[dont__panic]

I think it goes something like this:

- no Management Engine

- chips that don't turbo boost themselves into throttling

- not supporting a company with a toxic approach to business

I believe AMD outperforms Intel when you're targeting mobile performance/battery life, rather than "moar CPU" workloads. Though that might change now that Intel is using their own approach to performance cores. Still, given the last decade of Intel development, they don't exactly have my trust that they'll execute performance cores without serious hiccups.

[fsflover]

> no Management Engine

https://en.wikipedia.org/wiki/AMD_PSP

[gray_charger]

AMD PSP is NOT the same as Intel ME. AMD PSP is a "trusted execution environment" (the first sentence in your link). Intel's equivalent is Intel SGX. Trusted execution environments are a security feature that does not offer remote management. It's not a privacy concern like Intel ME is.

[dont__panic]

Thank you for clarifying this, all of the acronyms are difficult to reason about if you don't work with them every day.

[sascha_sl]

The PSP is a lot slimmer than Intel ME. It also doesn't randomly yank traffic for specific ports from your Ethernet.

[trinsic2]

I havent looked at the presentation yet, but are you saying the PSP, like intels ME could be doing nefarious things since its proprietary and closed? Do you have a link to information on the network capturing thing? I mean is that really a thing?

I have heard of these things before but I am not quite sure what the possibilities are. Do you have a link that can summarize what this actual means in terms of security concerns?

[sascha_sl]

CVE-2017-5689

[google234123]

How do you know there isn't an undiscovered CVE for AMD? There's probably maybe 10x more security research focused on Intel

[sascha_sl]

I don't, but they do a lot less with the PSP, especially if you're just using Ryzen Pro and not server SKUs. Intel put a web interface you can't disable with an offbrand networking gear level RCE vulnerability that needs nothing more than ethernet access into their security chip. I don't think AMD can exceed that anytime soon.

[fsflover]

Any good link with the details?

[sascha_sl]

There was a good talk with an overview (as well as owning it) at 36c3.

https://media.ccc.de/v/36c3-10942-uncover_understand_own_-_r...

[jeffbee]

> - chips that don't turbo boost themselves into throttling

Your level of understanding about how CPUs control their frequency, voltage, and power is evidently "none". Why spread comments like this which only serve to confuse and mislead readers?

[dont__panic]

Intel configured the chips such that they turbo boost so high that they overheat and downclock themselves to compensate.

Still "no" level of understanding? If there's something incorrect about my statements, feel free to correct me -- I do want to learn more, and I'm certainly no expert in CPUs. But it's just flat out rude (and against the contributor guidelines, I believe) to comment like this. Build other people up, don't tear them down.

[jeffbee]

> Intel configured the chips such that they turbo boost so high that they overheat and downclock themselves to compensate.

Yes, this is optimal and what literally everyone wants.

An airplane takes off at full power, reaches cruising speed, and reduces power to maintain cruising speed.

A CPU uses max power until it reaches its max operating temperature, then it maintains that temperature operating at lower power.

Why does the latter offend you when it's exactly the same as the former?

> Still "no" level of understanding?

Sadly, yes.

> don't tear them down.

This conversation started with you tearing down thousands of expert electrical engineers who make Intel CPUs.

[dont__panic]

Your airplane analogy is not what Intel CPUs do in practice.

A better analogy:

An airplane takes off at full power, reaches cruising speed, but its engines have overtaxed themselves and can't maintain altitude. The place descends to a suboptimal altitude until the engines can turn back on, and raise the plane back to the altitude it's supposed to cruise at.

Your CPU explanation is technically correct:

> A CPU uses max power until it reaches its max operating temperature, then it maintains that temperature operating at lower power.

Yep, this is a very high-level explanation of what CPUs do. The trouble with Intel processors today is that they use max power for too long, and have to throttle so heavily to "maintain that temperature operating at lower power" that you can notice the latency when the CPU downclocks. An ideal operating curve wouldn't use max power for so long that it causes obvious latency issues to an end user. That's why I have Turbo Boost disabled on my laptops -- the few seconds of "max power" it yields just aren't worth the massive downclock while the CPU cools down. Better to set a more conservative power level that doesn't get in my way. This is especially noticeable if you use emulation or a beefy IDE like Android Studio that turbo boosts your computer to a high temperature in the first few seconds of use, then turns text editing and code suggestions into a sluggish slideshow for the next few minutes because the CPU has downclocked. Or maybe I'm just imagining that?

> This conversation started with you tearing down thousands of expert electrical engineers who make Intel CPUs.

Did I say anything bad about the engineers? I have lots of disparaging things to say about the way Intel works as a business, mostly based around how product and sales operate. I think the engineers at Intel do the best they can under the constraints of a poorly run company. But there's a reason engineering talent has been fleeing for the better part of a decade.