The 21st version of this list does not show the SETI@Home project. The top entry is NEC at 35 terraflops. Today's SETI@Home average for the last 24 hours is 61 terraflops. It may be a virtual supercomputer, but it is producing real results.
There is. Someone has to want to scan the
document, run OCR, and edit it. It's pretty
time consuming. I know I wouldn't do that
for the Learned Elders. But, perhaps there
is someone who would.
I read Gutenberg books on my Palm with Weasel.
When I spot an error, I create a bookmark.
When I'm done with the book, I edit the
original (on my computer) and send context
diffs to the project. So, I'm only checking
books that I want to read anyway.
I think of myself as a pretty good proof
reader. However, in one of the children's
books that I've been
reading to my son, I continued finding new
minor glitches even after five readings.
If software has a bug per thousand lines,
despite the code being checked for syntax
and function automatically, imagine how
many errors there are in unchecked text.
This is an area where tens of thousands
of people working in an open environment
can contribute to quality, crushing any
proprietary effort.
I used to use
a Japanese Soroban (abacus). I've got these
two books (out of print now) with instructions.
It takes a couple months of 20 minutes a day
to get good at it. In the end, you can perform
mental arithmetic.
I once computed a 9 digit trig function (a sin())
using the Taylor series expansion (and converting
degrees to radians) in my head. It took
about a half hour - and it was correct.
I don't claim any special genius. Anyone
can learn this. I still believe it should
be taught in school. It's easier, faster
and much more reliable.
I ran a SETI@home unit on a 486/33. It was primarily an experiment to see if it would work. It did. The machine has 16 MB RAM, and 170 MB disk. I ran Peanut Linux on it. I stripped the kernel down a bit (I recompiled the kernel on a faster machine for speed - but I used to recompile kernels, etc., on my 486sx/25 laptop that also had 170 MB disk. That machine died, however). Running the SETI unit did cause the machine to swap a little. SETI's working set is something like 13 MB. Anyway, it took 26 days to run - but the results were accepted!
Yes, SETI@Home can run in 16 MB RAM. The effort was worth a few geek points.
I still run a Macintosh II. When I got it,
2 MB RAM was much more than needed, and 40 MB
disk was infinite space. You have to understand
that the Mac II was 5 times faster than any
other Mac at the time, and everything was written
to run from 800 KB floppies. 16.7 MHz 68020.
Now it has 20 MB RAM and 1.5 GB disk. It runs
Word 5, Canvas. The thing is, it's faster than 'modern' machines, due to bloatware. My Athlon 1800+ with 512 MB RAM is ostensibly 1000x
(or so) faster -
but the Mac starts things up quicker.
Disks are only a little faster now.
IDE hasn't hardly caught up with SCSI (before II).
Abiword claims tables now - maybe I'll switch
soon.
It made the leap to the internet in 1995 at age 8.
I'm stuck at Netscape 2, however.
It was just a year ago, or so, that I finally
moved my email to a new machine.
I have some childrens (educational) games that
run under DOS.
I have a 486/33 that runs them fine. I let
my 6 year old play with my modern machine
under my direct supervision only. Otherwise,
I could let him run them under DOSemu.
I programmed it with a Pascal language that
was incredibly sensitive to white space. It
used BASIC as an editor.
I also programmed it in C, using the C Power
compiler. It was a solid compiler, produced
the highest performance code of any language
I used (but I never wrote assembler for 6502).
It's main flaw was that it did not support
longs (32 bit ints).
I was just starting to do music via MIDI with
it when it died.
However, I remember enough of it to recall
how limited it was. Not all memories are fond.
In the US, if one company patents a compound,
say Aspirin, they must say what it's good for.
Let's say it helps headaches. Another company
is allowed to patent the same compound if
they can come up with a use for it - say
curing cancer.
Since doing all the toxicity studies has
already been done, drug companies are
interested in coming up with new uses for
old stuff. Therefore, they have an incentive
for trying to figure out new uses for stuff.
Incentives are what patents are all about.
That doesn't mean I think we ought to have
software patents. It can cost a billion
dollars to get a drug to market before the
first pill is sold.
The real problem is that the Webb scope (NGST)
does not replace HST. It is not a visible light
instrument. NASA does not have the funds to
keep HST going. No real surprise, it's expensive.
We keep putting up very expensive things: HST,
the Shuttle, ISS - it seems that a few expensive
things are easier to sell than a bunch of
cheap things.
It costs about $250 million per service flight.
Instead of servicing Hubble, send up a new
scope every five years or so.
How much scope can you get for $250 million?
If you drop IR, then you don't need active
cooling. This saves complexity and lots of
weight. It also means that the mission does
not degrade when you run out of coolant.
The Webb (NGST) and SIRTF cover this spectral
range. Use passive cooling. Use the solar
panels to shade the instrument. Don't have
them hanging out to wobble.
Next, use a single camera. For example, use
just an ACS like camera. This saves complexity
and weight.
Next, use the new lightweight mirror tech. It's
going in space, you don't need a thick mirror
because in microgravity, it won't bend under
it's own weight. Remember that it's $10,000
per pound to get it into space. Anything
you put there is worth more than platinum.
Use an offset Newtonian design, to eliminate
diffraction spikes.
I'd bet that a 2 meter scope could be on orbit
for under $200 million, mostly launch costs.
It could out perform HST, and stay on orbit
for a minimum of 5 years.
Canada recently put up a cheap space telescope.
MOST. Check it out.
The mirror article has a nasty headline, and
the first few paragraphs are overstated. The
rest of the article has the real information,
like how QQ47 probably would have a zero chance
in a few weeks (it had a zero chance the next
day).
It should be noted that jpl lists two Torino
scale "1" objects at the moment. Neither have
possible impacts any time soon... Should practical
imortality become available soon, I'll pay
more attention.
This one was examined in 1997 for 27 days.
One reason for no current data points is that,
as far as I can tell, it is lost in the
glare of the Sun. As near as I can tell
from the Minor Planet Center, 1997 XR2
might be spotted again in April or May of 2004.
As happens in these cases, as more observations
are made, the uncertainty in the orbit is reduced,
and the probabilities of impact tend to decrease.
Impact Risk for QQ47
http://neo.jpl.nasa.gov/risk/2003qq47.html
2014 has disappeared from the list, since the
probability of impact has gone to zero. There
are still 18 possible impact dates. The most
likely, at one in 5.2 million is in 2067.
If, in six months, 2067 is still in the list,
it's probability of impact will be greater.
Then, we'd have 64 years to do something about
it.
Everyone talks about the weather, but no one
does anything about it.
There's no shortage of problems to solve
or re-solve. At the moment, there's a big
push to outsource out of the US. While
outsourcing can work, I haven't seen it happen
yet. It doesn't matter if it's cheaper to
fail, it's still failure. When the current
fad runs its course, there will again be
plenty of jobs in this industry.
If Mr. Gates could distribute reliable products,
there'd be alot less jobs available.
NOTHING is of a higher and different state of
energy.
He's in good company.
The ontological argument goes something like this:
God is defined as a perfect being.
Nonexistence is an imperfection.
Therefore God exists.
My favorite formulation goes like this:
Nothing is better than life in Heaven.
A Ham sandwich is better than nothing.
Therefore, a Ham sandwich is better than
life in heaven.
At best, he's confusing energy and power.
The red shift does not destroy energy, but it
does reduce power. If you integrate the lower
power over time, you still get the same total
energy for the same photons.
In the original Big Bang, the photons aren't
redshifted by the stretching the fabric of the
universe, they're redshifted because, due to
expansion, the further away things are, the faster
they are receeding from our vantage point.
It isn't clear to me that today's dark energy
theories suggest any additional red shift.
I'm pretty sure that I do not want my toaster
powered by fusion explosions, even if they are
underground.
I knew my mom-in-law would want tech support.
Even though she would call me on her nickle,
I wanted to minimize it. So, I had her get
a Mac.
Little did I know, but Apple ships these things
with IE and Outlook. The only questions
have been about IE and Outlook. I have replaced
IE with Netscape, but have not replaced the email
client as yet.
No. If we wanted to do things in space cheaper,
we'd send robots.
If we're going to send humans into space which
costs easily 10x as much, it must be to learn
how to get people to space. Sure, we're going
to have accidents. If we don't learn from
them, there's no point in going.
It is interesting that in two shuttle failures,
we have discovered two failure modes. Now,
it may be that the shuttle is old, or that
the design is not fixable, and that this program
should be ended. As long as the next idea
benefits from the lessons, it's OK.
On the other hand, ISS has failed to advance
space station design, from any perspective.
This shows that we don't always learn from
our mistakes.
Sure, it's ten years since version 0.10. But
this isn't the first browser. The change log
talks about the previous version - 0.9 (which
sounds like 0.09 to me) and it
talks about some added features, but it already
was ported to several flavors of Unix.
It sounds as if it had achieved some maturity
in a prior version.
From 1989, we'd be at year 14, more or less.
I'd rather go with octal. 100 bits.
Hex has those alpha characters. If I have
a table of data, and I examine it programatically,
for example, so that numbers are right justified
and text is left justified, it's nice to
just look for digits. With hex, you either
have to use a lead zero (ugly) or run the
risk that your number looks like a word: like
BAD (5655 octal).
Can anyone doubt that 80% of the silicon is for
supporting legacy apps at this point?
I can. Die sizes have increased in area
exponentially. Lots of the area has to do
with speeding floating point, increasing
cache size, doing register renaming, etc.
The fraction of area used for the instruction
set is less and less as time goes by. Even
for the x86.
In the 16 bit to 32 bit transition, we didn't
always get more speed. For example Digital
had the PDP-11 (16 bit) and the Vax 11/780
(32 bit). The '780 had a PDP-11 compatibility
mode - it could run PDP-11 binaries as is
(except that they didn't support floating point
for PDP-11 binaries, for no real apparent reason).
Now, the PDP 11/70 was a little cheaper than
the new Vax 11/780. The compatibility mode
in the '780 was about as fast as the 11/70.
However, in my benchmarks from the time,
my 11/70 best times were about 20% faster
than my best 11/780 times.
Digital wanted their customers to migrate, and
though they came out with a PDP 11 on a chip
that was cheaper, they never came out with
a PDP 11 that was faster. Even so, the PDP 11
customers kept buying them for years.
The 16 bit data registers, data paths, etc.,
simply take up less die area on the chip.
So, it should be possible to make the 16 bit
processor faster, on any given chip technology
than a wider 32 bit design. But, the memory
size problem becomes a barrier to many problems.
On the other hand, I did have apps in those
days that really needed 32 bit addressing and
integer manipulation. These apps were faster
(or possible) on the Vax. On the whole, I
wanted to move forward.
I have over 512 MB RAM on my 5 year old Pentium II. I have apps (mostly that I've written myself)
that would be more capable if I could stuff
more RAM on the box, and if I could address it.
I want the new 64 bit chip now - even if I take
a speed hit - as long as the hit isn't too big.
For some of my apps, real 64 bit integers will
improve speed, overall.
I run Linux, and can, if I want to, recompile
everything. So, I'm ready to migrate to the
Alpha (64 bits) right now. But, I don't have
the budget (except perhaps in the used market).
I'd really like a commodity 64 bit capable
x86 chip.
Right now, 100% of desktop apps can be run in
32 bits. That's because 64 bits aren't generally
available. Back in the 16 to 32 bit transition,
there were lots of apps that were hard to write
in 16 bits. It wasn't that hard to write a
program that, when compiled, occupied more than
65536 bytes. In this transition, it's harder
for the instruction space to matter. 4 GB is
alot of code - even for bloatware. So, today's
big apps are mostly data. Think databases, big
simulations, etc.
When the 68000 came out, it had 32 bit
registers, but only 24 bits went out of the
chip to physical memory. So, it could only
address 1 MB RAM. Later models (like my
68020 Mac) can address 32 bits of physical
RAM (4 GB). But it took several years before
RAM prices came down so that I could afford
even 32 MB RAM (my Mac II's maximum). That's
only 25 bits.
The AMD chip will support 40 bits physical, and
a few more bits of virtual. So, a process will
be able to be larger than physical RAM.
Still, 40 bits of RAM is a terrabyte. That's
about $100,000 US of RAM at the moment, assuming
you can get a motherboard to support it.
AMD will be able to bump up the virtual and
physical RAM supportable without changes to
existing code - right up to 64 bits. That's
what Motorola did. At the moment, supporting
more physical bits would have non-zero cost,
with zero benefit.
One thing that additional virtual memory could
do for a large corporation is aid in LAN
distributed computing. Basically, one could
have a shared RAM dataset span hundreds or
thousands of desktop workstations. All machines
would have access to all of the data over the
network. The OS would translate virtual address
references to physical - including finding
the machine that has that chunk in it's RAM.
The OS for the machine that has that chunk of
RAM would coordinate changes to that RAM.
This could allow a corporation to utilize their
desktop investment for many supercomputing
projects - reducing their supercomputing needs,
saving them money.
It also allows one to more easily build a modern
supercomputer with the chipset.
The Commodore 128 used a funky bank switch
system to allow access to 128 KB RAM. This
is kinda/sorta how Intel wants to extend
the x86. IMO, this cripples the arch.
Intel seems to be doing it because they want
the Itanium to take off. However, the Itanium
appears to be an expensive dog. I'd rather
have any of it's 64 bit competitors, at the
moment. It appears to me that the Itanium
will cripple Intel, without regard to
the backing of big corporate customers.
Slashdot Mirror
The 21st version of this list does not
show the SETI@Home project. The top entry
is NEC at 35 terraflops. Today's SETI@Home
average for the last 24 hours is 61 terraflops.
It may be a virtual supercomputer, but it
is producing real results.
There is. Someone has to want to scan the document, run OCR, and edit it. It's pretty time consuming. I know I wouldn't do that for the Learned Elders. But, perhaps there is someone who would.
I think of myself as a pretty good proof reader. However, in one of the children's books that I've been reading to my son, I continued finding new minor glitches even after five readings.
If software has a bug per thousand lines, despite the code being checked for syntax and function automatically, imagine how many errors there are in unchecked text. This is an area where tens of thousands of people working in an open environment can contribute to quality, crushing any proprietary effort.
This could be broken down by religeon, income, and various other stats to check for biases.
I once computed a 9 digit trig function (a sin()) using the Taylor series expansion (and converting degrees to radians) in my head. It took about a half hour - and it was correct.
I don't claim any special genius. Anyone can learn this. I still believe it should be taught in school. It's easier, faster and much more reliable.
primarily an experiment to see if it would work.
It did. The machine has 16 MB RAM, and 170 MB
disk. I ran Peanut Linux on it. I stripped the
kernel down a bit (I recompiled the kernel on
a faster machine for speed - but I used to
recompile kernels, etc., on my 486sx/25 laptop
that also had 170 MB disk. That machine died,
however). Running the SETI unit did cause the
machine to swap a little. SETI's working set
is something like 13 MB. Anyway, it took 26 days
to run - but the results were accepted!
Yes, SETI@Home can run in 16 MB RAM.
The effort was worth a few geek points.
Now it has 20 MB RAM and 1.5 GB disk. It runs Word 5, Canvas. The thing is, it's faster than 'modern' machines, due to bloatware. My Athlon 1800+ with 512 MB RAM is ostensibly 1000x (or so) faster - but the Mac starts things up quicker. Disks are only a little faster now. IDE hasn't hardly caught up with SCSI (before II). Abiword claims tables now - maybe I'll switch soon.
It made the leap to the internet in 1995 at age 8. I'm stuck at Netscape 2, however.
It was just a year ago, or so, that I finally moved my email to a new machine.
I have some childrens (educational) games that run under DOS. I have a 486/33 that runs them fine. I let my 6 year old play with my modern machine under my direct supervision only. Otherwise, I could let him run them under DOSemu.
I also programmed it in C, using the C Power compiler. It was a solid compiler, produced the highest performance code of any language I used (but I never wrote assembler for 6502). It's main flaw was that it did not support longs (32 bit ints).
I was just starting to do music via MIDI with it when it died.
However, I remember enough of it to recall how limited it was. Not all memories are fond.
Since doing all the toxicity studies has already been done, drug companies are interested in coming up with new uses for old stuff. Therefore, they have an incentive for trying to figure out new uses for stuff.
Incentives are what patents are all about.
That doesn't mean I think we ought to have software patents. It can cost a billion dollars to get a drug to market before the first pill is sold.
We keep putting up very expensive things: HST, the Shuttle, ISS - it seems that a few expensive things are easier to sell than a bunch of cheap things.
It costs about $250 million per service flight. Instead of servicing Hubble, send up a new scope every five years or so. How much scope can you get for $250 million? If you drop IR, then you don't need active cooling. This saves complexity and lots of weight. It also means that the mission does not degrade when you run out of coolant. The Webb (NGST) and SIRTF cover this spectral range. Use passive cooling. Use the solar panels to shade the instrument. Don't have them hanging out to wobble.
Next, use a single camera. For example, use just an ACS like camera. This saves complexity and weight.
Next, use the new lightweight mirror tech. It's going in space, you don't need a thick mirror because in microgravity, it won't bend under it's own weight. Remember that it's $10,000 per pound to get it into space. Anything you put there is worth more than platinum.
Use an offset Newtonian design, to eliminate diffraction spikes.
I'd bet that a 2 meter scope could be on orbit for under $200 million, mostly launch costs. It could out perform HST, and stay on orbit for a minimum of 5 years.
Canada recently put up a cheap space telescope. MOST. Check it out.
It should be noted that jpl lists two Torino scale "1" objects at the moment. Neither have possible impacts any time soon... Should practical imortality become available soon, I'll pay more attention.
http://neo.jpl.nasa.gov/risk
Or maybe not.
Impact Risk for QQ47
http://neo.jpl.nasa.gov/risk/2003qq47.html
2014 has disappeared from the list, since the probability of impact has gone to zero. There are still 18 possible impact dates. The most likely, at one in 5.2 million is in 2067.
If, in six months, 2067 is still in the list, it's probability of impact will be greater. Then, we'd have 64 years to do something about it.
Everyone talks about the weather, but no one does anything about it.
If Mr. Gates could distribute reliable products, there'd be alot less jobs available.
I wanna see those quarks
Photons are quarks. Your eye can detect single photons, unaided.
My favorite formulation goes like this:
At best, he's confusing energy and power. The red shift does not destroy energy, but it does reduce power. If you integrate the lower power over time, you still get the same total energy for the same photons.
In the original Big Bang, the photons aren't redshifted by the stretching the fabric of the universe, they're redshifted because, due to expansion, the further away things are, the faster they are receeding from our vantage point.
It isn't clear to me that today's dark energy theories suggest any additional red shift.
I'm pretty sure that I do not want my toaster powered by fusion explosions, even if they are underground.
Their engine is tested horizontally, mounted on a wheeled trailer. See image
I knew my mom-in-law would want tech support. Even though she would call me on her nickle, I wanted to minimize it. So, I had her get a Mac.
Little did I know, but Apple ships these things with IE and Outlook. The only questions have been about IE and Outlook. I have replaced IE with Netscape, but have not replaced the email client as yet.
It is interesting that in two shuttle failures, we have discovered two failure modes. Now, it may be that the shuttle is old, or that the design is not fixable, and that this program should be ended. As long as the next idea benefits from the lessons, it's OK.
On the other hand, ISS has failed to advance space station design, from any perspective. This shows that we don't always learn from our mistakes.
Sure, it's ten years since version 0.10. But this isn't the first browser. The change log talks about the previous version - 0.9 (which sounds like 0.09 to me) and it talks about some added features, but it already was ported to several flavors of Unix. It sounds as if it had achieved some maturity in a prior version. From 1989, we'd be at year 14, more or less.
I'd rather go with octal. 100 bits. Hex has those alpha characters. If I have a table of data, and I examine it programatically, for example, so that numbers are right justified and text is left justified, it's nice to just look for digits. With hex, you either have to use a lead zero (ugly) or run the risk that your number looks like a word: like BAD (5655 octal).
I can. Die sizes have increased in area exponentially. Lots of the area has to do with speeding floating point, increasing cache size, doing register renaming, etc. The fraction of area used for the instruction set is less and less as time goes by. Even for the x86.
In the 16 bit to 32 bit transition, we didn't always get more speed. For example Digital had the PDP-11 (16 bit) and the Vax 11/780 (32 bit). The '780 had a PDP-11 compatibility mode - it could run PDP-11 binaries as is (except that they didn't support floating point for PDP-11 binaries, for no real apparent reason).
Now, the PDP 11/70 was a little cheaper than the new Vax 11/780. The compatibility mode in the '780 was about as fast as the 11/70. However, in my benchmarks from the time, my 11/70 best times were about 20% faster than my best 11/780 times.
Digital wanted their customers to migrate, and though they came out with a PDP 11 on a chip that was cheaper, they never came out with a PDP 11 that was faster. Even so, the PDP 11 customers kept buying them for years.
The 16 bit data registers, data paths, etc., simply take up less die area on the chip. So, it should be possible to make the 16 bit processor faster, on any given chip technology than a wider 32 bit design. But, the memory size problem becomes a barrier to many problems.
On the other hand, I did have apps in those days that really needed 32 bit addressing and integer manipulation. These apps were faster (or possible) on the Vax. On the whole, I wanted to move forward.
I have over 512 MB RAM on my 5 year old Pentium II. I have apps (mostly that I've written myself) that would be more capable if I could stuff more RAM on the box, and if I could address it. I want the new 64 bit chip now - even if I take a speed hit - as long as the hit isn't too big. For some of my apps, real 64 bit integers will improve speed, overall.
I run Linux, and can, if I want to, recompile everything. So, I'm ready to migrate to the Alpha (64 bits) right now. But, I don't have the budget (except perhaps in the used market). I'd really like a commodity 64 bit capable x86 chip.
Right now, 100% of desktop apps can be run in 32 bits. That's because 64 bits aren't generally available. Back in the 16 to 32 bit transition, there were lots of apps that were hard to write in 16 bits. It wasn't that hard to write a program that, when compiled, occupied more than 65536 bytes. In this transition, it's harder for the instruction space to matter. 4 GB is alot of code - even for bloatware. So, today's big apps are mostly data. Think databases, big simulations, etc.
When the 68000 came out, it had 32 bit registers, but only 24 bits went out of the chip to physical memory. So, it could only address 1 MB RAM. Later models (like my 68020 Mac) can address 32 bits of physical RAM (4 GB). But it took several years before RAM prices came down so that I could afford even 32 MB RAM (my Mac II's maximum). That's only 25 bits.
The AMD chip will support 40 bits physical, and a few more bits of virtual. So, a process will be able to be larger than physical RAM. Still, 40 bits of RAM is a terrabyte. That's about $100,000 US of RAM at the moment, assuming you can get a motherboard to support it. AMD will be able to bump up the virtual and physical RAM supportable without changes to existing code - right up to 64 bits. That's what Motorola did. At the moment, supporting more physical bits would have non-zero cost, with zero benefit.
One thing that additional virtual memory could do for a large corporation is aid in LAN distributed computing. Basically, one could have a shared RAM dataset span hundreds or thousands of desktop workstations. All machines would have access to all of the data over the network. The OS would translate virtual address references to physical - including finding the machine that has that chunk in it's RAM. The OS for the machine that has that chunk of RAM would coordinate changes to that RAM. This could allow a corporation to utilize their desktop investment for many supercomputing projects - reducing their supercomputing needs, saving them money.
It also allows one to more easily build a modern supercomputer with the chipset.
The Commodore 128 used a funky bank switch system to allow access to 128 KB RAM. This is kinda/sorta how Intel wants to extend the x86. IMO, this cripples the arch. Intel seems to be doing it because they want the Itanium to take off. However, the Itanium appears to be an expensive dog. I'd rather have any of it's 64 bit competitors, at the moment. It appears to me that the Itanium will cripple Intel, without regard to the backing of big corporate customers.