|Verfasst am: Di, 28.06.2005 07:37 Titel: Disillusions daily: Cell is not going to happen (engl.)
Ein kleiner Artikel auf englisch von AmigaWorld, warum es laut Verfasser unwahrscheinlich sei, daß der Cell-Chip eine Zukunft auf dem Desktop-Computer-Markt hätte.
|My initial enthusiasm regarding Cell has vanished almost completely. I followed a Cell seminar held in Barcelona via internet broadcast (http://www.power.org) and I visited a Cell seminar held at the Linuxtag (http://www.linuxtag.de) in Karlsruhe. Thanks to these seminars there are now more open questions than answers, and my prospect is that Cell is not going to happen on the Amiga or any other desktop computer.
Signals sent out by IBM are inconsistent and fragmentary to say the least. There were never clear ambitions towards a desktop future, but now I think that Cell may even turn out as a completely remote development that brills on the Playstation 3 for a couple of years and is discontinued afterwards.
Currently it is impossible for anybody outside large corporations to approach IBM for anything related to Cell; my impression is that they are still somewhat paralyzed from the Apple move and the damage done to the PPC reputation. It will be a big problem selling Cell for anything outside game consoles; a paradigm shift in programming is due to fully exploit the architecture, and IBM tried to propose such a thing on the Barcelona show. However, as it stands today, Cell is only a Power core with 8 coprocessors and a large company being unable to outline a possible use for it. As it has been demonstrated in Karlsruhe, Linux doesn't cut it; it can use the Cell only as a bank of coprocessors throughout file descriptors, which is the typical grue on Unix. IBM also demonstrated their lack of concept and the embryonic state that even the combination of IBM's prodigies (Cell and Linux) persists in at this time. There were "thoughts" about task virtualization on Cell under Linux, according to the developer running the seminar. Thank you, I had these thoughts myself.
Cell-based development workstations are no longer an issue as it seems, instead IBM showcased a motherboard for blades with its suggested use under Linux. That blade could crunch vast numbers of single precision floats -- provided that anybody would manage to find a task artificial enough to be encoded not dissimilar to PPC assembler under PowerUp, only 8 times more complicated and awkward (per Cell, of which there are two on the blade).
Anybody should reconsider if Cell is really a possible way out of the processor dilemma. The trap has been 68k and is now becoming PPC; If we can only replace one obsolete CPU with another then I would suggest to rebuild the next AmigaOS in code virtualization entirely, in 68k virtualization more precisely. The past and the future of the Amiga is 68k.
Cell-Chip auf dem Linux-Tag: http://www.pro-linux.de/news/2005/8313.html
kleiner Auszug aus dem Artikel
|Allerdings glaubt auch er nicht, dass Sony die PS3 pünktlich veröffentlichen kann. Angesichts der Stärken im Single Precision Floating Point-Bereich scheint der Cell eine etwas seltsame Wahl für eine Spielkonsole: Sony lässt sich ja schon von NVIDIA einen Mörder-Grafikchip bauen, ob der Cell bei Gameplay-Berechnungen mit den 3 PowerPC-Cores der XBox 360 mithalten kann war uns beiden ein Rätsel und nicht endgültig zu klären. Sony arbeitet jedenfalls ziemlich abgeschottet, wie auch die Microsoft-CPU-Entwickler bei IBM, und scheint lediglich die Prozessoren zu kaufen. |
http://webwood.de/winuae_1/index.html winuae download