Graduate student in Taiwan. Leader of KohakuLab. Researcher/Dev in ComfyOrg
Taiwan
Joined May 2021
- Tweets3.8K
- Following1K
- Followers5.1K
- Likes29.5K
Pinned Tweet
TT-VidT is accepted at #NeurIPS2026 🎉
A temporal-decoupled methodology in video pretraining to provide real "motion" info extracted from the video and only requires small scale training to reach high quality results.
Huge thanks to all collaborators, @NVIDIAAI and @spellbrush
琥珀青葉@KohakuLab retweeted
It is interesting that, ppl don't discuss the risk about use exploit to patch things inside GPU firmware may be dangerous but claiming "NV nerf your performance for stability" or so.
BRUH plz check every detail before you discuss anything, NV only nerf "some" specific dtype combination's performance.
1. Tensor core's most heat comes from multiplier, accumulator is very small compare to other part inside tensor core
2. in consumer card, FP16 matmul FP16 accumulation can run in full speed, but FP32 accumulation can only run in half speed. Which means your card can run multiplier in full speed
3. in 50-series, Mxfp8 and NVfp4 can run FP32 accumulation in full speed. which means your accumulator can run in full speed. (those 2 dtype even have more multiplier and need to handle the block scale lol)
4. so there is NO reason such nerf is designed for stability (the difference between 3090/3090ti can be about quality or stablity, some flaw in chip make NV need to block some SM and result in weaker version). And Usually the raw power of pro card is even worse than consumer card (like the SM's frequency, vector FP32 speed, if you compare p6k and 5090 you will find this)
The nerf is fully designed for spearate the market of pro card and consumer grade card. and use same design (with firmware cap) is actually cheaper in tape out.
The only risk of such "unlock" is if you don't prepare proper fallback entry inside your grub or UEFI, you may lost your OS. That's all.
The risk is even smaller than use apt to install nvidia driver.(X)
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
Latest news on 4090: fully reproduced the EFI applicaiton path and provided more details in my repo:
xitter.cf/KBlueleaf/status/21146…
Now thanks to community resources like gpu-unlock-ada or gpu-unlock-4080s. I successfully unlocked my 4090 on "Windows" system.
All the details are released in github.com/KohakuBlueleaf/Ko…
Plz check it out!
xitter.cf/KBlueleaf/status/21137…
琥珀青葉@KohakuLab retweeted
Now thanks to community resources like gpu-unlock-ada or gpu-unlock-4080s. I successfully unlocked my 4090 on "Windows" system.
All the details are released in github.com/KohakuBlueleaf/Ko…
Plz check it out!
xitter.cf/KBlueleaf/status/21137…
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
Now thanks to community resources like gpu-unlock-ada or gpu-unlock-4080s. I successfully unlocked my 4090 on "Windows" system.
All the details are released in github.com/KohakuBlueleaf/Ko…
Plz check it out!
xitter.cf/KBlueleaf/status/21137…
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
In the unlock I achieved +80% speed on mmapeak (my modification, which is more accurate, old mmapeak have all zero input which make result unclear)
and +50% speed on pytorch. which means 4090 now nearly catch up AIO version 5090 in fp16/bf16.
It is interesting that, ppl don't discuss the risk about use exploit to patch things inside GPU firmware may be dangerous but claiming "NV nerf your performance for stability" or so.
BRUH plz check every detail before you discuss anything, NV only nerf "some" specific dtype combination's performance.
1. Tensor core's most heat comes from multiplier, accumulator is very small compare to other part inside tensor core
2. in consumer card, FP16 matmul FP16 accumulation can run in full speed, but FP32 accumulation can only run in half speed. Which means your card can run multiplier in full speed
3. in 50-series, Mxfp8 and NVfp4 can run FP32 accumulation in full speed. which means your accumulator can run in full speed. (those 2 dtype even have more multiplier and need to handle the block scale lol)
4. so there is NO reason such nerf is designed for stability (the difference between 3090/3090ti can be about quality or stablity, some flaw in chip make NV need to block some SM and result in weaker version). And Usually the raw power of pro card is even worse than consumer card (like the SM's frequency, vector FP32 speed, if you compare p6k and 5090 you will find this)
The nerf is fully designed for spearate the market of pro card and consumer grade card. and use same design (with firmware cap) is actually cheaper in tape out.
The only risk of such "unlock" is if you don't prepare proper fallback entry inside your grub or UEFI, you may lost your OS. That's all.
The risk is even smaller than use apt to install nvidia driver.(X)
I have 2x2080ti-22G (modded) and I found 2080ti also have similar unlock stuff from community,
I'm trying to make it become initramfs module (easier to recover than the UEFI module design) and see if it works.
UwUb
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
2080ti success
xitter.cf/KBlueleaf/status/21141…
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
Remember to check my latest news on 2080ti!
xitter.cf/KBlueleaf/status/21141…
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
琥珀青葉@KohakuLab retweeted
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
Now I also confirmed that 2080ti unlock is successful with similar approach as the 30-series' method
it reach nearly same speed as 350w 3090(locked) with only 250w in mmapeak and mamf-finder
github.com/KohakuBlueleaf/Ko…
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
For those who don't know:
1. This unlock the "nerfed FP32 accumulator" stuff
2. you can rent a A6000/6000 ada to test the mmapeak to know what I'm saying (or check the FP16 accumulation speed of 3090/4090)
3. no unlock core, no overclock, it just remove the nerf
And as this unlock only happen in tensor core:
1. not help mining, not help gaming
2. only help AI usage (potentially DLSS but not guranteed)
琥珀青葉@KohakuLab retweeted
Confirmed that 3090 unlock is successful
with 360watt power wall, I got 75T -> 135T (measured in Pytorch)
Which is insane
I used github.com/Zagitta/gpu-unloc… with commit a9b0c75
琥珀青葉@KohakuLab retweeted
Something exciting is coming. 👀✨
We've been working on an anime-focused 2B pretrained DINOv3 model!
Our internal evaluations show promising improvements across a wide range of anime-related downstream tasks.
We're now making the preparations for release.
Stay tuned! 🚀
