Building a PowerPC Linux Box? 20
jglassco asks:
"Building an Intel Linux box is trivial, but what about building a PowerPC based Linux box? I have searched the /. archives yet have found nothing on this topic. My purpose here is to find sources for the two components that seem the hardest to locate, namely:
a motherboard (2 or 4 processors) with 4x AGP slot, and processors.
I want to build the rest using the common components found as commodities. Is anyone experienced with this?"
Check out LinuxPPC.com (Score:1)
check them out at linuxppc.com [linuxppc.com]
R5 looks amazingly like RH6.0
PReP/CHRP is pretty much dead (Score:2)
You might be able to find a board from this era, but it would probably be 603 or 604 only and certainly wouldn't have things like AGP and PC100 memory. It would probably just be easier to pick up an old Mac Clone or maybe a Motorola PPC machine.
* The conventional wisdom in the early 90s was that Intel x86 would fail to keep up with Moore's law. Lots of graphs were produced showing Intel's speed increases leveling out in the late 90s. IBM proposed PReP as a replacement to the x86 PC spec. While virtually everyone announced support, the only OSes that ever shipped for PReP were AIX and Windows NT (and Linux). The big problem was that Apple didn't buy in running their OS on an open platform, and of course, Intel never lost the price/performance lead for lower-end systems. PowerPC right now is pretty much an Apple and IBM midrange-specific CPU. I'd be nice if PReP had succeeded, but alas.
--
Re:I almost believed you... (Score:1)
You might replace "intelligently-written" with "backed up with verifiable facts" in my original statement and it would carry the same meaning, even if David does need to clean up his grammar a bit and use a spell-checker more often.
And Intel *is* behind the curve. Just take a look at the numbers. BTW, I have lots of respect for the guys at Intel. It's just very unfortunate that the company seems hopelessly stuck on "backwards compatibility" for it's own sake.
The performance of non-Intel x86 variants is irrelevant in the marketplace. The other companies are entirely dependent on Intel to produce what little innovation there is in the x86 space.
As far as IA-64 goes, well, let's not count unhatched chickens. The thing only taped out the other day, and its a long, long way from shipping. Intel really backed themselves into a wall with EPIC.
The "Big Windows Crowd" will go wherever billg@microsoft.com tells them they're "going today".
Clarification (Score:2)
Midrange ~= "Minicomputer" ~= Custom Unix box ~= not a commodity or "desktop" machine.
BTW, while the PowerPC certainly has the technical advantages you speak of, the "true status" of PReP/CHRP is that it's dead as a consumer platform.
--
Re:Clarification (Score:1)
PReP and CHRP are not the same thing. IBM created PReP as a counter-offensive/alternative to CHRP. The two were somewhat reconciled in the revised CHRP, otherwise known as PPCP.
Perhaps you are correct that all three specs, as they originally stood, are dead, but that doesn't mean the idea of an open PPC platform is dead.
Re:The reality of PowerPC (Score:1)
You came off as just a bit self-reightous (not to mention off-topic), so I would like to address a few of your points (as well as go even further off-topic).
How is SSE (Intel's floating point vector instruction set ) inferior to Altivec? They are both instruction sets that increase floating point performance via parallelization. There is nothing inherent in SSE that restricts it to multimedia applications. If you are refering to MMX (Intel's integer vector instruction set), that was a kludge that didn't allow use of the FPU at the same time. There is also AMD's 3DNow which is another floating point accelerator (it doesn't have any performance hit on normal FP ops).
I don't understand why you bring up the price of the Xeons with huge cache. All chips with big, fast cache are incredibly expensive. If cache memory were cheap we weould use it for RAM! The reason people buy these more expensive chips is that they provide huge performance boosts in many applications (serving static data comes to mind).
You call AGP and faster RAM "hacks," but then brag that future Macs will sport the same features. I don't know how you could conceive that more unified memory access (AGP allows graphics cards to use system RAM as texture memory) and faster RAM as "hacks." Also your description of the future Macs sounds like a less scalable AMD Athlon (K7)
I do agree however that in a vacuum, Motorola's chips have better technology. But when you factor in cost and industry support I believe Intel comes out on top. The performance hit that comes with backwords compatability is a tradeoff.
If you are going for pure performance though, DEC Alphas blow both Motorola and Intel out of the water.
Re:The reality of PowerPC (Score:1)
Sorry about the self-righteousness; I got a little annoyed. It seems like all anyone can do these days is blindly follow the crowd.
Isn't SSE otherwise known as MMX2? Isn't it just *more* instructions piled on top of the same core? AltiVec is an entirely separate processing unit, not a replacement for existing instructions. AltiVec (and the Power3 vector unit) provides more functionality without the trade-offs that Intel's schemes demand. If I'm wrong about SSE, I'd like to see documentation of it.
As far as 3DNow goes, I have little doubt that it will be an x86 marketplace failure if it's not supported by Intel. Remember all the complaining when Intel moved to Slot 1? Seems to have died down, doesn't it?
The point of the Xeon price comparison was to point out the extremely low price/performance ratio of the fastest Intel chips currently shipping. Anyone can make a chip faster by throwing lots of expensive cache at it, or reducing the die size. It's much, much harder to actually improve the architecture, something Intel hasn't done in nearly *five years*.
I didn't call AGP and faster RAM "hacks", I was using the term in a vastly different context, one referring to pundits of inferior technologies. I was pointing out that AGP and faster system RAM do not translate into proportional performance increases.
As a good example, many first- and second-generation PCI Power Macs, when upgraded with G3 processor cards are faster than the first-generation G3 systems, which have faster system busses; this is due to the superior architecture of the earlier machines. Unfortunately, the lower cost of the later machines was bought at the price of performance.
I'm not sure I understand the Athlon comparison, but if you look at the power consumption of those Alpha's you'll be shocked. If I threw that much electricity at *any* processor it would be whole truckloads faster.
Backwards compatibility? What for? Are you still running DOS 5.0? (Oh yeah, that's right...MS *still* requires DOS 5.0 for some stuff...) I'd rather have *forwards* compatibility, thank you very much.
I could care less who's "on top". I just hate hearing people bash a platform they don't know well. The original thread of this discussion was from a person seeking PPC components; comments to the effect that the seeking is a waste of time are not helpful.
Re:Clarification (Score:2)
(Actually curious) -- Do the new iBooks run Linux out of the box?
What I'd like to see is an actual server hardware plaform from Apple. (Rumors of such exist on the Mac sites.) This group would probably have the political pull to ensure good Linux support. If you're leaving it up to the iMac marketeers, I'm not so sure (despite one Linux evangalist).
--
Re:I almost believed you... (Score:1)
As for getting the Windows crowd to hop to PPC, you've got to get them off Windows, onto a portable, multiplatform (hardware) system like Linux before you'll ever get them off x86. That's because to be competitive with x86 using emulation, you'd have to be ~10x faster (native), and with Intel's $$$$'s, you'll never be _that_ much faster.
Not just for Linux? (Score:1)
I once read that Macos 8 would run on chrp/prep boxes that didn't have Apple roms, though of course, in an unsupported fashion. When I say heard, I don't meen the early Apple promises, but rather users saying so after system 8 was released. Has anyone ever seen this or have anything else to back it up? I've found a single article about it, but at the time of this writing the server appears to be down. here [eenn.net]. I also remember hearing it on Macosrumors way back when, but I'm not sure how reliable Macosrumors is . . .
I'm really curious about this, I've kinda had it on the top of my head for a while(yes I don't have much to do!) So if you could tell me anything I don't know, I'd be happy.
I almost believed you... (Score:1)
I would really like the PowerPC to succeed. Its a nice design (fast, low power, clean architecture), but in the current market you pretty much have to buy a Mac (or IBM workstation) to get a functioning system with one. The chip itself comes at a very reasonable price, but that doesn't help us, the end consumers, very much.
And Intel is *not* behind the curve...yet. They may be shortly, but even if they do fall behind, there are other x86 makers. I have great hopes for AMD. I'm not sure how much more Intel can squeeze out of the dated P6 line, and they probably don't either. They intended to be transitioning to IA-64 by now. If the delays continue much longer, they could start to lose market share. But lose market share to whom?
I think the Windows crowd (read: the BIG market) is too tied to x86 to hop over to PowerPC or Alpha without a fight. I don't think that sort of change will happen so long as *someone* produces a decent x86.
I'm curious to see what will come out of the next few quarters. I think it will be an interesting time for microprocessor enthusiasts to watch.
--Lenny
Re:Clarification (Score:2)
My main problem with your original post is that the question was where parts for a commodity PPC box could be found. Read between the lines, and I assume that the guy doesn't want to buy a Mac and install Linux on it. Yet you essentially told him to go buy a Mac.
(Not that I have a real problem with Macs, since the price is certainly reasonable even if don't want to run MacOS. However, Apple is in the business of selling Macintoshes, not general purpose boxes, so hardware support is always going to be more iffy than a truly 'open' plaform.)
--
Re:Clarification (Score:1)
Buying a Power Mac to run Linux is only *one* option. Why would you assume that Apple would advertise a job for someone whose duties would include "making sure Linux runs on all Apple hardware platforms" if they intended to provide "iffy" support?
Re:I almost believed you... (Score:1)
I don't have to do anything to get the Windows crowd to move to a different platform. Microsoft will do it if and when they feel it necessary. Windows users, as in everything else Microsoft, will have no choice but to follow, just as Mac users had no choice but to move to PPC.
I hope for the sake of sanity that Microsoft is as capable of handling the transition as Apple proved to be.