the most magical exploit i've ever seen

Not surprised that there are many bugs in intel branch prediction.
High frequency trading code optimization basically shows that the branch system is completely broken.

Again branch prediction exploide?

This is pretty incredible. I wonder how they were able to time when the cpu switched over to kernel mode.

I thought this was going to be new. Yes it's rampant it's bin rampant and apple and Microsoft have bin milking it like nothing I've ever seen in my life.

🗣️🗣️ ETH MENTIONED 🗣️🗣️

normies who dont know: 🤔
schizos who do know: 🧐

stuff been going quite bad for intel as of recent
I really hope nothing more terrible happens (both jokingly and really - we need competition on the market - because of how market works - but I still am a AMD person)

That was very interesting, thank you.
A little context for US-based viewers, the ETH Zürich (spelled out E T H, not like the crypto 😉) is the most prestigious technological University of Switzerland and currently ranked 7th best University world wide.
For Europe it's akin to the MIT prestige-wise, so they really get some of the brightest researchers in the world.

I wonder what intels secret cpu inside some of its cou running the management engine vulnerability will be discovered to have since it can be accessed remotely

Everyone complaining about Intel security vulnerabilities are the same ones complaining about Intel year over year speeds.

Imagine how fast we could make stuff if we didn't have so many malicious people in the world.

Using address arguments for non-lr instructions always felt like a mistake tbh

YEAAH Intel SUCKS (however, almost the same exploits were also applicable for AMD and even ARM). And I sell courses, guys!

Great teaching marred by gay shit.

If it involves branch prediction, then it doesn't affect EVERY Intel CPU. I mean, a 486 doesn't have branch prediction.

There is such thing in ARM architecture - MPU (Memory Protection Unit) which defines memory region accessible by CPU in non privileged mode. Done. No brunch prediction bug or other bug can go beyond this region. SO given MPU is working and have no hw bugs and enabled and OS uses it - no such bugs are possible. Right?

But it is prediction? So altering prediction is not to equal to executing predicted altered brunch?

gud i got amd

ah only smart people use chrome

Eat Zürich?

Is there any fix for this exploit?

Computerphile has a great video on branch prediction.

I still think the xbox 360 RGH is the most magical :)

NOO MY INTEL CORE i5 1334U!! WHY IIINTELLLLLLL

AMD bros just keep winning

branch prediction is truly a hacky and buggy optimization

How much money is TSMC paying you?

....you don't think anyone is an expert in microprocessors? wtf?

It's E-T-H Zürich, not Eeethh Zürich

Basically for Spectre and attacks like these it was the question "What gets reversed if a prediction is wrong?"

Alright. Throw it all away. A return to 6502 is in order.

Were Spectre and Meltdown not mitigated?

"We found a new Ring 0 exploit", meanwhile Im over here laughing because there are ring -5 exploits that let you reflash the microcode and run custom instruction sets all from ring 3.

J Lo gives taco flavored kisses to her Ben.

This is insane.

Psst. Intel knew this. For a while. Think agencies

Clearly nobody is a CPU microarcitecture expert, since we have vulnerabilities in CPU microarcitectures 😂

"C the language that runs all other languages" .. ASM is like 🤨
How old are you - org 0x100 Lulz

I don't have to worry about this bug because my mini pc is Ryzen and my laptop is a Snapdragon ThinkPad.

lololol, laughs in threadripper

calling 'etc' by the name of 'etsy' is amazing. Never heard that before

At the start of the video I understood that I have Intel, what is a race condition and what's branch prediction. I didn't have a clue on how they would combine. The rest I learned from LowLevelEddie. Also somehow Intel's gauntlet is like a list of Bond movies.

I'm now tempted to try to run that in my work pc so I can finally use it as a normal human being, so no updating it for a while.
I have a work's pc more limited than a Internet Cafe in the 2000's and no way I'm doing the same repetitive job for 8 hours each day. And I'm tired of two computers to download/send to personal computer -> send to work computer.

Branch prediction and speculative execution has always been an obvious invitation for security vulnerabilities. Basically, any time you allow the computer to go ahead and do things before you know whether or not it is allowed to do them, you have a problem.
AMD is not completely immune to it, but it does seem as though they have generally tried to engineer to minimize these kinds of conditions. Even then, I think a lot of it will come down to compartmentalizing use of systems. Security should be understood as something arising from tight control over; the operating conditions, hardware setup, and applications to the use case. You have to assume that hardware virtualization and privilege levels is contingent on good faith and that access by bad faith action can compromise it.

"I knew that". Except for the "exploit" part of course ¯\_(ツ)_/¯

kiss 4 u?