. This is only the start.
Ghost In The Stack retweeted
That gap between seeing the pace and having the vocabulary to explain it is brutal. It’s like being fluent in a language everyone else is still learning.
Ghost In The Stack retweeted
The core issue is not that SuperGrok Heavy “always included only $100 of Other Models.” When SuperGrok Heavy users first claimed Cursor Ultra, they were given the full Ultra third-party pool — $400, the same Other Models allowance that paid Ultra subscribers get. That is what people actually received in the first cycle.
What changed is the second month. After the user renewed SuperGrok Heavy and the Ultra grant was re-issued, that same Other Models pool was cut from $400 to $100. Support is now treating $100 as if it had always been the Heavy-linked amount, and saying $400 “only applies to paid Ultra.” That rewrite ignores what users were actually given on first claim.
That is the lack of good faith. If the bundled Ultra was always a reduced product, Cursor should have said so at claim time — on the dashboard, in the promo, and in billing copy. Instead, users were shown Ultra, used it as Ultra ($400 Other Models), then hit a lock at ~$155 in the next cycle while being told the $400 figure never applied to them.
Calling this a “dashboard distinction” does not fix it. The distinction was not disclosed when the benefit was first granted. Reducing the pool after renewal, then acting as if $100 was the original terms, is the problem.