Weekend Discussion: How Concerned Are You If Your CPU Is Completely Open?

[cynic]

I'd pay double for something completely open.

Not sure if up to double, but I'd pay significantly more for an open hardware as well.

[Tim Blokdijk]

I suspect that 'real' use of cryptocurrency will make open hardware more of a thing. Once losing your "privacy" directly translates to losing your money.. people will care.

[coder]

From that comment dimmers hope that the ...

I'm not familiar with the verb form of "dimmers". Is that what you meant to write?

[coder]

NUVIA = new VIA ?

The new VIA seems to be Zhaoxin.

[birdie]

An open CPU is a myth unless you personally control everything: design, masks, production, distribution - at every point someone may interfere and make your design contain the features you didn't quite think about ... um, backdoors.

[coder]

We live in a world where digital privacy is virtually non-existent, devices are back-doored en masse
https://en.wikipedia.org/wiki/Clipper_chip

How does that prove anything?

The 1990's called... it wants its issue back.

[milkylainen]

He is absolutely right.
I've been watching the "I want free and open hardware" community for over 20 years now.
(I actually started investigating free and open CPU designs 20 years ago).
Nothing, absolutely nothing has come out that was worth an iota.
OpenRISC, Sparc, Power, etc, etc. All free ISA's. Dead. And yeah. Power is dead to me aswell.

[Space Heater]

hardware and CPUs should be open for the many reasons we prefer open source, fixing bugs, security, making changes, innovation you name it. And yes PowerPC is dead, it was never that great, and the more modern AND OPEN RISCV will dominate the near and mid term future ;-) And in the wake of all the Intel & co CPU bugs, who would not want a truly open, verifiably and fixable CPU design?

RISC-V's openness only applies to the ISA, there is no requirement that the internal implementation of a RISC-V processor is open. In fact, it is highly unlikely that high performance RISC-V chips will have an open microarchitecture, given the cost of development and the culture of hardware companies. You simply cannot make any strong claims about RISC-V's immunity/vulnerability to hardware security bugs, given the myriad of implementations it has and will have.

Also, how is OpenPOWER less open than RISC-V in any meaningful way?

[milkylainen]

An open CPU is a myth unless you personally control everything: design, masks, production, distribution - at every point someone may interfere and make your design contain the features you didn't quite think about ... um, backdoors.

You're right. But having the actual implementation and RTL design verifiable by your own eyes _significantly_ reduces any opening for easy manipulations and backdooring. This is why China and Russia is rushing to produce their own verified implementations of even x86 cpus.

[milkylainen]

Also, how is OpenPOWER less open than RISC-V in any meaningful way?

You're right. We also have Sparc and OpenRISC f.ex.
Also. A free ISA implementation like the Sifive cores are not guaranteed to have less backdoors than your average x86 implementation.
ISA != Implementation which seems to be lost on a lot of people.

It's like saying that arbitrary cryptography implementation is safe because the theory behind the math looks sound.