h a l f b a k e r yGood ideas at the time.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Please log in.
Before you can vote, you need to register.
Please log in or create an account.
|
My idea is thus:
A computer in which each 256
bytes of memory is put into
"blocks" which each have their
own processor (very simple one of
course).
Even with such a small and simple
cpu, the computers power would be
amazing as it could have millions
of processors judging by the
amount
of memory in computers
these days AND with these
processors so small and simple
they will each run around a few
gigahertz themselves.
I have be working on this idea
for a while and my main problem
is how the communication between
blocks would work.. heirarchy,
chain, clusters?
Also, networking would be
wonderful on this as you just
have a link between the two
computers in the same way a link
is put between two groups of
memory in one computer. So..
what do you all think?
[link]
|
| |
Dollar $ign$ in your eye$, huh. Gooood luck, even if it could -teehee-happen your power bill would be a$tronomical e$p. in light of the power ogilopoly flexing $ome mu$cle $oon. In term$ of block$ of data between computer$ though, Internet 2 will not only allow you to regurgitate info a$ intended (chunk-$tyle) but will allow receiving end to get it'$ meal without getting $natches of crumb$ here and there. This will be accompli$hed through marking packet$. Too much infra$tructure has been laid for thi$ to happen $oon. No link for you, Nothing! $ee all tho$e dollar $ign$? |
|
| |
no need the link. ive been
keeping up with the internet 2
link for quite some time now :) |
|
| |
About the power bill.. nope.
wouldnt take that much power at
all.. whole fast, small, and
simple processor thing ya know. |
|
| |
I've got an old Connection Machine somewhere here in the back closet I'll sell ya. |
|
| |
Baked and not very workable. See: Connection Machine, "Intelligent RAM" (IRAM), massively parallel / distributed computing, etc.. |
|
| |
Problems: Most computation doesn't parallelize perfectly, communication becomes the bottleneck, reliability is difficult to achieve (without very complex fault tolerance systems), heat dissipation issues, etc.. |
|
| |
You say "networking would be wonderful on this as you just have a link between the two computers in the same way a link is put between two groups of memory in one computer"; I say "programming would be horrible on this as you have to worry about the link between two groups of memory in one computer just as much as a link between any two computers". |
|
| |
You also seem to assume that using a "small, simple processor" means you can run them at "a few gigahertz". Not true, or even close to it. |
|
| |
Small and simple processors are
faster. Why? Because they are
SMALL and the electrons have
much smaller distences to go,
makes sence, doesnt it? |
|
| |
Actually, simple processors are not necessarily faster. Much of the complexity in Pentium-class CPU's (from all vendors) has to do with allowing them to perform multiple operations simultaneously or out of sequence. If I write the code: |
|
| |
ld ax,[1234h]
ld [1236h],ax
ld bx,[si]
ld [si+2],bx
|
|
| |
If memory address 1234h doesn't happen to be cached, it may take many cycles to retrieve it from memory. On a Pentium, program execution would be suspended at the first instruction until that address was read; on today's processors, however, it's possible for the CPU to continue onward with the following instructions and process them while waiting for the first memory fetch to come back. Being able to process instructions while waiting for memory operations to complete saves a lot of time, but adds a lot of complexity. |
|
| |