Hypercomputers ...

Hypercomputers ...

Post by Tor Arne Bratl » Sun, 31 Dec 1899 09:00:00



Did you read the story on Hypercomputers at CNN? If not, read on ...

Star Bridge Systems claims having technology that enables them to
produce a computer 60.000 times faster than a PII 350-based PC, for only
$1000, and that it will be available in about 2 years. They also claim
it's capable of emulating just about every CPU there is... They are
announcing Hypercomputers for sale now, but in a much higher price range
(intended for current users of Supercomputer systems).

If AI could have a deal with these guys, the Amiga coming after the NG
could be awesome :)  (given that Star Bridge Systems are actually
capable of doing this, which I find kinda hard to believe, but you never
know ...).

Read about it at:
www.cnn.com/TECH/computing/9906/15/supercomp.idg/

Or at Star Bridge Systems homepage:
www.starbridgesystems.com

Enjoy :)
Tor Arne
--

Tor Arne Bratli
Games programmer
Norway

 
 
 

Hypercomputers ...

Post by Vanni Torel » Sun, 31 Dec 1899 09:00:00


Hi, Tor Arne Bratli! The *space is huge, still I spotted you writing
as follows:

Quote:> Star Bridge Systems claims having technology that enables them to
> produce a computer 60.000 times faster than a PII 350-based PC, for only
> $1000, and that it will be available in about 2 years. They also claim
> it's capable of emulating just about every CPU there is... They are
> announcing Hypercomputers for sale now, but in a much higher price range
> (intended for current users of Supercomputer systems).

  They want to make the gigs, don't they? :-)

Quote:> If AI could have a deal with these guys, the Amiga coming after the NG
> could be awesome :)  (given that Star Bridge Systems are actually
> capable of doing this, which I find kinda hard to believe, but you never
> know ...).

  Reconfigurable computing (using FPGAs to implement a chip design "on
the fly") is a pretty old area of research (several years, starting from
early 80s). Whether or not this SBS will suceed, there's little doubt
this will be a viable technology for future computers (AmigaNG and non-
AmigaNG as well). Indeed, it is rumoured that the mythical Transmeta
project will be a low-cost RCU with algoryhtm optimized specifically for
Windows NT 5.0.

--

. ..................................................... ... .. .
  \ _/__      vanni  torelli  mind  soul  whatever  in  between
   \/         electronic engineering student  -  audio engineer

. .. ... ..................................................... .

 
 
 

Hypercomputers ...

Post by Tor Arne Bratl » Sun, 31 Dec 1899 09:00:00



> Hi, Tor Arne Bratli! The *space is huge, still I spotted you writing
> as follows:

[snip]

Hi Vanni,

Quote:>   Reconfigurable computing (using FPGAs to implement a chip design "on
> the fly") is a pretty old area of research (several years, starting from
> early 80s).

[snip]
it seems you're quite right on that one; I talked to a friend of mine that's
working in a tech-company after writing the first piece to this newsgroup, and
he says they use FPGA's for specialized types of algorithms in their realtime
systems to obtain what he referred to as "speeds far higher than CPU-based
solutions, and cheaper too". But, he also thought that making the technology
work for general purpose computing would be a tough proposition (as a lot of
the experts that where quoted in the articles related to SBS' technology
seemed to think). I really hope they make it, would be awesome having that
kind of power on (or under) the desktop ;)

Quote:>Indeed, it is rumoured that the mythical Transmeta
>project will be a low-cost RCU with algoryhtm optimized specifically for
>Windows NT 5.0.

Given the fact that Linus Thorvalds is working there I should think the
processor would have a reasonable degree of performance in Linux as well - and
who knows, it might be a kick-ass processor for Amiga NG, although some people
probably wouldn't appreciate the "link" to the x86 side of things...

Tor Arne :)

--
Tor Arne Bratli
Games programmer
Norway

 
 
 

Hypercomputers ...

Post by Vanni Torel » Sun, 31 Dec 1899 09:00:00


Hi, Tor Arne Bratli! The *space is huge, still I spotted you writing
as follows:

Quote:> it seems you're quite right on that one; I talked to a friend of mine that's
> working in a tech-company after writing the first piece to this newsgroup, and
> he says they use FPGA's for specialized types of algorithms in their realtime
> systems to obtain what he referred to as "speeds far higher than CPU-based
> solutions, and cheaper too". But, he also thought that making the technology
> work for general purpose computing would be a tough proposition (as a lot of
> the experts that where quoted in the articles related to SBS' technology
> seemed to think). I really hope they make it, would be awesome having that
> kind of power on (or under) the desktop ;)

  FPGAs can easily reproduce logic ports and transfer functions, but they
do lack features like registers and high speed arithmetic units
(multipliers or adders). Some companies are devising CPUs which have the
best of the two worlds, the features of a standard CPU + an FPGA area
that can be reconfigured on the fly to meet specific computational
purposes. However, this is just pioneering technology. I read the SBS
page after your link, and they seem to be delivering a general purpose
supercomputer by means of several RPU board (36, I read) in parallel.
Such an architecture may effectively cater for the power and flexibility
of a standard processor with incomparably higher speeds, but don't forget
the cost, way above the home user.

  Anyway who knows, an RPU could even be chosen as the MMC we have been
rumouring eversince. Whatever the AmigaNG will have, I would only be glad
if they could redefine the paradigm of home computing (and computing in
general) just as they did years ago with the A1000.

Quote:> Given the fact that Linus Thorvalds is working there I should think the
> processor would have a reasonable degree of performance in Linux as well - and
> who knows, it might be a kick-ass processor for Amiga NG, although some people
> probably wouldn't appreciate the "link" to the x86 side of things...

  I haven't heard about Linux optimization: after all Windows NT 5 is
Linux direct rival, so to optimize the chip for two competitor OS would
make little sense.

--

. ..................................................... ... .. .
  \ _/__      vanni  torelli  mind  soul  whatever  in  between
   \/         electronic engineering student  -  audio engineer

. .. ... ..................................................... .

 
 
 

Hypercomputers ...

Post by Tor Arne Bratl » Sun, 31 Dec 1899 09:00:00



>   I haven't heard about Linux optimization: after all Windows NT 5 is
> Linux direct rival, so to optimize the chip for two competitor OS would
> make little sense.

You're probably right; I was just guessing anyway. What I'm a bit curious about is
what kind of optimization Transmeta could make that are OS-specific(?); I would
think that any performance increasing stuff they add in should be possible to make
use of in Linux as well, even though it might require compiler / kernel updates. Or,
shouldn't it? Please enlighten me :) (Linux is one of my favourite OS'es, and I
would like to see it do well, almost as well as the NG... :)

Tor Arne
--
Tor Arne Bratli
Games programmer
Norway

 
 
 

Hypercomputers ...

Post by Vanni Torel » Sun, 31 Dec 1899 09:00:00


Hi, Tor Arne Bratli! The *space is huge, still I spotted you writing
as follows:

Quote:> You're probably right; I was just guessing anyway. What I'm a bit curious about is
> what kind of optimization Transmeta could make that are OS-specific(?); I would
> think that any performance increasing stuff they add in should be possible to make
> use of in Linux as well, even though it might require compiler / kernel updates. Or,
> shouldn't it? Please enlighten me :) (Linux is one of my favourite OS'es, and I
> would like to see it do well, almost as well as the NG... :)

  Not a clue! :-) I don't work for them, nor did I ever read anything
that went further than those (quite noisy indeed) rumours. Personally, I
think actual technology is still too bound to ASICs, at least due to pure
cost-performance reasons, and an ASIC is what we should expect as the
MMC. Consider that the first machines to be launched by GW2K will be a
game console and an information system appliance, which focus on two
pretty narrow applications that request extreme competitively priced
devices. So again, I see little sense in endowing  those machines with an
RCU. As for the real AmigaNG we all are longing, that could be a viable
solution and, who knows...GW2K's plans to release the big computers later
on will give them a little more scope for a wiser choice.

--

. ..................................................... ... .. .
  \ _/__      vanni  torelli  mind  soul  whatever  in  between
   \/         electronic engineering student  -  audio engineer

. .. ... ..................................................... .