Is there any reason why IBM doesn't sell 'consumer' CPUs?

[chrisjames61]

IBM's x86 license was part of their work and deals with Cyrix.

I'm fairly certain that once Cyrix went defunct (and ended up in the hands of Via), IBM no longer had an x86 license.

VIA is still producing ultra low power x86 chips.

Intel doesn't like license transferals and these days is against companies working with an x86 license holder to create their own x86 chips.

Wow, a reply to a comment over 12 years old lol!

 

[chrisjames61]

The 'B' stands for business. That's where their focus is.

If x86 had completely dropped the ball, and another ISA like Risc-V or or Acorn failed to fill the vacuum, I think IBM could have very well adapted Power to fill the void. In the late 90s and early 00s PowerPC chips actually were in consumer computers (eg. Apple, Atari, Amiga, NeXT).

But since this hasn't happened I think they're completely happy focusing on banking and business.

The first PowerPC Macs hit the scene in 1994.

 

[chrisjames61]

Neither Atari nor NeXT ever used PPC, and NeXT never made consumer computers. The NeXTStation started at US$4995, and that was the low-end system - that's equivalent to $9800 today. The Cube was about double that.

68040. Motorola.

 

[Abwx]

well, I beleive IBM is one of the four companies who are legally allowed to make x86 processors. (64bit is still an x86 processor, not an x64)

From memory (!) there was IDT wich was bought by VIA, Cyrix, IBM, Texas Instrument, STMicro, and of course AMD, possibly a few other firms.

 

[SarahKerrigan]

68040. Motorola.

68040 is a 68000, not a PPC. Radically different things.

 

[amd6502]

Neither Atari nor NeXT ever used PPC, and NeXT never made consumer computers. The NeXTStation started at US$4995, and that was the low-end system - that's equivalent to $9800 today. The Cube was about double that.

Oops my memory is failing me. Thank you for the correction. I confused Motorla 68000 family with the PPC family. That leaves pretty much just PPC based Apples.

 

[DrMrLordX]

Some probably do and I bet there are quite a few here that blow 3k on a high end PC setup with a top of the line GPU and mobo.

It's a rich man's alternative (proper modern day alternative) to a FX 83xx secure computing setup and could be used as an actually a real server. 4c/16t at that. For enterprise that's top dollar value.

It's kind of cool that you can buy something like that, but at the same time, you can see that economy of scale is not working in their favor. Just the motherboard price . .. y0w. I doubt they move many Talos II units. You really have to buy into the idea that AMD's Trustzone is insecure if you need one of these as an alternative to one of their machines. For the rest of the world, you can probably get a lot more performance out of AMD or even Intel products for the money, depending on what you are doing with it.

But hey, at least they're trying.

 

[amd6502]

It's kind of cool that you can buy something like that, but at the same time, you can see that economy of scale is not working in their favor. Just the motherboard price . .. y0w. I doubt they move many Talos II units. You really have to buy into the idea that AMD's Trustzone is insecure if you need one of these as an alternative to one of their machines. For the rest of the world, you can probably get a lot more performance out of AMD or even Intel products for the money, depending on what you are doing with it.

But hey, at least they're trying.

And the other problem is that I think they may have stopped developing AIX and are relying on clients using linux. I wouldn't trust many linux distros to be highly secure, and the choice of distros becomes very limited when you have a not widely supported CPU. (Maybe something like Gentoo would work well for that workstation. )

Besides the OS, the library of software also becomes an issue. It's kind of a shame that the Sun model or vision of Java to liberate us from ISA dependent applications never reached close to what was hoped for.

Oops, I stand corrected, AIX is still alive and well (it just has a long support cycle).

AIX 7.1	September 10, 2010; 8 years ago	2022-04-30
AIX 7.2	December 1, 2015; 3 years ago	TBA

 

[DrMrLordX]

@amd6502

I'm pretty sure Arch and Gentoo support PPC64 (actually PPC64le) so you can produce a secure machine depending on your level of technical expertise. And because Arch supports it, I think you could get BlackArch running on a Talos II if you tried hard enough (or you could just replicate BlackArch using the base Arch distro).

In fact, I think you can, if you're willing to compile tools from source:

Download BlackArch

Downloads the netinstall or the live iso by torrent, http or ftp

blackarch.org

As part of an alternative method of installation, you can build the blackarch packages from source.

Now whether you really need all those penetration-testing tools to feel "secure", I don't know. Also I would imagine getting all that stuff to compile properly on a Talos II could wind up being tricky. Or maybe not. Regardless, many people consider BlackArch to be one of those most secure Linux distros out there, above even Kali (that does not have a PPC64 distro available).

Depending on whom you ask, the only truly secure OS out there is OpenBSD; sadly, I do not think Theo has any plans to support PPC64Le at all.

 

[NostaSeronx]

For consumers? How about niche consumers? Amiga X5000/Amiga A1222.

Not IBM, but IBM IP. Hopefully, IBM/Linux Foundation can make POWER an alternative to RISC-V/ARM. (I have a feeling IBM will get AMD to port Zen to POWER)

 

[NTMBK]

As far as I know, the last consumer device using a PowerPC CPU (outside of little embedded cores) was the Wii U.

 

[moinmoin]

As far as I know, the last consumer device using a PowerPC CPU (outside of little embedded cores) was the Wii U.

Indeed, and that Espresso CPU was quite an odd chip, being a custom three cores version of Broadway (Wii's CPU also released by IBM as PowerPC 750CL) which again was very close to Gekko (GameCube's CPU based on IBM's PowerPC 750CXe which was also used by Apple in their last PowerPC based systems as "PowerPC G3").

 

[SarahKerrigan]

And the other problem is that I think they may have stopped developing AIX and are relying on clients using linux. I wouldn't trust many linux distros to be highly secure, and the choice of distros becomes very limited when you have a not widely supported CPU. (Maybe something like Gentoo would work well for that workstation. )

Besides the OS, the library of software also becomes an issue. It's kind of a shame that the Sun model or vision of Java to liberate us from ISA dependent applications never reached close to what was hoped for.

Oops, I stand corrected, AIX is still alive and well (it just has a long support cycle).

Ubuntu, Fedora, SLES, and RHEL all run just fine on PPC64le and are officially supported. (Can't speak for more niche things like Arch.)

AIX and iSeries are alive and well too but don't run on OpenPower gear; the hardware is the same but the firmware is not. (PowerVM stack vs PowerNV/OPAL.)

 

[DrMrLordX]

Ubuntu, Fedora, SLES, and RHEL all run just fine on PPC64le and are officially supported. (Can't speak for more niche things like Arch.)

The implication may be that he does not consider mainline Linux distros to be secure. If you're really a fanatic about these sorts of things, they really aren't. But how many of us run OpenBSD?

 

[whm1974]

I wonder why all the various vendors's PPC systems used something like three different firmware and not just a single one. As I recall Apple used their own as did IBM with AIX systems, and Motorola and others that run WinNT had ARC bios.

Wouldn't far more PPC CPUs and systems just used single one? And if they are used the same platform specs?

 

[whm1974]

The implication may be that he does not consider mainline Linux distros to be secure. If you're really a fanatic about these sorts of things, they really aren't. But how many of us run OpenBSD?

Wouldn't security be more dependent on what the Root and regular users are doing anyway? I mean if the Root doesn't keep up wth sercurity and users have Bad Practices then it is just a matter of time before siad system is hacked and compromised. Right?

 

[DrMrLordX]

Wouldn't far more PPC CPUs and systems just used single one? And if they are used the same platform specs?

Maybe it has something to do with some generations of POWER being bi-endian?

Wouldn't security be more dependent on what the Root and regular users are doing anyway? I mean if the Root doesn't keep up wth sercurity and users have Bad Practices then it is just a matter of time before siad system is hacked and compromised. Right?

Some security problems come straight out of the box though. Examples:

Security notices | Ubuntu

Ubuntu is an open source software operating system that runs from the desktop, to the cloud, to all your internet connected things.

usn.ubuntu.com

Ubuntu has already dealt with those problems; still, 18.04 LTS has been out for over a year, and some of those problems were only posted a few days ago on the security notices page. Plenty of users might not be running exim, but systemd is default for everyone. You can't just get rid of it easily, either. Most folks that want to avoid systemd gravitate to other distros. Of course you'd have the same problem under Arch by default. Not sure if Artix can be made to work with PPC64. Gentoo defaults to OpenRC so there is always that option.

 

[whm1974]

Maybe it has something to do with some generations of POWER being bi-endian?

Well that to, and the all the vendors should have agreed to use little endian like x86 did. By the way doesn't little endian even uses less memory and storage?

 

[DrMrLordX]

Well that to, and the all the vendors should have agreed to use little endian like x86 did. By the way doesn't little endian even uses less memory and storage?

It might. Supposedly POWER9 can perform a little better in big endian mode though. So, you know, tradeoffs.

 

[whm1974]

It might. Supposedly POWER9 can perform a little better in big endian mode though. So, you know, tradeoffs.

Maybe, but I can't really say as I'm not an expert on endianess. And the Endian Wars are long over due to X86 winning the ISA Wars.

 

[scannall]

I bought one of the Cheese Grater Macs when they came out, with the G5. That thing was a beast compared to the Intel offerings at the time. It was competing against the P4, and just flat out smoked it.

 

[SarahKerrigan]

I wonder why all the various vendors's PPC systems used something like three different firmware and not just a single one. As I recall Apple used their own as did IBM with AIX systems, and Motorola and others that run WinNT had ARC bios.

Wouldn't far more PPC CPUs and systems just used single one? And if they are used the same platform specs?

Enterprise Power and PowerNV use different firmware because of different requirements; AIX and iSeries generally run virtualized (even if there's only one LPAR on the system) and so Enterprise Power firmware has a lot of hooks to support the HMC, as well as PowerVM and VIOS components. PowerNV - Power NonVirtualized - systems have no HMC, and boot Linux on bare metal. That was essentially a clean sheet for OpenPower, and is fully open-source, unlike the firmware used in enterprise Power systems.

Apple and IBM both used Open Firmware; IBM still does for the enterprise machines, although it has diverged somewhat to support stuff like the features mentioned above. ARC was a standard that went well beyond just PPC, as it was also used on MIPS and, occasionally, x86 systems. It was sort of a generalized "1990s things that are expected to run NT, especially non-x86" industry standard that didn't quite take off.

 

[SarahKerrigan]

Well that to, and the all the vendors should have agreed to use little endian like x86 did. By the way doesn't little endian even uses less memory and storage?

It shouldn't; it's just a representation difference. The primary consideration is just compatibility with x86-oriented data structures - which was a lot less of a consideration when Power was new (1990ish, when x86 had almost no serious server footprint) than today; most of the processors of the era were either big-endian-only (SPARC) or bi-endian but universally used in big-endian mode (PA.)

 

[whm1974]

It shouldn't; it's just a representation difference. The primary consideration is just compatibility with x86-oriented data structures - which was a lot less of a consideration when Power was new (1990ish, when x86 had almost no serious server footprint) than today; most of the processors of the era were either big-endian-only (SPARC) or bi-endian but universally used in big-endian mode (PA.)

If I recall didn't DEC have some MIPs workstations and servers that were little endian?

 

[SarahKerrigan]

If I recall didn't DEC have some MIPs workstations and servers that were little endian?

A few; Alpha was usually LE too. But if you look at every other major server processor at the time, it's near-universally BE. (Most MIPS systems, Power, SPARC, PA, 370...)

This is not something where there's a clear better or worse, and compatibility with x86 data structures was not on the minds of most processor architects in the late 80s as a major feature. x86 servers at the time mostly meant NetWare (plus a smattering of low-end Unix), which competed in a somewhat different territory than the RISC and mainframe processors of the day.

Power being big-endian is probably justifiable purely from the standpoint of 370 being big-endian, ignoring all other considerations, as affinity with 370 mattered somewhat more for Power's target market than affinity with x86. And again, there is not a technical right answer here.