HNHacker News
TopNewBestAskShowJobs

Zagitta

493 karma · joined May 25, 2015

submissionscomments
Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
I wouldn't hold my breath for 50 series, hopper & blackwell has a different boot architecture with more security layers
Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
These firmwares are already fully encrypted but nvidia has been publishing encrypted debug images that turns out could be decrypted with one of their public test keys. And those debug firmwares turned out to be close enough to the production ones that they could be used to construct exploits from.

In this case it's a DMA buffer overflow that writes to the stack which in turn can be exploited with Return Oriented Programming. So no unsigned code is ever executed but we still have code execution control.

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
Maybe they got hold of the vbios for those higher vram versions similar to the 48gb 4090s?
Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
Hopper and blackwell have a very different security chain so those are out of scope for this approach at least.

But I would suspect China has figured something out given the reports of them buying 5090s in droves and seemingly modifying them.

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
If they are forked from my repo whatever driver version they pin doesn't matter, the pin is just clanker stuff really. What it's used for is just extracting the right booter firmware and loading it, but the whole booter part is reset before handing over to nvidia driver so the versions dont have to match at all.
Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
Blackwell and hopper have a whole new boot process so those are unlikely to fall over anytime soon from what I've seen.

If you can't wait for 4090 someone has a fork with a seemingly working unlock :D

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
Yeah I hate the patched drivers too, I have a CMP170 as well I want to see if I can find a way to get fully working the same way without or minimal driver patching.

4000s is WIP, forks has already unlocked it but I'm doing some refactoring to better support different families and after that I'll port it over :-)

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
It's quite nuanced. In a nutshell the unlock just flips the 2 bits that halves how often the core can issue the specific instructions Nvidia locks to full rate on the professional and datacenter cards.

Obviously doing more compute requires more power but what it does is shift the card down on the frequency curve to keep the wattage in check. So you're right that at stock power limits it hits a cap where the full issuing rate isn't reached but it depends heavily on the exact workload how much juice is left.

The frequency curve also takes heat into account so in general you end up with a nice perf boost for a slightly hotter card.

I've been running this on my 2 3090s for a month now with no issues but they are also water cooled :-)

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
It's also useful for prefill in inference, VLLM sees up to 30% boost depending on the exact kernels and prefill size
Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
Hey this is my work, although the exploits etc aren't my ideas by any means. This is just a way to wrap the whole thing up in a much less annoying package than having to patch the nvidia driver. I originally got inspired by the CMP90/50 windows unlocks that run as standalone UEFI modules that chainloads into the OS.

Happy to answer questions!

Zagitta··on GPU-unlock: FP/BF16 instruction issue rate on Nvidia 3000's cards
It's the FP/BF16 FMA to FP32 instructions.

And also it does work on most 3000 series cards as they all run the same GA102 booter that's vulnerable :-) There are some reports of cards with weird vbios layouts having issues but i'll try to get that fixed

Zagitta··on Elon Musk Just Gave Europe a Reason to Replace American Technology
Musk constantly LIES, a tweet from him means literally nothing
Zagitta··on Gemini 4 Argon
Yeah that was abundantly clear when Bjarne Stroustrup published "A call to action: Think seriously about “safety”; then do something sensible about it"[0] as a reaction to NSA's recommendation to no longer use C/C++.

[0] https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2023/p27...

Zagitta··on ASML says it sold 'absolutely nothing' in Europe in 2026
So now we're suddenly talking about living standards and not environmental regulation?

Your first 3 countries are also incredibly poor comparisons for living standards as they have horrendous work/life balance.

Zagitta··on ASML says it sold 'absolutely nothing' in Europe in 2026
Everyone loves a good race to the bottom with the environment
Zagitta··on Samsung is expected to more than double output of its HBM4 and HBM4E DRAM
Racing to idle is a very common power optimization technique
Zagitta··on Suspected sabotage causes major Netherlands rail disruption
Farmers and right wing people in all of Europe are actively being disinformed by Russian intelligence service and either wilfully ignores it or too incompetent to realize it.
Zagitta··on Apple Watch Series 12
I mean it's nonsense, apple can update the "secure exclave" software at any time to exfiltrate the audio if they want or get ordered to.

If you read that quote carefully you'll also notice it's only the audio that's ever mentioned as inaccessible to apple etc but never mentions anything about the actual transcript, wanna bet that goes into their little advertisement venture?

Zagitta··on Apple introduces M6 and M5 Ultra
There are kernel patches to enable P2P communication via PCIE on 3090 which is almost as fast as NVLink for vllm
Zagitta··on Apple announces changes for apps in the European Union
yet https://keepandroidopen.org/
Zagitta··on Apple's App Tracking Transparency treated its own apps better than rivals
Well you could argue all bundled OS apps should have a dialog similar to the browser selection Microsoft was forced to implement.
Zagitta··on Apple's App Tracking Transparency treated its own apps better than rivals
Regulations for user privacy shouldn't be conflated with regulations for market competition.

I also delight in this highlighting Apple's hypocrisy

Zagitta··on European digital ID wallets rely on safety services of Google and Apple
Can you imagine the collective screeching, across the White house, HN and Apple reality distortion field, that'd happen if EU attempted to breakup the American monopolies?

Electing to not do something impossible and framing it as a surrender is strange to me.

Zagitta··on Apple reveals new AI architecture built around Google Gemini models
This is moving goal posts, DMA is not requiring Apple to run anything on their cloud but merely provide third parties with same device API access.
Zagitta··on Apple reveals new AI architecture built around Google Gemini models
Ah yes something a trillion dollar tech company definitely is too inept to solve!
Zagitta··on Apple ignores DMA interoperability requests and contradicts own documentation
Why is it you feel the need to defend a trillion dollar company?
Zagitta··on Fix the iOS keyboard before the timer hits zero or I'm switching back to Android
Because thirdparty keyboards are equally bugged. Sometimes keyboards simply fail to show up or cover app content without rhyme or reason. This happens across multiple thirdparty keyboards so it must be an ios issue.
Zagitta··on Fix the iOS keyboard before the timer hits zero or I'm switching back to Android
It's hardly surprising when a large percentage of HN users will whiteknight a trillion dollar company whenever EU tells them to stop anti-competitive practices.
Zagitta··on Fix the iOS keyboard before the timer hits zero or I'm switching back to Android
Because thirdparty keyboards are equally bugged. Sometimes keyboards simply fail to show up or cover app content without rhyme or reason. This happens across multiple thirdparty keyboards so it must be an ios issue.
Zagitta··on iOS 26.3 and macOS 26.3 Fix Dozens of Vulnerabilities, Including Zero-Day
Tell that to my keyboard that doesn't show up half the time and the other half covers random parts of the app
Page 1 of 3Next →