General > General Technical Chat
Linus Sebastion Turned Down $100M Offer
<< < (18/35) > >>
hans:

--- Quote from: wraper on August 16, 2023, 07:32:21 pm ---SCR latch-up does not happen because you run the chip at voltage somewhat higher than normal. It generally happens with large transients on I/O, like ESD discharge. Not due to running the chip at 0.15V higher than its safe maximum or exceeding its thermal limits as was with faulty BIOSes. You cannot fix an SCR latch-up susceptibility by applying a BIOS update.

--- End quote ---

A transient such as ESD is certainly not the only cause. Try to blow up a chip just from applying some power onto a chip with ill defined LV inputs :-/O CMOS chips specify all inputs to be e.g. VCC+0.3V max. If VCC=0V, then keep everything below 300mV. I did with a FT4232H some time ago.
Modern multi-rail chips that operate near 1V still demand 3-5% tolerance, including ripple. Max may be only a few hundred mV over. 150mV is 12% of Vsoc nominal, which is a mile length. Give some latch-up path a small push, and it will be self-sufficient till something fails. I don't expect you need 700mV of an ancient 1N4148.. could be way less.

I agree you cannot fix a susceptibility of the silicon, but you can fix the issue by enforcing a more strict SOA in e.g. BIOS or VRM firmware. That's what AMD did and their quick fix was a hard voltage cap, but I suspect there was also some sequencing timing/ramp up violation somewhere that was compounded by the high DC voltage. I also cannot say for sure what happened, but the failure mode of the CPU consuming an enormous amount of power while being "off" and then burning itself up in a runaway situation does exhibit a lot of the symptoms of a latch-up.

Anyhow, this is quite off-topic so I will leave it at that.
wraper:

--- Quote from: hans on August 16, 2023, 11:05:30 pm ---I also cannot say for sure what happened, but the failure mode of the CPU consuming an enormous amount of power while being "off" and then burning itself up in a runaway situation does exhibit a lot of the symptoms of a latch-up.

--- End quote ---
If you're talking about the GN video, they fried it by running at high SOC voltage for a prolonged time. What you're talking about was not while being off, it was motherboard partially shutting down once failure happened but still applying SOC power to already dead CPU. And that's not a latch-up situation, it's already fried silicon. By such definition of a latch-up, any failure with a shorted die would qualify as such. That runaway with socket pins frying was not due to CPU though but due to motherboard still pushing power into a shorted CPU even though reasonably it should trip OCP protection.
Shonky:
All the Youtubers are piling on now. Never miss a chance for views.

And the ex social media manager has come out again with a massive rant....
EEVblog:

--- Quote from: wraper on August 16, 2023, 04:38:12 pm ---There is very little money at TikTok, top 7 TikTokers combined supposedly earn about $50M annually. And the money TikTok itself pays per view is like 2% of Youtube.
--- End quote ---

That's because the views aren't real. Shh, don't tell anyone, you aren't supposed to know.
EEVblog:

--- Quote from: Shonky on August 17, 2023, 01:34:52 am ---And the ex social media manager has come out again with a massive rant....

--- End quote ---

For those playing along at home:
https://twitter.com/suuuoppp/status/1691693740254228741?s=20
Navigation
Message Index
Next page
Previous page
There was an error while thanking
Thanking...

Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod