Startup Ideas Bank

A clever kernel hack that doubles GPU math speed, but it is a free tool with no business attached and a legal cliff underneath.

AI roast score: 38/100 (F)

The idea

You found a way to unlock hidden speed in Nvidia cards, then gave it away free, with no price, no buyer and no plan. Nvidia can kill it with one firmware update. Great hack. Not a company.

In the founder's own words

Zagitta/gpu-unlock — Unlock FP/BF16 instruction issue rate on Nvidia 3000's cards to their professional equivalents

GPU unlock 1.0.0

Unlock FP/BF16 instruction issue rate on Nvidia 3000's cards to their professional equivalents

This Linux kernel module combines the CMP unlocker's V67/PLM technique with the pre-driver timing of the UEFI approach. It runs the unlock, writes the speed selectors (SS0 0x88888888 , which selects the full BF16 rate, and the 4-bit SS1 0x8 IMLA4) and performs FLR before the unmodified NVIDIA driver binds the card. It is neither a UEFI application nor a patched NVIDIA driver.

The benefit of this early kmod driver is that it does not require manual patching each time nvidia releases a new driver version.

The effect of this unlock is that it doubles the instruction issuing speed of FP/BF16 FMA into FP32 operations on GA102, GA103, GA104 and GA106 cards.

3090 Ti dense 8192³ GEMM at 450 W, five-sample medians with FP32 accumulation :



Input
Before
After



FP16
87.46 TFLOPS
118.39 TFLOPS


BF16
87.50 TFLOPS
157.41 TFLOPS


TF32
42.58 TFLOPS
82.61 TFLOPS



Tested cards

Results reported by users. The module accepts every GeForce ID listed in
kernel/gpu_unlock/main.c , but the cards below are the ones someone has
reported on. Please add yours with a
compatibility report ,
whether it worked or not.



Chip
Card
PCI ID
Result
BF16 GEMM gain
Reports



GA102
RTX 3090 Ti
10de:2203
Unlocked
+68–80%
#1


GA102
RTX 3090
10de:2204
Unlocked
+25–67%
#5 , #9 , #10 , #17 , #25 , #27 , #33


GA102
RTX 3080 Ti 20 GB (OEM, GA102-250)
10de:2205
Unlock completes, BF16 stays at half rate
—
#28


GA102
RTX 3080 10 GB
10de:2206
Unlocked
+70–73%
#14


GA102
RTX 3080 20 GB (memory upgrade)
10de:2206
Unlocked
+56–59%
#10 , #15


GA104
RTX 3070
10de:2484
Unlocked
+73–81%
#10 , #12


GA104
RTX 3070 Ti
10de:2482
Not tested with this module (the same unlock worked from a UEFI port on Windows)
—
#32


GA106
RTX 3060 12 GB
10de:2504
Unlocked
about +80%
#13 , #31


GA103
any
10de:2414
No reports yet
—
—



The gains are dense GEMM throughput from each report's own benchmark, so
they are not directly comparable. The power limit matters a lot: the lowest
figure (+25%) is from cards capped at 225 W.

Matrix-heavy work speeds up: LLM prefill (+14–36% in
#9 ,
#15 ,
#25 and
#33 ), diffusion and
FlashAttention. LLM decode, which is limited by memory bandwidth, and INT8
work stay about the same. The reports that compared outputs found them
bit-identical before and after.

Reports also cover up to four cards in one machine, the card driving the
display, and cards passed through to a VM

The roast

The engineering is real. You show BF16 going from 87.50 to 157.41 TFLOPS on a 3090 Ti, with output reported as bit-identical. That earns respect from developers. It is not a business, though. You picked free pricing, you have not decided how to earn money, and your own deal size is under 10 dollars. Your whole product depends on Nvidia not caring. It undoes a deliberate speed cap that Nvidia uses to separate GeForce from its pro cards. One firmware change, a driver check or a legal letter and the module is dead. The gains are also uneven. One 3080 Ti 20 GB card unlocks but stays at half rate, and the 3070 Ti is untested with this module. Your buyers are hobbyists and small teams running local AI. They tinker and they hate paying. Your plan to grow through articles and search will bring GitHub stars, not paying customers. Being a solo founder means nobody else can handle support for 20+ card variants.

Red flags

Verdict

Decide whether this is a portfolio piece or a business. If it is a business, pick a paying buyer such as small AI labs or GPU renters before Nvidia closes the door.

Start with your own startup idea →

More ideas roasted by AI

Browse all startup ideas →