AMD Publishes Platform QoS Patches For Next-Gen Processors

[phoronix]

Phoronix: AMD Publishes Platform QoS Patches For Next-Gen Processors

This afternoon AMD sent out their first Linux kernel patches for what might end up being a new feature for the "EPYC 2" / Zen 2 processors...

http://www.phoronix.com/scan.php?page=news_item&px=AMD-Platform-QoS-RFC-Patches

[Guest]

My wallet is ready for Zen 2 and Navi.

[humbug]

My wallet is ready for Zen 2 and Navi.

Mine isn't, but I am determined to make it ready in time.

zen2 should be good, I really hope they get the GPU side of things back on track with Navi.

[BoMbY]

They already published a documentation last month: https://www.amd.com/system/files/Tec...Extensions.pdf

You sure this is Zen2 only?

[cybertraveler]

My most important question for ZEN2:

Does it have an always on, all-pervasive, secret processor running code that cannot be easily inspected and cannot be modified?

They call this thing the Platform Security Processor (PSP). Security for who?

[TemplarGR]

My wallet is ready for Zen 2 and Navi.

Mine as well. I am long overdue for an upgrade from Piledriver and Tonga, but 7nm being so close and RAM prices still being ridiculous made me wait...

[DavidKL]

My most important question for ZEN2:

Does it have an always on, all-pervasive, secret processor running code that cannot be easily inspected and cannot be modified?

They call this thing the Platform Security Processor (PSP). Security for who?

Always a pertinent question but the bottom line is is that basically everything of a certain level of complexity is intentionally insecure — hardware and software... The only exceptions are things like NASA probes.

Assume insecurity and then prove security (if you can). That's the motto for our times. People are far too often doing it the wrong way, assuming security then responding with surprise when it's made known to them that the product isn't secure. This latter mentality is lazy at best and intentionally naive at worst. Even without the known fact of there being many people who want to exploit insecurities by managing to get them baked into systems intentionally, bugs are hardly so rare that people should not assume insecurity by default — simply due to the presence of unknown errors. Distrust first, then trust, if able to prove trustworthiness.

Another point is that so many products, especially in tech, are designed to use the "user" these days. Planned obsolescence is the primitive version of that but things have become more sophisticated. IT is called informational technology for a reason. Inform without consent, the growing "telemetry cloud" game. Knowledge is power, so why would any of us want any? It's so much more pleasant to give it all away to our betters. Surely, they have only our best interests at heart, not their own competing interests.

[starshipeleven]

My most important question for ZEN2:

Does it have an always on, all-pervasive, secret processor running code that cannot be easily inspected and cannot be modified?

Do you even need to ask?

The answer is yes, and unless they remove the feature they had in Zen you can probably disable maibox registers in the UEFI (i.e. block any access to the PSP).

They call this thing the Platform Security Processor (PSP). Security for who?

Answered many times already here and elsewhere.

It ensures the client device is secure for content providers and other third parties that need to protect their software or media from potentially malicious consumers.

[starshipeleven]

Inform without consent, the growing "telemetry cloud" game.

Can you please stop calling "telemetry" what is actually tracking or even snooping?

Actual telemetry is good, it sends over only technical data used by developers or system administrators to diagnose issues since most users can't be bothered (or genuinely don't know how) to open bug reports and run debugging when issues are encountered.

It's when the applications are sending over other stuff to track the users, or extract information from the users (snooping), that is bad.

[cybertraveler]

It's when the applications are sending over other stuff to track the users, or extract information from the users (snooping), that is bad.

I also consider it bad if it's done without clear warning or without being granted explicit consent.