| Power Developer https://powerdeveloper.org/forums/ |
|
| new PAsemi-based server blades https://powerdeveloper.org/forums/viewtopic.php?f=3&t=1436 |
Page 1 of 1 |
| Author: | markos [ Mon Nov 26, 2007 4:59 am ] |
| Post subject: | new PAsemi-based server blades |
Seems that PowerPC is not dead after all. http://www.ppcnux.com/?q=node/7141 at least in the mid/high-end markets. We only need a PowerPC workstation damn it! PS3 is nice, but it's non-expandable. Even so, if anyone would produce a Cell-based motherboard with SATA connectors and RAM expansion, even then it would be a winner! We _need_ a properly priced workstation! Konstantinos |
|
| Author: | SoundSquare [ Mon Nov 26, 2007 5:41 am ] |
| Post subject: | |
for what market ? There is no market for a PPC workstation. |
|
| Author: | Neko [ Mon Nov 26, 2007 12:58 pm ] |
| Post subject: | |
Quote: for what market ? There is no market for a PPC workstation.
I agree. Most people - the ones who get excited when they can buy a $200 1.5GHz Via PC with a crap version of Ubuntu on it - do not need one.However most PPC developers would kill for one. It's not like you can buy Apple PPC boxes that easily anymore.. eBay, refurbs, they are not ideal, and not that current. Unfortunately that market is tiny. Genesi pretty much sewed it up with the Pegasos. We know EXACTLY how many people will buy a new one. I'm not sure personally if it's worth making one. I'd love it if there WAS one, but in reality it'd have to do something else besides be a PPC developer workstation so that it became a profitable product. We certainly did not get any support from Power.org at the time we suggested this, and I doubt they will particularly impress with their upcoming G5 box (it is hard to imagine it being anything but unaffordable). |
|
| Author: | SoundSquare [ Mon Nov 26, 2007 1:20 pm ] |
| Post subject: | |
actually that's a neverending techno-centric topic. Of course the MorphOS users and devs (including myself) would love to see a new Pegasos-like machine popping up from BPlan/Genesi. It's a logical and legitimate wish but only a few users make the effort to look further the need and understand that Genesi (or any other hardware designer/manufacturer) has to make a living from it. I've been roving like a dharma geek on various hardware/software "scenes", more or less underground, (as a simple and curious user) and about the amiga/morphos one i've never seen such hardware-starving users. everytime a new PPC board is released, there are always a pack of starving amigawolves claiming they want AmigaOS4/MorphOS on it. That's the way we got trapped into PPC in a way. It would have been much different (and easier) with X86. What is truly missing in the amiga scene nowadays is a bit of lucidity. Of course dreams and hopes can be a great energy to move forward, but some must keep in mind the limits of alternative computing and economics toward it. Now i wish long life to the PPC, i'm sure there are still great things to do with it, like the Efika, if not a workstation. |
|
| Author: | Neko [ Tue Nov 27, 2007 6:32 am ] |
| Post subject: | |
Quote: actually that's a neverending techno-centric topic.
Lucidity would be nice :){snip} What is truly missing in the amiga scene nowadays is a bit of lucidity. Of course dreams and hopes can be a great energy to move forward, but some must keep in mind the limits of alternative computing and economics toward it. Now i wish long life to the PPC, i'm sure there are still great things to do with it, like the Efika, if not a workstation. There are plenty of uses for MorphOS and AmigaOS, and the more profitable ones are definitely NOT tending to the existing Amiga community. It's quite hard these days, though, to encourage companies to use MorphOS over and above readily available Linux solutions, even on the basis of performance. Vendors expect a lot more - bluetooth, wireless, hardware support, all ready and waiting and open source, which MorphOS does not give, and neither does AmigaOS, without some significant time and effort behind it. It is the best solution in some ways, but if you are on the clock, you will always be disappointed with the small development team and the small number of changes you can make to customise it to a market - compared to something like Linux which has 1000s of developers working on it, or even QNX or VxWorks which has a commercial backing. The second problem, then, is once you pick Linux instead of MorphOS, what is the point of staying with PowerPC? The same thing that gives PowerPC it's edge over x86 - you can design a device on PowerPC and run the same open source applications higher performance at lower power consumption with greater variety of SoC and discrete processor solutions - gives PowerPC no edge when the OS is processor agnostic to that degree. ARM, MIPS, even Blackfin is a contender now. The whole thing has to come as a solution package, now all we need to do is find some packaged solutions that work. This was the whole point of the Projects program, after all - interested users with great ideas can work on this and possibly bring something exciting to market. It has been a success, to some degree, as there have been some great projects which prove the usefulness of the Efika at least, and it has done a great job of highlighting all the bad points of the solution so we can work to fix them in a future revision.. you could not hope for such information to flow so freely in most markets. |
|
| Author: | markos [ Wed Nov 28, 2007 8:58 am ] |
| Post subject: | |
Quote:
The second problem, then, is once you pick Linux instead of MorphOS, what is the point of staying with PowerPC?
Well, if we can't answer ourselves that, then what's the point of staying with PowerPC in the first place. Myself? I find AltiVec and the Cell SPEs, something that just does not exist in any form in any other architecture, save perhaps the NEC vector computers which are basically SIMD mainframes.I've tried to convince others that exactly this feature of PowerPC should be exploited to its limits. But most have failed. It's not enough to have a couple of enthousiastic developers to work on such a thing. There has to be a strategy for this. It's not enough to sit and wait. But many people just wait. Most users seem to be comfortable in waiting the auto-vectorization in gcc. Hell, people compile everything with -ftree-vectorize and think they see a difference. Well, anyone that has done any non-trivial SIMD code will tell you that it will take a long long time before full auto-vectorization works, and even then i bet human SIMD code will outperform the compiler output. But then again, PowerPC might just not exist then. Well, here's an idea for a solution package. If EFIKA (1/2/3/whatever) would offer some kind of SIMD, I would port Blender or any other rendering software to utilize that SIMD engine, and one could chain 100s of efikas into a tiny cluster that consumes very little energy, offers great redundancy and performs greatly and at a very cheap price. With the proper custom PSU, 100 EFIKAs would not use more than 500 Watts of energy, and would outperform a PC cluster that consumes 100x the energy. But, alas efika doesn't use SIMD. Here's an idea, that I would actually try to sell, IF efika used SIMD. I'd put everything in a nice custom box, sth like a bladerack, with each blade carrying several efikas as replaceable modules. Hm, I actually start to like this idea. How much would it cost to produce a SIMD-ready headless pci-less EFIKA, with gigabit ethernet? And a PSU that would power 10s or more of these? Just a CPU module in fact. Seriously. Konstantinos |
|
| Page 1 of 1 | All times are UTC-06:00 |
| Powered by phpBB® Forum Software © phpBB Group http://www.phpbb.com/ |
|