Comparison Power PC architecture VS ARM ?

[ZetaZuds]

Hello Guys,

Please, can someone give a comparison between Power PC architecture and ARM, what are the differences, advantages, what is the best one should I learn?

Thank you in advance
Best regards

[legacy]

Please, can someone give a comparison between Power PC architecture and ARM, what are the differences, advantages, what is the best one should I learn?

PowerPC is used in avionics, our fly-boards are PowerPC-440 and 460. It's also used in automotive, e.g. Valentino Rossi's motorbike uses a PowerPC e500 dual-core. It's also used in Formula One, e.g. Redbull uses a PowerPC e500 quadri-core.

If you are looking for a job in these areas, you'd better learn PowerPC.

The PowerPC architecture is described by four books by IBM (red, yellow, green, gold), covering everything from PPC601 to Cell, as well as the last POWER9 architecture used by Darpa, FBI, CIA, and other federal agencies in USA is described in books by IBM dedicated to the POWER-line. We have now low-cost POWER9 Workstations running Linux. They are a bit costly, a maxed out unit costs around 5K USD.

PowerPC is also used for some routers applications, as well as for some storage applications (i.e. encrypted, at the low level, hard drives).

I am using an old (and deprecated) 405GP core for this multi-node NAS machine. But it's a hobby project 

[ogden]

Hello Guys,

Please, can someone give a comparison between Power PC architecture and ARM, what are the differences, advantages, what is the best one should I learn?

Thank you in advance
Best regards

Here you go

- Definitely there are many someone's, you just have to find one of them. Not necessarily in this forum.

[MT]

The PowerPC architecture is described by four books by IBM (red, yellow, green, gold), covering everything from PPC601 to Cell, as well as the last POWER9 architecture used by Darpa, FBI, CIA, and other federal agencies in USA is described in books by IBM dedicated to the POWER-line.

Mind elaborate some more on why Darpa, FBI, CIA, and other federal agencies uses Power PC/POWER9? Thanks!

[legacy]

Mind elaborate some more on why Darpa, FBI, CIA, and other federal agencies uses Power PC/POWER9? Thanks!

Perhaps because they think that x86 has too many back-doors and crap in the silicon and they don't want to deal with Intel. News like this might have some basis of credibility onto their feet 

Anyway, I don't know the exact reason, but I know that Darpa is sponsoring and promoting the POWER9 project giving a lot of money to IBM to continue the support.

I know that FBI, CIA, and other federal agencies have already bought a lot of these POWER9 workstations, and server, as there are contributions to Linux for that purpose. As well as requests for AIX.


Anyway, here are some low-cost RACK version of the new POWER9 machine (Raptor is the company that makes POWER9 for the mass), besides we now have advanced ARM64 workstations, here are some tests, POWER9 vs ARM64

[Bassman59]

Hello Guys,

Please, can someone give a comparison between Power PC architecture and ARM, what are the differences, advantages, what is the best one should I learn?

Except for specialist and legacy applications, PowerPC is dead. I used to work for a company whose main product were PowerPC-based SBCs and they were solid parts and compared to the x86 parts at the time, they were simpler, had lower power consumption and were solid performers. (ARM then wasn't what it is now.) Unfortunately the end of the line was basically the G5.

At the next job, we liked them when they were the hard cores in the Xilinx Virtex 4 and even migrated to a standalone PPC glued to a simpler/cheaper FPGA, but the vendors stopped making them and wouldn't support us, so we migrated to ARM.

[MT]

Mind elaborate some more on why Darpa, FBI, CIA, and other federal agencies uses Power PC/POWER9? Thanks!

Perhaps because they think that x86 has too many back-doors and crap in the silicon and they don't want to deal with Intel. News like this might have some basis of credibility onto their feet 

Anyway, I don't know the exact reason, but I know that Darpa is sponsoring and promoting the POWER9 project giving a lot of money to IBM to continue the support.

I know that FBI, CIA, and other federal agencies have already bought a lot of these POWER9 workstations, and server, as there are contributions to Linux for that purpose. As well as requests for AIX.

Perhaps China security agencies also run POWER9 machines? What if all Stasi organisations on planet earth runs POWER9 then all spying is obsolete and we can give each other flowers instead! My Huawei smart phone runs Android by Google known for spying so there we are both China and US spying on me when i watch porn, they must be pleased
qoute:
Because if Lenovo is doing this, are we supposed to be so naïve to presume that Google, Apple, AT&T, etc are not?”

[westfw]

PowerPC was a very early player in the (true) RISC microprocessor/microcontroller areas.  Motorola was there, and various vendors went from 68k systems to PPC, especially if they needed a big-endian architecture (like many networking equipment vendors.)   And Apple, too!  Chips like the "PowerQUICC"  (PPC core with 4 high-speed communications interfaces) were pretty much routers-on-a-chip.

Alas (?) PPC was perhaps "unfocused", chasing desktop, gaming, embedded, and supercomputer markets all at once.MIPS took over most of the Embedded (ie network appliance), Intel won the Desktop, and ARM won phones (and used  the resulting influx of cash to expand ... everywhere.)

While I worked on some of that PPC network equipment, I never had cause to figure out the "Architecture" any further than "supported by GCC, and someone else as ported the tiny number of assembly-language routines needed for our OS to work."

(And let's see... Motorola became Freescale, and was then acquired by NXP (and almost Qualcom), IBM got out of most small-scale computers, P.A. Semi was acquired by Apple, Apple switched their desktops to Intel, TI bought Luminary, and Intel released a bi-endian compiler.  Most of the major players - out, or at least diluted.  I don't recall any major complaints against the architecture, but "they" sure blew the business side.)  (Hmm.  It may be that they hit a power (as in Watts!) wall.  I don't recall there ever being much focus on "low-power ppc" work, and that may have become important even for larger scale CPUs, as clock rates went up.)

[legacy]

Perhaps China security agencies also run POWER9 machines?

Dude, you can't export the same class of POWER9 machines in use by federal agencies in USA, only the underpowered version (Talos™ II Secure Workstation) offered to the mass, and the whole line of POWER9 CPUs by IBM does not have any backdoors inside the chip, whereas x86 CPUs by Intel do have a lot of them.

I am not a fanatic, I don't really care and I am still using an Intel x86 laptop, but these were two points I am happy to see, simply because they open new doors.

Because if Lenovo is doing this

Intel is doing that. Intel makes CPUs.

[legacy]

While I worked on some of that PPC network equipment, I never had cause to figure out the "Architecture" any further than "supported by GCC, and someone else as ported the tiny number of assembly-language routines needed for our OS to work."

AMCC is one of the interesting partners we have in avionics.

But working on fly-boards means you have to deal with tons of low-level stuff that needs to be written in assembly, and you can forget to use GNU(4). Here you start using professional tools like Greenhills for both C, Ada, and assembly, and they help you with the toolchain, and with their business contract there is always someone to call when you have a problem with the compiler or with the assembler, and it's a great help, except when you have to deal with superscalar-CPU, multi-core, and coherency of the hardware.

But at some point, the PowerPC guideline makes itself confusing and a bit frustrating, since the four holy books (with thousands and thousands of pages to read) claimed they would have defined the PowerPC ISA in all declinations in a coherent way, from the PPC601 to Cell, while there are random differences in the implementation of this and that chip even within the same class, and with/out the superscalar version of the chip you might have to deal with strange (sometimes undocumented, unless you pay for the documentation) bad-effects.

Definitively not nice.

Something similar happened with the first and second generation of MIPS CPUs, R3K-R4K, with a mess made by different vendors about the cache behavior, cache instructions, and COPs stuff: all different from a MIPS-CPU made by NEC, a MIPS-CPU made by Toshiba, and so on. Then they unified the ISA with the guideline for the implementation of the new MIPS32-R2 and MIPS64, and nowadays we are all happy with that. Really!

IBM has never done anything similar, and I have literally seen people save the day by adding some sort of voodoo code, i.e. EIEIO-opcode followed by iSYNC and some cache manipulation in a magical order, in the low level of an Avionic Automa (1) as well as in CBIT chunks of code(2), and the solution that was OK for a board with a CPU (say for 440), still needed to be adjusted (and re-tested) for another CPU (say for 460).

Linux on a PPC750(G3) is almost the same as Linux on PPC7550(G4), but Linux on a G5 is completely a different story, and a different profile: G3 and G4 are Linux/PowerPC32-BE, the G5 is profiled as Linux/PowerPC64-BE

Oh, here the story makes itself more intriguing because nobody knows the reason why IBM switched from big-endian to little-endian, and this breaks all the legacy compatibility with AIX, Linux, and all the existing software, in fact, the POWER9 is profiled as Linux/PowerPC64-LE, with a couple of instructions to handle the endianess, thus QEMU on POWER9 is virtually able to execute the BigEndian code with hw acceleration 



Anyway, with ARM I have never experienced this embrace, and I am an ACORN(3) fan, as well as an owner of an old RISPC/600 loaded with a StrongArm@200Mhz CPUmodule


1) sort of bootloader with paranoid diagnostics
2) CBIT = continuous built-in tests, they are low-level tests injected into the ram by the Automa, and periodically executed by the OS to check that the hardware is still OK when the aircraft is flying
3) it's the British company that created the first ARM design, to be used for their RISCPC computer
4) that is also largely bugged, you can see the Linux sources for that, e.g. for core like 405GP, gcc & binutils don't correctly support alignments

[Muxr]

POWER9 CPUs by IBM does not have any backdoors inside the chip, whereas x86 CPUs by Intel do have a lot of them.

[Citation Needed]
If a government is demanding backdoors into one architecture why wouldn't they demand backdoors into the other?

Besides you're completely forgetting another major player AMD, with its Epyc series of x86 CPUs. Ie. Intel is no longer the only game in town as far as x86_64 compute in the datacenter is concerned.

[legacy]

Besides you're completely forgetting another major player AMD

A lot of feral agencies are not using common x86 computers. This is a fact. AMD is the same as Intel with the backdoors politics. A bit less dirty (or simply less documented about backdoors), anyway everything that is x86-compliant has hw backdoors.

It might sound like a conspiracy theory 

[MT]

Dude, you can't export the same class of POWER9 machines in use by federal agencies in USA,

But dude, im guessing, i basically know nada regarding rules around ultra security! You link gave the fact
U.S. State Department had bought 16000 Leneovo things so clearly federal gov didnt know what its was doing!

only the underpowered version (Talos™ II Secure Workstation) offered to the mass, and the whole line of POWER9 CPUs by IBM does not have any backdoors inside the chip, whereas x86 CPUs by Intel do have a lot of them.

Well then the Chinese maybe is using the Talos™ II ""Secure"" Workstation, better then x86 at least.

Trusted Security
Talos™ II drives the state of the art of secure computing forward. Talos™ II gives you — and only you — full control of your machine's security. Rest assured knowing that only your authorized software and firmware are running via POWER9's secure boot features. Don't trust us? Look at the secure boot sources yourself — and modify them as you wish. That's the power of Talos™ II.

OpenBMC
In an industry first, Talos™ II ships with fully open and auditable BMC firmware, based on the Open BMC project. Gone are the days when you had to carefully isolate the buggy, insecure BMC port from threats at the firewall level. With Talos™ II, the BMC is just another Linux system that can be maintained as part of normal workflow. Find a bug or vulnerability? No problem; just patch, recompile, and instal

[/i]

I am not a fanatic, I don't really care and I am still using an Intel x86 laptop, but these were two points I am happy to see, simply because they open new doors.

Is fanatism a requirement? For what? I dont understand what you mean! Im also on a backdored x86.

Intel is doing that. Intel makes CPUs.

Yes and those are inside Lenovos that U.S. State Department bought! Also known CIA and NSA asks for backdoors as well as to try to get their encryption code accepted by the encryption organisation but was rejected recently. Story is on the web.

[MT]

A lot of feral agencies are not using common x86 computers. This is a fact.
It might sound like a conspiracy theory 

Which feral agencies?

[helius]

Which feral agencies?

That's a dogcatcher, I think.

[helius]

Oh, here the story makes itself more intriguing because nobody knows the reason why IBM switched from big-endian to little-endian, and this breaks all the legacy compatibility with AIX, Linux, and all the existing software

Someone made the decision that there was more code that depends on little-endian and losing big-endian support wasn't important.
Webassembly, for example, has a fundamental little-endian dependency.

[westfw]

you start using professional tools like Greenhills for both C, Ada, and assembly, and they help you with the toolchain, and with their business contract there is always someone to call when you have a problem

Yeah, yeah.  You can sit in your aviation-spec tower with you DoD-style budgets and your ability to whine at your "pro" compiler vendors, and I'll sit over here with my gcc-based product and my 300x-appreciated stock, even if we had to pay CodeSourcery and hire our own set of compiler gurus. :-)

Now, admittedly, planes weren't going to fall out of the sky and kill people if gcc had some compiler bug that we didn't notice during testing.  But I'm really tired of the frequent admonitions of the "gcc isn't good enough for professional work" form...

[legacy]

Well then the Chinese maybe is using the Talos™ II ""Secure"" Workstation, better than x86 at least.

The software for x86 (i.e. Windows) is not re-addressed to POWER9, this makes Federal Agencies happy, and they do no use Talos™ II (Talos™ II is a reduced version), they use more advanced POWER9 workstations and servers.

These machines are exclusively requested by DARPA to IBM, as the main funding source to develop the architecture, and - they say - it's an important exclusive point for computing performances, especially for their crypto-stuff 

Is fanatism a requirement? For what? I dont understand what you mean! Im also on a backdored x86.

A lot of fanatic guys (e.g. guys at Nekochan) on the internet claim they use { PowerPC, POWER9, HPPA, MIPS, Alpha, SuperHitachi, SPARC, ARM, Itanium, ... } ... non-x86 computers because they believe they are "better than x86".

When you ask to define "better", they answer tons of bullshit, and they don't know absolutely nothing about the ISA, neither they program in assembly, rather they use non-x86 machines simply because - oh, ah uh! it looks cool

This means being fanatic, especially when you are not able to support opensource, and you don't have software to do something useful with your computer 

My new pocket calculator is made by CASIO, and it has a SuperHitachi SH3 chip on it. I am not happy/unhappy to know the CPU is an SH3 chip. I am happy to know that there is an SDK released by CASIO, and I can develop some good app for it.

I might "hate" x86 only if I have to program it in assembly (it rarely happens), or if I have to implement it in FPGA ( in this case I will refuse to do it, there is too much complexity), otherwise (for the user point of view) Linux on x86 is exactly equal to Linux to { PowerPC, POWER9, HPPA, MIPS, Alpha, SuperHitachi, SPARC, ARM, Itanium, ... }

yet another Linux box, except that:

• out of the area of interest (that nowadays is mainly around x86 and arm), there is less support for everything, for the Linux/kernel and for the userland might have more unfixed bugs (HPPA has already shown that it's a problem)
• there might be no bootstrappers (e.g. there is Ada bootstrapper(1) only for { x86, arm })
• a lot of PCI/ePCI cards (e.g. SATA controllers, VGA cards) do need a BIOS initialization, and during this process, they inject x86 code, thus can't they work on a non-x86 machine unless you put an x86 emulator into the Linux kernel, or into the userland, that is crazy
• a good point about non x-86 machines is that the PCI doesn't come with all the legacy stuff that x86 computer must have in order to support old and deprecated peripherical, such as old VGA cards used in 386 and 486 computer. This stuff, if you read the Linux source, is a mess of irritating and annoying layers of crap that only make the BAR's code confusing and bugged Non-x86 machines don't need it, thus their source code is neat

Anyway, I don't care about Federal Agencies unless they want to pay me to help at supporting Linux on POWER9. It will never happen, I have a VISA permission to work in the USA, but I am and will ever be a stranger ... might be I will be paid to support Talos™ II's Linux stuff .. mumble ... and in the meanwhile ->  x86 is, and it will be, a good choice for me because it's largely diffused and used, it has brilliant performances, it runs all the software I need (including Windows Xp, Altium, Dina, and these x86-apps), and doesn't cost too much.


(1) Ada bootstrapper, a short story
Since paid by a couple of customers to do that, I worked hard to cross-compile an Ada compiler for HPPA, and it took several months of effort. In the end, I used an old Gnat compiler released for HPUX v10.20, from which I cross-compiled for Linux/HPPA, and then I recompiled the compiler natively. 

PowerPC has already removed the support for modern Gnat compilers, and on HPPA it looked like a horror story. Now Gnat works on Linux/HPPA, but the funny point is that for x86 Gnat has always been ready out of the box and the effort you have to invest to get Gnat working is equal to zero

That's life

[legacy]

and I'll sit over here with my gcc-based product and my 300x-appreciated stock, even if we had to pay CodeSourcery and hire our own set of compiler gurus

eh-eh-eh, like the making bat-signal call for the opensource compiler gurus?

you are in trouble, and you put on the sky lamp "hey, oh? compiler gurus?  help me, please, come here to defeat the villain compiler", and then your phone rings, "it's Denx here, we offer professional support for GCC and Linux" - oh, good ... GNU is free, this is it for free? !!! - "emm, sir, our price for consultancy is no less than one hundred bucks per hour"

one hundred bucks? per hour? whaaaaaaaaaaat?!?!? X___________X

Sigh, d'Oh, sad reality, in comics books, Batman has always answered the call for free 

[iMo]

PowerPC is used in avionics, our fly-boards

The only reason why powerPC is used in avionics, space, military is they are produced in RADHARD versions.
Flying high or through a nuke mushroom cloud with commercial chips is not recommended... 
There is no other technical reason to not replace them with ARM/MIPS or any other modern arch.

[legacy]

There is no other technical reason to not replace them with ARM/MIPS or any other modern arch.

Yup. Also because PowerPC has been certified for the purpose, and contracts with vendors have always guaranteed replacement parts done in the exact way they were certified, this for 20 years of coverage.

And 20 years have already expired, but in avionics we are still using 440 and 460, they are G3 with steroids (500Mhz? 1Ghz? yeah, we have some of them for MACH3 aircrafts) and redundancy (oh, about that, we have "voters", ASIC chip outside the CPU), whereas for space missions they use the old PPC601 and PPC603, redesigned for radiation hardened, but the core is a G1.

Maybe their contracts have considered longer coverage 

Anyway, with ARM and MIPS you can't have these two assumptions since they change too frequently, and there is no vendor (chip maker) that want to honor such a contract. Why not? I have never understood it 

I believe - take it as my personal speculation - that it's a bit hard for MIPS, but probably they will reconsider the business for the ARM stuff. But here it's not going to happen for 2019, anyway.

[langwadt]

Please, can someone give a comparison between Power PC architecture and ARM, what are the differences, advantages, what is the best one should I learn?

PowerPC is used in avionics, our fly-boards are PowerPC-440 and 460. It's also used in automotive, e.g. Valentino Rossi's motorbike uses a PowerPC e500 dual-core. It's also used in Formula One, e.g. Redbull uses a PowerPC e500 quadri-core.

all motogp teams use a Magneti-Marelli ECU, all F1 teams use a McLaren ECU, they are required to in the rules

[legacy]

all motogp teams use a Magneti-Marelli ECU, all F1 teams use a McLaren ECU, they are required to in the rules

I worked for them between 2013 and 2017, inside motoGP and F1 control boards there was a PowerPC e500.