Get your royalty-free soft-core OpenPOWER processor core blueprints here. Extra, extra – read all about it

Here are a couple of chip-related news bytes for you. An IBM engineer has released the first soft-core implementation of an OpenPOWER CPU since Big Blue's decision to open up the specification, and Intel says it is "shipping" its first 10nm Agilex FPGAs. Watt the heck? Earlier this month, IBM announced it will follow in the …

[Christian Berger]

Probably a pass for most

The Power architecture is not particularly elegant, performant or has particularly well compiler support, so most people will go for RiscVI instead.

However it still is a positive step, though it feels like IBM is trying hard to keep their Power architecture relevant with this. The times when the fastest Microprocessors were Power ones are long gone, and today the fastest amd64 processors outperform the fastest Power ones by a factor of 2.

https://www.phoronix.com/scan.php?page=article&item=rome-power9-arm&num=1

However it'll be nice for people doing FPGA reimplementations of Macs.

[whitepines]

Re: Probably a pass for most

RISC-V, the ISA that still doesn't have vector instructions? The ISA that you have to keep various compilers around for due to the incompatible silicon between vendors? The ISA that has its best cores topping out below even decent ARM cores in performance?

Does anyone remember the DOS days when software ran on some PC clones but not others, and everything was a fragmented mess? I suspect the current RISC-V proponents do not, or they wouldn't be pushing things like "modular ISA" or "fork and make proprietary silicon for vendor-specific software" as positives.

I also take some issue with Rome being used as the comparison. Its true competition is Power10, not the older Power9 CPU that is currently somewhere in middle age. The releases have been staggered to make AMD look good, and it *is* fundamentally a good core on a leading manufacturing node, but it has two serious drawbacks:

- It implements the closed, proprietary, x86_64 instruction set

- AMD forces you to blindly trust their massive, binary only firmware, which exists primarily for DRM purposes and to fundamentally limit what you can do with the system (a digital nanny)

I may be in the minority, but those two issues alone are deal breakers for me. Why would I want to work with something as closed and locked as the x86_64 ISA when I don't need to? So that leaves ARM, RISC-V, and now Power as fairly available candidates -- of those three, I can buy good enough CPUs for daily work that implement ARM and Power. The ARM ISA requires licensing and ARM can deny access if they want for basically any reason (ask Huawei how that's working out for them!), so basically that means Power is most interesting to me right now.

By all means, downvote me to oblivion for daring to say anything bad about Team Red. They became another Intel years ago, and the fan base just won't see it. Must be the same human factors at work that keep people frothing at the mouth defending a mediocre sports team, or more to the point one that keeps cheating.

Icon 'cause that's where we go if people choose to keep Intel/AMD on top despite the decent alternatives that have appeared in the past 5 years.

[Bronek Kozicki]

Re: Probably a pass for most

Speaking of AMD, I wonder if we will see (before 2030) a POWER chip from AMD. It would be an interesting move in taking the fight to Intel.

[whitepines]

Re: Probably a pass for most

Speaking of AMD, I wonder if we will see (before 2030) a POWER chip from AMD.

If they'd position it as a true server chip and get rid of the bloody PSP for those SKUs, plus release enough documentation for coreboot to be ported to it, then I'd be all for it. My suspicion is that the DRM contracts bind the entire company though, not just the x86 CPU business, so this is all just fantasy. ISA doesn't matter when the USA vendor still has control of your data and system via the CLOUD act, closed signed firmware, and anything that acts even remotely like the PSP.

[Korev]

Re: Probably a pass for most

Thanks for the interesting link

[whitepines]

Re: Probably a pass for most

Here's another interesting link:

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

Much like when the USA introduces laws with names like the "Patriot Act", you can be sure the actual intent of the "Security" processor here is exactly opposite of what the name says. Oh, and it's completely unremovable, it's your personal nanny on Epyc, Ryzen, and Rome whether you want it or not.

And with that, I've said my piece. Flame away!

[dajames]

Re: Probably a pass for most

... you can be sure the actual intent of the "Security" processor here is exactly opposite of what the name says.

Methinks the name is honest enough -- the aim is to keep the platform secure by preventing the user from doing with it what he might wish.

Yes, it's all about DRM. It's provably impossible to implement an unbreakable DRM scheme in software alone, so they're trying to push it into the hardware.

[Aitor 1]

Re: Probably a pass for most

No flame from me.

I do not trust AMD, Intel etc not to offer backdoors to the US gov. In fact, I would be surprised of the contrary.. just look at the laws, and they have to follow them!

[Christian Berger]

Re: Probably a pass for most

Opps I meant to write RISC-V instead of RISC-VI. I should sleep more. :)

[Bronek Kozicki]

"Intel says it is shipping its first 10nm Agilex FPGAs"

So, in principle, you could load a softcore FPGA implementation of POWER into an Intel FPGA and have an Intel FPGA running POWER instruction set, at a good speed.

[Anonymous Coward]

Cool

Looking forward to seeing the processors that come out from this. Also - would be nice to see some OS/2 back in the world.

We used to develop software on various platforms and the PPC one was one of the nicest.