On Sat, Jan 22, 2005 at 07:45:37PM +0000, Douglas James Dunn wrote > I believe that 64 bit computing will just begin to gain momentum > when quantum computing makes it obsolete.
"The Cell" has been vapourware for a while. However, the STI group (Sony/Toshiba/IBM) have recently been granted a patent for it, so there is actual info available now. Nicholas Blachford has a very good writeup at http://www.blachford.info/computer/Cells/Cell0.html where he discusses it in detail. The only part I don't like is "Unfortunately the patent reads like it was written by a robotic lawyer running Gentoo in text mode, you don't so much read it as decipher it".<G> I resemble that remark, composing this email in mutt in a real text console. The suggested specs for The Cell are * 4.6 GHz * 1.3v * 85 Celcius operation with heat sink * 6.4 Gigabit / second off-chip communication ======================================================================= If over clocked sufficiently (over 3.0GHz) and using some very optimised code (SSE assembly), 5 dual core Opterons directly connected via HyperTransport should be able to achieve a similar level of performance in stream processing as a *SINGLE* Cell - Admittedly, this is purely theoretical and it depends on the Cell achieving it's performance goals and a "perfect" application being used, it does however demonstrate the sort of processing capability the Cell has. The PlayStation 3 is expected to have 4 Cells. ======================================================================= I repeat, 1 Cell = 5 dual core Opterons. 4 Cells on a PS3 will walk all over X86-based CPUs, or anything this side of an IBM z-series mainframe. Unlike "Quantum Computing", The Cell should be affordable very soon. 4 Cells per PlayStation is going to mean really *BIG* volume production, and the resultant economies of scale. Superiority won't guarantee adoption. The Amiga walked all over the IBM PC in every aspect except sales. However, the 8088/8086 CPU's address space was segmented to look like a bunch of Intel 8085's and it's Zilog-8000 (aka Z80) equivalant. This allowed very quick porting of a whole slew of CP/M business apps (Wordstar, Lotus 1-2-3, dBASE II, etc) and remember what the "B" in IBM stands for. There are two things that can kill Cell on the desktop... 1) Built-in DRM. The "S" in "STI Group" is Sony, who have a lot of media properties and "intellectual property". If they try to turn it into a "Fritz-chip" by having the DRM code unblockable, it's going to hurt sales. Remember the Pentium III serial number fiasco? 2) Microsoft, for only the second time in its history, puts out a new version of its OS that takes *LESS* CPU and RAM resources than its predecessor, making the extra power of Cell superflouous, except for gamers. OK, I'll admit to being in my fifties and experiencing... - the moaning and groaning about the switch from DOS 3 to DOS 4 and its heavier memory usage. - when DR-DOS came up with loading stuff into vacant memory between 640K and 1088K, leaving more low memory for DOS apps, its sales started taking off. The Windows-3-may-not-run-on-DRDOS FUD by MS was minor in comparison to the improved memory-management ability of MS-DOS 5, when it came to taking away DR-DOS marketshare. On the other hand, MS would just love DRM hardware to run Windows on, so that they can implement customer lock-in galore. If anything, I can see them rapidly abandoning Intel for Cell. We'd see an overnight shift from "Wintel" to "WinCell". Maybe the modifiable, free-wheeling linux that we know and love may run on the first-generation Cell, but I would urge people to be very, very afraid of Cell in the long run. Linux must be kept... a) portable, and b) lean and mean enough to run speedily on CPUs other than Cell. I hope that the Microsoft fanbois developing the GNOME and KDE "desktops" take this to heart. -- Walter Dnes <[EMAIL PROTECTED]> An infinite number of monkeys pounding away on keyboards will eventually produce a report showing that Windows is more secure, and has a lower TCO, than linux. -- [email protected] mailing list
