Apple's PowerMac G5 has been out of sale for nearly two years now, which some find a sad thing. As we all know, the desktop PowerPC market is more or less dead by now, which means getting your hands on a PowerPC workstation is either difficult, or very expensive. Terra Soft Solutions, the company behind Yellow Dog Linux, is about to launch its YDL PowerStation, the unofficial successor to the PowerMac G5. "Not just a simple replacement, but a well designed, perfectly packaged, readily upgradable, and far, far, more open source friendly system. The YDL PowerStation is four cores of unleashed Power in a solid, affordable package."
Post a Comment
Member since:
2005-07-10
I love the idea of this. Yellow Dog on an optimized machine would be very nice.
I wonder if you could get it to run IBM's Linux as well?
SLOF firmware:
http://www.openbios.org/SLOF
http://www-128.ibm.com/developerworks/power/pa-slof/
Power Systems: (IBM 970MP is not exactly POWER5)
http://en.wikipedia.org/wiki/POWER5
IBM Linux:
http://www-03.ibm.com/systems/power/software/linux/index.html
http://www-03.ibm.com/systems/browse/linux/?c=serversintro&n=Li...
Edited 2008-06-12 22:49 UTC
Member since:2008-06-12
IBM does not have it's own Linux. It certifies Red Hat (RHEL) and Novell (SLES) will work on it's systems. So I'm guessing they would work just fine if they use standard hardware that is well supported (as it sounds like they are). The only thing might be that although it is open source, SLES and RHEL may not include those firmware/drivers yet. If you are talking about AIX, I highly doubt it. I don't think IBM sells AIX unless you get it with their hardware or upgrade.
Also, It appears to me that it is a customized IBM system. The case looks very similar to a ThinkCentre/ThinkStation which IBM sold off to Lenovo. The front grill is very telling. http://shop.lenovo.com/SEUILibrary/controller/e/web/LenovoPortal/en...
I wouldn't be suprised if it wasn't an IBM designed machine that TerraSoft rebrands.
Actually, it looks identical to IBM's now discontinued Intellistation Pro Series:
http://images.google.com/imgres?imgurl=http://cdn.overstock.com/ima...
Identical case, just has a PowerPC chip inside instead of intel.
Member since:2006-09-30
Member since:2006-09-30
Member since:2006-08-02
Heh, I have this computer sitting next to me at work running AIX 5L
http://www.cce.de/homepage/siteupload/Intelli185/IntelliStation~*~@...
Member since:2006-01-03
Member since:
2005-07-06
Member since:2005-07-22
Member since:
2006-09-30
Member since:2005-07-12
Member since:
2005-07-07
Member since:2005-07-07
None. Practically all software that you run on PPC Linux has to be open sourced thus they will also be available on x86. On the other hand, there is a lot of stuff that runs on x86 that will not run on PPC. What made PPC Linux a no-go for me was the lack of commercial software (i.e. closed source) like GIS packages, MATLAB, Flash, multimedia codecs, etc. that ran on x86 Linux, Windows and Mac but didn't run on PPC Linux due to the lack of demand.
Member since:2006-01-22
But there is gnash, octave,scilab, yorick etc.
Why commercial software is a no-go? I cannot understand it. Even open-office is available there and icedtea.
The only no-go is PPC prices. They are far more standard efficient and useful than x86.
These crazy Romans!!!!
Edited 2008-06-13 10:21 UTC
Member since:2005-07-07
Have you ever used Matlab? Even ignoring Simulink for which there no alternative, the following are the features which I find Octave to be lacking in:
1) Plotting. Matlab totally trounces Octave when it comes to data analysis and plot manipulation. It's just much much easier to customize your graph in Matlab than it is in Scilab or Octave.
2) The IDE. The whole Matlab IDE makes debugging simple like in Visual Studio. Hovering your mouse over a variable brings up a tool tip showing its current value. The workspace enables you to have a quick look at what variables are set and what values they hold. The new versions come with on the fly syntax highlighting and checking, similar to what you get in Eclipse and Netbeans. Using Octave makes you think you're in the late 80s.
3) The performance. My code typically takes 5 - 10 seconds to run in Matlab. Octave took 30 minutes to run that exact same code. Apparently, having slightly complex loop with some branches inside causes Octave much grief. The solution suggested in the mailing lists was to rewrite the code in C and use the MEX interface. That slightly defeats the purpose since the whole reason I'm using Matlab is to rapidly prototype ideas before implementing them in C++. I might as well not bother if that's the solution to the problem.
4) Toolboxes. Matlab comes with loads which are of better quality and performance than anything that's free. Take neural network packages. Netlib runs on both Octave and Matlab. Matlab runs it slightly faster. However, Matlab comes with its own neural network toolbox and that is typically 10x - 20x faster than Netlib. And once you increase the size of your dataset to go from "toy" to "real world" scenario, Matlab is the only solution.
The whole reason for using a product like Matlab instead of coding up everything in C++ is to allow for rapid prototyping of algorithms. Mathworks understand that, which is why Matlab is the way it is. It is also why a professional license for Matlab costs thousands of dollars and yet they are the premier maths package for people engaged in numerical analysis.
People on these boards like to argue that you don't need Photoshop since Gimp is just as good, you don't need Office since OpenOffice is just as good, etc. That may be, that may not be and I'll leave it out to the various advocates to yell it out. You are sorely mistaken if you think that Octave is anywhere near being a replacement for Matlab.
Member since:2005-07-17
because PPC is still the rage in embedded systems so it's good to have a native box to work on. Also, many kernel level programmers consider Power PC an easier system to program advanced features for versus X86. They also like to have access to other platforms to keep things cross-platform.
Member since:2006-03-13
Member since:
2006-05-26
But in the real world, it's a solution in search of a problem. While the hardware seems capable/expandable enough from what this brief article states here on OSNews, x86-64 hardware is available much cheaper for the closest equivalent performance, and has better support in terms of being sure available software will build correctly for it and run. Also, the G5 is... another efficient combination of computational power and furnace. If it could run another more mainstream desktop OS (say Windows and/or OS X) it'd help - but not much. It's just too customized in too many ways to develop critical mass in terms of unit sales that would drive the price down to being more competitive in the market, and it doesn't even matter how "perfect" the machine is for the task. Unless a machine has monumentally greater capacity to do something than the common-slob hardware that's comparable for price, it's hard to justify why people will switch to it, as opposed to buying that common-slob hardware, which is more readily replaced/maintained and where software is more readily acquired and used. Creating a truly custom hardware base for an OS that's got a small part of the total software market just isn't commercially viable: did they not research Be, Inc. amongst others for a reference? Heck, even Apple went towards more commodity-based hardware, and their computers are still a small percentage of the market!
So, unless there's a lot more to this system than meets the eye in terms of unique software/hardware capabilities, it's commercially dead before it was born. A pity.
Member since:2006-01-03
Jonathan: "But in the real world, it's a solution in search of a problem."
It is a powerpc developer workstation; maybe your mom didn't quite wait for it, but many of us did.
Expensive, no support you say? I understand that you don't know how this industry operates, but it's exactly what we want.
Jonathan: "just isn't commercially viable"? Wow, with that kind of insight you could make a very good living in financial services. And spare us a couple of dumbass comments.
Member since:2006-01-11
Member since:2006-05-26
Yes, but my statement was about usage as a more common desktop box, or something that really requires market mass to sell in a commercially-viable way: I'm fully aware (unlike the idiot poster that asserts otherwise, that posted before you) of just how prevalent the PPC architecture is in consoles and other embedded markets: I've developed software for embedded systems, and keep up with that sort of thing. However, this machine simply isn't an embedded system that's sold by the millions and can compete on price, partially because the amount of CPU power to the power budget is right-sized with the CPU and SoC costs: it's effectively attempting to compete against an existing larger market of comparable machines for the desktop/workstation market. Historically, other than the Macintosh line, there's not been a single system that's sold in viable numbers for most businesses to survive and do well off of it. For economies of scale in the desktop/workstation/server realm for cheap boxes, AMD and Intel and other x86 manufacturers (and those that rely on that) have the price/performance ratio advantage: you need to make the hardware/software combination be especially cost-efficient for a stated task (like the Sun Niagara processors) to make meaningful inroads, and having the name and corporate backing of a long-established industry titan like Sun, IBM, etc. doesn't hurt, either, and this becomes especially important for a corporate environment for IT support: IT departments are notorious for not wanting to take more risk than they can avoid when it comes to providing software/hardware: what track record does this company have that'll stand out as a good easy "check, we've got that covered" experience? It used to be stated, "Nobody ever got fired for buying IBM" and that was for a good reason: they were big, not going away, and they were very good at supporting their customers. For that matter, they still are: credit where credit is due.
Other than the fact that IBM makes the processor, this company doesn't have a lot to back it up, based on computers sold in the last 15 years for their general market. It'll be a hard sell to make major sales numbers: I hope they're setup in such a way they can be profitable on low-volume sales: that's the most realistic thing they can expect, and it'd be nice to have a richer computer ecosystem than currently exists.
Member since:
2007-12-22
Member since:
2006-02-06
Member since:
2006-01-22
Member since:2006-03-17
Part of it is available in this discussion: http://www.powerdeveloper.org/forums/viewtopic.php?t=1559 on PowerDeveloper.org.
Member since:
2005-07-06
Member since:2006-01-04
Member since:
2005-07-12
Seriously, in 2002 this would have been hot ****, but today that's like ponying up that type of money for a dual 3ghz P4D... Though in a number of ways that's always been the G5's problem once the playing field of OS is removed from the equation.
With linux builds and benchmarks it truly starts to show how... dated the 970 series, more popularly known as the G5 is. In terms of raw computing power it's contemporaries were the 3ghz Galatin Xeons and the Opteron 244... Bleeding edge in 2001 - too bad the G5/970 didn't hit until '02
Compared to today - it shows it's age and anyone looking at this would be better served by most ANY modern dual core chips from either intel or AMD. For linux a phenom 9550 or Q6600 should flat out OWN this thing... So unless you are SPECIFICALLY looking at it for something like AmigaOS - I can't see a market for it at that price.
Seriously, $1900 for a pair of six year old dual core processors, a SINGLE 70 gig SAS drive, an equally dated and anemic video card that is little more than a rebranded Trident 9880 (there's a reason the XP10 was originally the Volari Mobile 8300), and a measly two gigs of RAM?!? Does the term 'ripoff' ring a bell? then for laughs they slap it in a leftover IBM case?
Hell, apart from the disk access model, a $750 dell or HP special is twice the computer...
Though that is where 80% of the price is coming from. SAS controllers will run you $600+, the drive is $350ish... Really those are the only parts of it WORTH anything in terms of buying a modern machine - and likely the only things what would make it FEEL fast. (though being 3gbps of a standard SATA exceeds the mechanical limits of a 15K RPM drive, you do have to question the point of SAS in a single or even dual drive setup - it only shines when you have MORE than two)
Member since:2006-02-15
then for laughs they slap it in a leftover IBM case?
I actually like the case. It's clean and it looks very sturdy, not like those cases with a gazillion blinky "cool" light effects and the case which breaks down completely if you accidentally nudge it over.. I rather have a functional and clean-looking case than something that sticks in the eye all the time. But well, ymmv 
Member since:2005-07-06
Member since:
2008-04-21
This discussion is total bull****. MatLab? Octave? Get real! Do you seriously think that anyone who would need to buy a Quad-G5, they would do it for running MatLab? Linux/PPC didn't fail for non-availability of MatLab, that's ridiculous. It failed because there was no hardware available (since Apple made the switch, Linux/ppc installations didn't increase substantially), while there was a need. The need is still there, and now there is also hardware -and there will follow more.
These CPUs are quite powerful even by today's standards, though they miss on efficienty (850W PSU is a bit much). Whoever will buy these, will not buy them for their 3d graphics power, nor for their software. But there are MANY applications apart from that. Has anyone done embedded development? It's far easier to do native compiles for ppc embedded CPUs rather than use cross-compiling. Plus, with Nvidia/ATI cards getting cheaper (and nouveau on ppc for that matter) I don't think it will take long to have other options for gfx cards. Were I to buy this system, I would most probably swap the XGI card for a nvidia card and experiment with nouveau (It's already usable on ppc, but just not 3d yet).
The only problem I see is the future evolution of the G5. Price is not a problem, but it would be sad to see this product as an one-off solution. At the very least if it succeeds it would show to IBM that there is a market and they should probably evolve G5 to next levels.
Edited 2008-06-13 13:47 UTC
Member since:2005-07-06
Member since:
2006-08-13
Member since:2005-11-17
Member since:2006-08-13
Member since:
2005-07-06
I'm confused by all this to be honest. For pure geek-love i would love to own one of these to go with my other alternative systems I collect.. but from a business point of view what's the benefit of running Linux on a PPC rather than X86? Is it that much more stable or something? Or is it only for programmers who want to compile PPC binaries for other platforms? Like when MS used PPC workstations to write XBOX360 games?
Member since:2005-10-12
Yes, its a good question. It may not have a rational answer. Don't it go back to the days when Apple was marketing PPC as a superior alternative, better quality, higher thruput processor? For a lot of people the move to Intel was a great shame. It was the loss of the one remaining source of distinctiveness. It was felt as selling out to one part of the evil duopoly that had deprived Apple of its rightful inheritance.
There are still people who simply will not buy a macintel, in the same way that no matter who the candidates are, there are some who will never vote Republican (or Democrat). They went to OSX fairly readily, because that was Apple. But buy from Intel, after looking at them with a mixture of loathing and contempt for so long? More time has to go buy before that becomes acceptable. And so the excitement about every attempt to revive PPC desktops.
Member since:
2006-05-09
Member since:
2006-03-27
Member since:
2008-04-07
This really reminds me of my IBM M Pro boxes, at least as far as looks.
I'm impressed by the spec except for:
*I hate dealing with Adaptec RAID devices. In my experience rebuilds are slow and the interface is terrible. They're very buggy under any Linux/BSD too. And for an "open" system, they should have gone with a more cooperative company.
*The specs are pretty terse. Exactly what video card do we get?
Maybe I could overlook those two things, use my own RAID card and graphics card, but that would make purchasing this a waste of at least $500 USD.
