UX, AIX, FLAYX...

Liam Proven lproven at gmail.com
Thu Sep 6 09:58:41 CDT 2007


On 04/09/07, Patrick Finnegan <pat at computer-refuge.org> wrote:
> On Tuesday 04 September 2007, Liam Proven wrote:
> > On 04/09/07, Patrick Finnegan <pat at computer-refuge.org> wrote:
> > > On Monday 03 September 2007 17:41, Liam Proven wrote:
> > > > On 03/09/07, William Donzelli <wdonzelli at gmail.com> wrote:
> > > > > > I have a number of PowerMacs for that side of things. :¬)
> > > > >
> > > > > The PowerPCs in Macs are the retarded little brothers of the
> > > > > real POWER processors.
> > > >
> > > > Arguably, this may be so, but I'd submit two rather important
> > > > riders on to that.
> > > >
> > > > [1] Whereas this was the case with the 601/603/604 and so on, I'm
> > > > not sure it's really true any more. The "G5" is pretty much a
> > > > 2-core POWER4, AIUI
> > >
> > > As I've read, the 601 is more POWER than the rest of the PPC chips.
> > > Still much slower and dumbed down than the proper POWER (non-PC)
> > > versions.
> >
> > Can you give some supporting evidence for that assertion? I've been
> > following the development of POWER and PowerPC since their first
> > announcements, and as far as I can see, the process has been one of
> > gradual convergence. The PPC601 was a single-chip implementation of
> > what was still a processor /chipset/ on the IBM side - IIRC, I think
> > the contemporary IBM processor was spread over 5 chips or something.
>
> I don't feel like looking this up.  Go look at SPECfp numbers for PPC
> and POWER chips of the same era, for a starting point.  There's also a
> number of POWER architecture descriptions floating around the internet
> which detail differences.
>
> > > And you have that backwards, the G5 is 1/2 of a POWER4 chip.
> >
> > I have nothing whatsover backwards and kindly don't make assumptions.
> > I stated that the "G5" - the PPC970 family - is a dual-core POWER4;
> > given that I was talking to someone who claims good knowledge of the
> > POWER processors but seems not to be so /au fait/ with PowerPC, I did
> > not see any necessity to spell out that POWER4 is a quad-core design.
> > PPC970 also adds Altivec-compatible SIMD instructions to POWER, but
> > is a highly capable 64-bit implementation of POWER.
>
> The POWER4 is dual core, the original G5 is single core.  The newest G5
> is dual core.  IBM didn't make ANY quad-core processors until the
> POWER5+.

Whoops!

I'm sorry, you're absolutely right. I was getting my early versus my
late models muddled up. I humbly apologise.

> > I'm not aware of much that modern PPC processers can't do that POWER
> > can, other than things intended for the support of legacy IBM OSs
> > such as OS/400 or zOS.
>
> The POWER chips have a bunch of hypervisor goo, and many, many MBs of L3
> cache on the multicore versions.
>
> > > I guess a
> > > dual-core G5 is close to a POWER4, but it's still lacking some
> > > things, like cache.
> >
> > (?) The 970 has onboard primary and secondary cache. In the first
> > model, 64KB of direct-mapped L1 instruction cache and 32KB of L1 data
> > cache, plus 512KB of 2-way associative L2 cache.
>
> I wasn't say it was without *any* cache.  It just has a lot less than
> POWER4.  No one builds modern processors without ANY cache, that'd just
> be silly.

Fair call. I think your phrasing was ambiguous, though: "it's still
lacking some things, like cache."

I have seen references to a cacheless Intel Celeron Ultra-Low-Voltage,
but I can't find mention of it now.

> > >  And, the G5 has VMX/Altivec, which the POWER4 doesn't.
> > > Mostly, the POWER4 was designed to be a enterprise server grade
> > > CPU, and the G5 is designed to be a consumer-grade CPU.  There's a
> > > lot of trade-offs that they made when designing one vs the other.
> >
> > They discuss the tradeoffs between POWER4 and PPC970 in some detail.
> > The PPC is and always has been a desktop processor for the retail
> > consumer market; its rivals were the Athlon64 and Pentium 4D. As such
> > it's tuned towards different demands than the POWER4, which is aimed
> > for expensive, non-cost-sensitive IBM servers, minis and mainframes -
> > not that there's a lot of difference between those 3 categories
> > today.
> >
> > To copy the 1st article I link's summary of the differences: "In sum,
> > the 970 is made to be faster, cheaper, and significantly less
> > reliable than the Power4."
>
> Why should I read those?  That was the exact point I was trying to make.

Well, because it makes points like this:
"One of the MPF seminar attendees, David Wang of RWT, revealed some
interesting information in a *very recent article* on the 970. Allow
me to summarize: As it turns out, at some tasks the PPC 970 will
actually outperform its big brother, the Power4, because the 970 is
made to run at higher clock rates. Apparently, in order to increase
the reliability of the Power4 for the high-end server market, IBM used
much thicker gate oxides on the chip's transistors. The trade-off for
this decreased failure rate and improved reliability was that the
Power4's transistors have slower switching speeds, so even with
process shrinks it's harder to push the design to higher clock speeds.
Since the 970 is made for the desktop market, there's no need for such
measures and therefore the new chip's clock speed will scale much
higher than the Power4's. In sum, the 970 is made to be faster,
cheaper, and significantly less reliable than the Power4. (Of course,
when I say "significantly less reliable than the Power4," you have to
understand that this puts the 970's product life and failure rate on
par with other mainstream CPUs, since the Power4's increased gate
oxide thickness makes it significantly more reliable than most
mainstream CPUs.)"

The *very recent article* is here:
http://www.realworldtech.com/page.cfm?AID=RWT101502203725

Which states that IBM, at the PPC970 launch, said:
"In the case of the PowerPC 970, the processor does not need to meet
similar reliability requirements as the POWER4 processor, and as a
consequence, circuit and process technology can be tweaked to obtain
higher performance by trading away the near-absolute reliability
required by the POWER4 processor."

> > > The POWER5 and POWER6 are quite a bit more interesting than any
> > > PowerPC chip.

In a somewhat theoretical sense, perhaps, yes.

> > Yes, arguably, but now Apple has abandoned the PPC, it's dead in the
> > water as far as the desktop is concerned, so it won't really evolve
> > any further. (Which I think is a great shame.)
>
> POWER5 and POWER6 have evolved quite a bit from where POWER4 was.  IBM
> made a lot more money selling POWER processors, systems, and consulting
> to go with it, than they ever did with selling PPC chips to apple.  IBM
> was dragging their feet with PPC features on the G5 (and whatever would
> have been G6), which is why Apple abandoned them; if IBM was making
> enough money from Apple on their G5s,  I'm sure that they would have
> payed more attention to what Apple wanted (like low-power noteboot
> CPUs).

Right now, PowerPC is selling like never before, as it's the basis of
the Xbox360, Playstation 3 and Nintendo Wii. It's shipping millions of
units - maybe tens of millions.

After their year or two in the sun, though, I think it might subside
into something of a minority niche.

> > On the other hand, in the embedded market, there are fascinating
> > chips like PA Semi's PWRficient line.
> >
> > http://news.zdnet.com/2100-9584_22-5907281.html
> > http://www.theregister.co.uk/2005/10/24/pasemi_power/print.html
> >
> > These will carry the torch forward for PPC, but we won't find them in
> > any affordable desktop machines.
>
> Depends on what you mean by "affordable".  Easily more affordable than
> an inflation-adjusted IBM 5150 or similar.

About the only company even claiming to make personal computers based
around PowerPC now are Amiga-compatibles. Amiga's products are
vapourware, as far as I can tell. At least you can buy Genesi EFIKA
boxes.

They're years away from any of the state-of-the-art stuff we've been
discussing, though. They're looking at well-sub-1GHz SoC processors.

> > > > [2] All the interesting apps and the desktop-relevant stuff is on
> > > > PowerPC. Specifically, on OS X. Even a 10y old G3 Mac with OSX
> > > > makes a pleasant and usable machine for the Web today and for
> > > > day-to-day use. I can't think of a lot of use for a 10yo RS/6000
> > > > except as a server - as a workstation running Motif or something,
> > > > it wouldn't be much use on the desktop today.
> > >
> > > Interesting depends on what you mean by it.
> >
> > Well, of course, but I went on to define what I meant in the same
> > sentence on the same line. Desktop-relevant stuff that I can actually
> > use. I'm reading and writing this on a web page; for the 2007 Web, I
> > need a modern CSS2-capable browser, Flash, Javascript, Java,
> > RealPlayer, QuickTime and Windows Media support. OS X delivers those;
> > even PowerPC Linux, probably the most-widely-supported desktop Unix
> > environment on PPC, does not.
>
> Whether or not you need Flash is a matter of debate... but a lot of that
> can be handled by firefox, IBM's PPC JREs, and mplayer/xine.  Who needs
> three separate media players, when you can use one of two OSS players,
> and play almost anything.  I hear that the OSS flash player is even
> very close to being able to play youtube/google flash videos.  But,
> alas, this still doesn't make a platform intersting to me.

OK. Clearly we find very different things of relevance.

I found PowerPC of interest because it was the Other Desktop Platform.
It is no longer. Now, pretty much, the desktop is x86 from horizon to
horizon, with a soupcon of ARM in a few low-power handheld and pocket
devices. I think that's a damned shame.

I personally use towers and notebooks. My clients, generally small
businesses, use PCs and the occasional Mac. That means x86 all the
way, with x86-64 starting to show on the horizon now. So what happens
in big enterprise iron is of little direct import or interest to me.

> > > POWER has AIX and i5/OS / OS/400.  by themselves, much more
> > > interesting to me than anything that runs on MacOSX.  Of course,
> > > I'm "one of those people" who runs Linux on their work-provided
> > > PowerMac G5.  I just can't seem to get my boss to want to pay $10k
> > > for a proper POWER5 workstation from IBM. ;)
> >
> > A snag!
> >
> > I find these to be academically interesting OSs, ones I'd like to
> > work with and know more about, but neither I nor my small-business
> > clients have any direct personal or professional use or need for
> > them. Unlike a Mac.
>
> Who said personal/professional use?  I though that people collected
> computers because they were "interesting", not just because they were
> useful.

I daresay most do.

Me, I don't collect it unless I have some *use* for it, even if that's
just in theory. I have Atari machines I've never turned on, because I
wanted to play with MINT and the like. Maybe one day I'll have time...

-- 
Liam Proven • Profile: http://www.linkedin.com/in/liamproven
Email: lproven at cix.co.uk • GMail/GoogleTalk/Orkut: lproven at gmail.com
Tel: +44 20-8685-0498 • Cell: +44 7939-087884 • Fax: + 44 870-9151419
AOL/AIM/iChat: liamproven at aol.com • MSN/Messenger: lproven at hotmail.com
Yahoo: liamproven at yahoo.co.uk • Skype: liamproven • ICQ: 73187508




More information about the cctalk mailing list