DavidFMayerPhD
Registered Member
Intel Cites Breakthrough in Transistor Design
http://dailynews.yahoo.com/h/nm/20011126/tc/tech_intel_dc_2.html
http://dailynews.yahoo.com/h/nm/20011126/tc/tech_intel_dc_2.html
Intel Cites Breakthrough in Transistor Design
- Thread starter DavidFMayerPhD
- Start date
- Status
This is cool, but....
Sure, you can have faster & faster computers but at some point it doesn't make any difference. Computers have outstripped office applications long ago (word processing, spreadsheets, etc). The only things that still challenge computer speed are high-polygon-count 3D graphics and np-complete chopping algorithms. Computers get faster & faster, but what besides the conventional applications is filling the need?
What breakthroughs have been made in programming? What about AI, or self-aware, or environment-aware computers? Is processing speed really the main roadblock in these areas? What obstacles exist for programming such things?
Sure, you can have faster & faster computers but at some point it doesn't make any difference. Computers have outstripped office applications long ago (word processing, spreadsheets, etc). The only things that still challenge computer speed are high-polygon-count 3D graphics and np-complete chopping algorithms. Computers get faster & faster, but what besides the conventional applications is filling the need?
What breakthroughs have been made in programming? What about AI, or self-aware, or environment-aware computers? Is processing speed really the main roadblock in these areas? What obstacles exist for programming such things?
There are still many problems that need to be addressed with High speed CPU's. One is the capability of parallel processing threads through the CPU rather than just a thread.
Most of the time some of these fast CPU's mess themselves up with a HHGTTG "42". They try compiling an answer without fully following the question which creates this fractal variable that becomes an error. This is most noticable witin the older windows environments when you start getting BSOD (Blue screens of Death) and Kernel32.Dll errors.
In truth the main concern with CPU's is similar to that of Semi-Conductors, namely the more energy that is needed to compute causes Heat which ca cause material degradation.
This is where you get back to the notion of allowing your CPU's to be of a lower speed, but you have more than one CPU in your system. (Of course the older windows systems didn't have the capacity to use multiple CPU's)
Next you have the point that Graphics cards no longer need to send too much info through the CPU (They can have their own built-in ones).
What would concern me also is as the processors get faster the RAM (VRAM/DRAM/SDRAM etc) seems to get left behind, so you can process at a stupid speed but your memory still restrains your system. (Thats why AMD systems have a speedier more Reliable system than it's Intel Counterpart.)
I could of course mention of the first Pentium MMX chipsets, which had one major problem.... they couldn't calculate (they had a problem with floating point)
Most of the time some of these fast CPU's mess themselves up with a HHGTTG "42". They try compiling an answer without fully following the question which creates this fractal variable that becomes an error. This is most noticable witin the older windows environments when you start getting BSOD (Blue screens of Death) and Kernel32.Dll errors.
In truth the main concern with CPU's is similar to that of Semi-Conductors, namely the more energy that is needed to compute causes Heat which ca cause material degradation.
This is where you get back to the notion of allowing your CPU's to be of a lower speed, but you have more than one CPU in your system. (Of course the older windows systems didn't have the capacity to use multiple CPU's)
Next you have the point that Graphics cards no longer need to send too much info through the CPU (They can have their own built-in ones).
What would concern me also is as the processors get faster the RAM (VRAM/DRAM/SDRAM etc) seems to get left behind, so you can process at a stupid speed but your memory still restrains your system. (Thats why AMD systems have a speedier more Reliable system than it's Intel Counterpart.)
I could of course mention of the first Pentium MMX chipsets, which had one major problem.... they couldn't calculate (they had a problem with floating point)
Re: This is cool, but....
Success_Machine,
Don't forget real-time voice identification, translation, and conversation...They need quite a few MIPS to get it right and advances in that area of technology are ongoing.
Success_Machine,
Don't forget real-time voice identification, translation, and conversation...They need quite a few MIPS to get it right and advances in that area of technology are ongoing.
Computer is the single most dramatic event to happen
to mankind since the beginning of ages, no question there.
This thing helped us understanding and mastering our
environment at a totally insane rate since it's introduction in the
middle 50's
I'm of those who believe that we will se more science advance in
the next 20 years than we had in the last 100. Computer play a
major role on this.
Right now as we speak, we can simulate nuclear explosion
with a very high accuracy on computer like ASCII white from IBM
This monster take 2 basketball field worth of space.
With the latest advance in nano related science, like carbon
nanotube, it is clear that the new Intel discovery just add to the
mixt of very promising evolution of the silicon based CPU to
appear in the near future.
Considering that silicon based CPU will last an other 10-12 years
at most, nano technology is set to be the next big thing in
computer science.
It is estimated that the human brain can do 30 million MIPS. In
order to have a convincing Cyborg, we would need at least a 30
Millions mips CPU. It is now estimated that we will reach such a
computational capability on a single CPU somewhere around
2015 to 2020.
As far as programing this 30 Millions MIPS CPU is concerned, there
will be no problem. Just take a look at what AND corporation
( http://www.andcorporation.com/ ) is doing right now whit his
HNET programming language. It is just amazing if you ask me.
This is the future, and it's just the beginning.
Think about it: The more we know, the less to understand remain.
Also, the more we know, the faster we learn new thing. At some point in time in this learning curve, you get to a point where
the graph shift from ''mostly horizontal'' to ''Definitively vertical''
This is the exponential era. We are about to get there, thank's to
computer.
The list of application requiring more processing power is
countless. Real time voice translation, Robotic surgery, Data base
search, artificial intelligence, Automated intelligent space probe
and robot etc...
The need for power is there and will be fulfill soon. Thinking
machine crunching data at totally out of this world rate will
enable us to make light years jump in many science field like
genetic, physic, math, space exploration, artificial life and
countless other.
If you ask me, I would like to have an Humanoid like DATA in
Star Trek at my disposal. Imagine what a couple of DATA could
do with scientist to make science progress. Mind blowing, it will be.
Can't wait to 2020
Enjoy,
Steph
to mankind since the beginning of ages, no question there.
This thing helped us understanding and mastering our
environment at a totally insane rate since it's introduction in the
middle 50's
I'm of those who believe that we will se more science advance in
the next 20 years than we had in the last 100. Computer play a
major role on this.
Right now as we speak, we can simulate nuclear explosion
with a very high accuracy on computer like ASCII white from IBM
This monster take 2 basketball field worth of space.
With the latest advance in nano related science, like carbon
nanotube, it is clear that the new Intel discovery just add to the
mixt of very promising evolution of the silicon based CPU to
appear in the near future.
Considering that silicon based CPU will last an other 10-12 years
at most, nano technology is set to be the next big thing in
computer science.
It is estimated that the human brain can do 30 million MIPS. In
order to have a convincing Cyborg, we would need at least a 30
Millions mips CPU. It is now estimated that we will reach such a
computational capability on a single CPU somewhere around
2015 to 2020.
As far as programing this 30 Millions MIPS CPU is concerned, there
will be no problem. Just take a look at what AND corporation
( http://www.andcorporation.com/ ) is doing right now whit his
HNET programming language. It is just amazing if you ask me.
This is the future, and it's just the beginning.
Think about it: The more we know, the less to understand remain.
Also, the more we know, the faster we learn new thing. At some point in time in this learning curve, you get to a point where
the graph shift from ''mostly horizontal'' to ''Definitively vertical''
This is the exponential era. We are about to get there, thank's to
computer.
The list of application requiring more processing power is
countless. Real time voice translation, Robotic surgery, Data base
search, artificial intelligence, Automated intelligent space probe
and robot etc...
The need for power is there and will be fulfill soon. Thinking
machine crunching data at totally out of this world rate will
enable us to make light years jump in many science field like
genetic, physic, math, space exploration, artificial life and
countless other.
If you ask me, I would like to have an Humanoid like DATA in
Star Trek at my disposal. Imagine what a couple of DATA could
do with scientist to make science progress. Mind blowing, it will be.
Can't wait to 2020
Enjoy,
Steph
kmguru
Staff member
Instead of using a single chip with billions of transistors, we should look at distributed computing architecture where a large number of processors with only a few million transistors work together to solve problems. It works for super computers, why not PCs? A single die can have 1024 subprocessors and one can have 8 chips for the processor unit....or something like that....
Using the space
Hello everybody, I am new here and I wish to participate.
I apologize if my english is not good enough.
I agree with kmguru in the fact that the time spent solving a problem is not reduced only by increasing the CPU speed or its number of transistors. An important thing to consider is the CPU architecture, it can be a parallel, superscalar or superpipelined architecture for example.
An refering to the architecture, there are at least four proposals to the use of the extra space in the CPU, that result of the highest integration scale achieved with the new technology of transistors.
The first proposal is to increase the amount of on-chip cache. It has been widely proven that cache memory accelerates the execution of any program.
Another idea is, as kmguru said before, to replicate the same processor several times, to have a parallel computer on a chip.
The next idea is to add several specialized functional units to perform specific and sophisticated operations, for example fourier transforms, encrypting and decrypting units, image compressors and decompressors, etc. , all performed in hardware and exploting parallelism between instructions.
The last proposal is to add to the CPU a reconfigurable unit (also called FPGA), such that instead of using a limited number of different specialized units, the FPGA could be configured to implement any special function. Of course we will need a new kind of compilers, loaders and even operating systems.
We have to keep in mind that not all the ideas exposed before are useful for every kind of application and there must be evaluated to determine which of them is the best.
Hello everybody, I am new here and I wish to participate.
I apologize if my english is not good enough.
I agree with kmguru in the fact that the time spent solving a problem is not reduced only by increasing the CPU speed or its number of transistors. An important thing to consider is the CPU architecture, it can be a parallel, superscalar or superpipelined architecture for example.
An refering to the architecture, there are at least four proposals to the use of the extra space in the CPU, that result of the highest integration scale achieved with the new technology of transistors.
The first proposal is to increase the amount of on-chip cache. It has been widely proven that cache memory accelerates the execution of any program.
Another idea is, as kmguru said before, to replicate the same processor several times, to have a parallel computer on a chip.
The next idea is to add several specialized functional units to perform specific and sophisticated operations, for example fourier transforms, encrypting and decrypting units, image compressors and decompressors, etc. , all performed in hardware and exploting parallelism between instructions.
The last proposal is to add to the CPU a reconfigurable unit (also called FPGA), such that instead of using a limited number of different specialized units, the FPGA could be configured to implement any special function. Of course we will need a new kind of compilers, loaders and even operating systems.
We have to keep in mind that not all the ideas exposed before are useful for every kind of application and there must be evaluated to determine which of them is the best.
Last edited:
The Mhz is nothing indeed. The IPS (instruction per second)
is what count. No matter how we achieve a better IPS throughput
it will greatly benefit our capacity to simulate real life in real time.
Anyone has ever figured how much processing power it would
need to electronically simulate an entire human being with billions
of cells, in real time ? This is were it's getting interesting and we
are still 1 or 2 decades away to get there.
The human body is a 4.5 billion years old evolution from the single
cell that appered back then. Nature provide us with 1 CPU, the
brain. I don't see any advantage to spread the computing power
over a multi CPU setup. We do that today because this is the
only way to get more processing power.
The day we can fit ASCII white power to a single CPU, we may
not bother to build a 1024 CPU monster anymore...
Steph
is what count. No matter how we achieve a better IPS throughput
it will greatly benefit our capacity to simulate real life in real time.
Anyone has ever figured how much processing power it would
need to electronically simulate an entire human being with billions
of cells, in real time ? This is were it's getting interesting and we
are still 1 or 2 decades away to get there.
The human body is a 4.5 billion years old evolution from the single
cell that appered back then. Nature provide us with 1 CPU, the
brain. I don't see any advantage to spread the computing power
over a multi CPU setup. We do that today because this is the
only way to get more processing power.
The day we can fit ASCII white power to a single CPU, we may
not bother to build a 1024 CPU monster anymore...
Steph
Let us consider something. The transfer of information in the brain, even when it is based in electrical signals, is done in miliseconds, and this transfer in an electronic CPU is performed in nanoseconds.
How we can imagine that the brain is better?, The answer is that the brain is a highly paralell machine. The brain is always performing several tasks in parallel, I mean, multiple tasks are performed in different sections of the brain at the same time.
That is why I defend the paralellism at every level.
How we can imagine that the brain is better?, The answer is that the brain is a highly paralell machine. The brain is always performing several tasks in parallel, I mean, multiple tasks are performed in different sections of the brain at the same time.
That is why I defend the paralellism at every level.
Last edited:
Steph ...
I think you'll find that the following:
The 'brain' itself consists of three major structures (CEREBRUM, CEREBELLUM and BRAIN STEM) and two minor (medulla oblongata and thalamus). So in a sense you have a neatly integrated collection of specialized processors that are programmed not only to work together, but also independently. Thinking of 'the' brain, rather than the 'nervous system' is kind of like trying to envision a network where all you're seeing is a single PC.
Take care.
I think you'll find that the following:
isn't quite correct. What might be more appropriate would be think of the various ganglia, both parasympathetic and sympathetic, and the autonomic nervous system as sub-processors which, in a sense, parallel process sensory and other inputs prior to transmitting the result to the 'brain'.
The 'brain' itself consists of three major structures (CEREBRUM, CEREBELLUM and BRAIN STEM) and two minor (medulla oblongata and thalamus). So in a sense you have a neatly integrated collection of specialized processors that are programmed not only to work together, but also independently. Thinking of 'the' brain, rather than the 'nervous system' is kind of like trying to envision a network where all you're seeing is a single PC.
Take care.
Last edited:
Intel chip.
All the talk about that tera transistor, Is along way off.
It will be years yet till they have that.
There is alot more that goes into making the chip, they still get alot wrong.
There just know getting everything ready for .16 micron.
All the talk about that tera transistor, Is along way off.
It will be years yet till they have that.
There is alot more that goes into making the chip, they still get alot wrong.
There just know getting everything ready for .16 micron.
Hi Chagur ! Hello Yiech !
I like both of your view. I never consider the problem of
computing under this angle. Of course, I already knew that parallelism is the key to a flexible and efficient computing
system, just like the brain.
My point at this stage was more like let's get the most power
on smallest scale possible for the the CPU, and then incorporate parallelism into the spec instead of the other way around.
Considering the nervous system has a sub-processor might not
be exact. I think that the nervous system is a peripheral of the brain. The nervous system doesn't ''compute'' on it's own, only the brain in the human body can do this, with the assistance and
input from the nervous system. The cognitive task is the premise
of the brain.
regards,
Steph
I like both of your view. I never consider the problem of
computing under this angle. Of course, I already knew that parallelism is the key to a flexible and efficient computing
system, just like the brain.
My point at this stage was more like let's get the most power
on smallest scale possible for the the CPU, and then incorporate parallelism into the spec instead of the other way around.
Considering the nervous system has a sub-processor might not
be exact. I think that the nervous system is a peripheral of the brain. The nervous system doesn't ''compute'' on it's own, only the brain in the human body can do this, with the assistance and
input from the nervous system. The cognitive task is the premise
of the brain.
regards,
Steph
One of the problems with CPU's is the necessity for Cooling. I've mentioned this once before also.
In original PC's of about 100Mhz the CPU doesn't need to have a fan, because it manages to stay cool enough for use, once you start running at higher speeds, it produces more heat resulting in having a fan as a necessity.
The way to defeat the heating problem and boost a systems performance is pretty much in the Parallel processing which you have been mentioning, but rather than using a whole bunch of 500mhz-1Ghz CPU's you best re-utilising the old CPU's that give off less heat.
This is what I have mentioned in other threads about "Why we only use 10% of the brain", what people neglected to understand was that 10% isn't a 1-in-10 BLOCK, but that 10% is spread through the neural networking processes so as not to OVERHEAT the brain.
When looking at using Parallel Processing systems with multiple CPU's this is what should be considered to keep the temperature low. (SPACE your CPU's out.)
In original PC's of about 100Mhz the CPU doesn't need to have a fan, because it manages to stay cool enough for use, once you start running at higher speeds, it produces more heat resulting in having a fan as a necessity.
The way to defeat the heating problem and boost a systems performance is pretty much in the Parallel processing which you have been mentioning, but rather than using a whole bunch of 500mhz-1Ghz CPU's you best re-utilising the old CPU's that give off less heat.
This is what I have mentioned in other threads about "Why we only use 10% of the brain", what people neglected to understand was that 10% isn't a 1-in-10 BLOCK, but that 10% is spread through the neural networking processes so as not to OVERHEAT the brain.
When looking at using Parallel Processing systems with multiple CPU's this is what should be considered to keep the temperature low. (SPACE your CPU's out.)
And your experience of computers during 2001 equates to...?
I'm not intending to judge your nearly decade old opinion. I just thought it would be interesting to see if your opinion has changed. It's rare you get to hear from someone who has forum tenure of a decade.
In 2001 I was very young. My experience was on a Pentium II running Red Hat Linux with some proprietary multics terminal emulation which connected to an aeronautical processing center down-town.
In 2001 I was very young. My experience was on a Pentium II running Red Hat Linux with some proprietary multics terminal emulation which connected to an aeronautical processing center down-town.
Wha??? Jesus, if I told you I ate 6 chili dogs, would you say that your brother ate 10?
- Status