Slashdot Mirror
Where's My 10 Ghz PC?
An anonymous reader writes "Based on decades of growth in CPU speeds, Santa was supposed to drop off my 10 Ghz PC a few weeks back, but all I got was this lousy 2 Ghz dual processor box -- like it's still 2001...oh please! Dr. Dobbs says the free ride is over, and we now have to come up with some concurrency, but all I have is dollars... What gives?"
We've found the limits of silicon and hard drives and they are being approached asyptotically. Relax...
More
Multi-processing is the way to go. We need to do that to help heat dissipation...
My wife doesn't listen to me either...
...you shouldn't do drugs.
Since when did Slashdot start mimicking Hunter S. Thompson?
People in Soviet Russia, however, appear to be afflicted with amusing juxtapositions of the aforementioned situation.
A programmer is a machine for converting coffee into code.
system buss, hard drives, etc too slow anyway
It was just an observed trend. The trend is breaking, as far as retail availability, and thus we are not seeing our 10GHz rigs. (I believe that Moore's law is still trending fine in the labs.)
I remember the old days, when programmers nudged every
single bit of speed and capability out of the machines they had.
When computer engineers, faced with limits, still made magic
happen.
I hope this ushers that habit back into the profession. We have a lot of great technology, right now, let's find a better way to use it and make it more ubiquitous.
...I cannae change the laws'a'physics!
According to most predictions we were meant to be enjoying lives of leisure by this point - working a 5-hour week in the paperless office, and driving to work in our hovercars.
Sigs cause cancer.
It'll still crash, just faster. how about having M$ catchup on the 2005 technology brfore jumping ahead?
"I cannot think of any need in childhood as strong as the need for a father's protection." -- Sigmund Freud
There will always be points where technology slows down because it invariably will have to go through some total redevelopment instead of just building upon current products (like what they will be doing with the space program.)
News Reporters Make Tasty Polar Bear Treats!
we now have to come up with some concurrency, but all I have is dollars... What gives
CONcurrency is short for Confederate Currency. Your ancestors used the old notes to light cigars after the Civil War.
Then again, you might need to use laundered money. That's also called concurrency.
If the main processor can no longer be relied on to become continually faster to support perpetual feature bloat and inefficiency, how long until someone offers a FBPU?
that clock speeds are the one true and accurate method of determining computing performance, so if the clock speed isn't faster, it must not be better.
Gee, thanks Intel Marketing Department.
...running nearly twenty times slower sounds a little like hooking a 1,200 horsepower supercharged nitromethane-burning Hemi to a set of bicycle tires. With either one, if you can't "hook it up", what's the point?
About the word "if": If bullfrogs had wings, they wouldn't bounce around on their little green butts.
To my mind it _might_ be a good thing if the rampant speed-advance slowed (a lot).
Consider:
We might get some return to efficient coding being the norm, instead of writing systems anyhow and throwing more/faster hardware at it until it runs acceptably (Microsoft; its you I'm looking at!)
Your (and your business') desktop machine might _not_ become obsolete in no more than 2 years, and mmight continue in useful service as something more sensible than a whole PC doing the job of a router...
Processor designers might spend more time (i know they already spend some) on innovating new ideas, rather than solving the problems with just ramping up clock speeds.
Cooling/Quietening technology might have a snowball's chance in hell of catching up with heat output?
(and the wild dreaming one)
Games writers might remember about gameplay, rather than better coloured lighting...
since the mid 90s thats all I have built - they really do extend the time before you feel compelled to upgrade. Sure there are not that many apps that run threads on each CPU. But to me a large part of it is that I run many applications simultaneously. With 2 CPU's I rarely get any sluggish feel. And if one app is being especially hoggish I can set it to run on one cpu and flip another important app to the other cpu.
This time around I also sprung for a hardware raid card and set up a 10 array. That has helped quite a bit with system responsiveness.
I've also turned off as much eye candy as possible. After a couple days its really not missed and things are much snappier.
yeah it would be great if I could run out and get some 10GHz chips to fry a few eggs on, but I think my dual MP2200's still have a bit of life in them.
Did anyone else notice they sourced their Figure 1 from Wikipedia? I know the trolls are already going to agree with the article just to get picked on, but then they'd have to agree with wikipedia, which they can't do because they already are picking on Wikipedia.
It's a paradox of trolling. What will we do!
For context, click Parent.
I'm not sure that graph in the article has any real significance - Nearly all of the 'flattening out' of the curve is in the future, and the remaining part doesn't look *that* discouraging.
Apart from that.. I'm not sure if it's a major shift - quite a good amount of apps already are multi-threaded. Not to mention the increasing complexity (and thus amount of work done) of operating systems and api layers, which are already running as seperate processes.
and I am not being negative and glib just passing off the need for faster, PC. When you have speed things will come up to use that horsepower!
But now say we have 2.5gihz athlon's and G5's and 3.8gihz pentiums and other processors all at respectable speed. They are really very good for many things!!
What do you all people see as uses for faster PC's if you could have 10gihz on a desktop machine or even laptop. Well video would be one thing that is good. What else would it help to have the speed for? Would it make a real advantage for you to step up for a 10gihz computer? Is it more, important, to have your 500gib and 1tib hard drive space and 8gib of ram with a big monitor??
Mac Security News
How about a double intel xeon 3 ghz ? should be better.. hehe
Music is the sedative for mind...
flying car.
Where else would it be?
#SickNotWeak
but all I have is dollars...
They're probably even Canadian dollars...
Why the size restraints on processors? Could a processor be made twice as fast if it could be made twice the size? When we hit the limit on how small transistors can be made, could processors continue to increase in speed by making them larger? I see no need why computers need to keep a processor size to two inches square.
There are so many areas in which we can improve technology; it's not time to get fixated on GIGARHURTZ when we're coming up against a wall.
Against stupidity the Gods themselves contend in vain.
Moore's law has nothing to do with processor frequency. It says that semi-conductor capacity doubles every 18 monthsm, not frequency. (With the corollary that there is no appreciable change in price). As we all know, semi-conductor capacity is roughly proportional to speed, so saying processor speeds double every 18 months is not quite wrong, just a little inaccurate. On the other hand, saying that we're not seeing 10 ghz processors, so Moore's law is broken is wrong.
To make laws that man cannot, and will not obey, serves to bring all law into contempt.
--E.C. Stanton
Ramping up clock speeds is hitting some serious limitations as far as increasing the work done by a machine is concerned. There are lots of ways to get work done faster. They are just harder to market without some good, popular, and independent benchmarking standards. At some point engine manufacturers realized that increasing the cubic centimeters of displacement in an engine was not the best way to make it faster or more powerful. Now most car reviews include horsepower. Clock speed is analogous to CCs.
Normal user software will get no improvement of cpu speed like this.. They day to day computer user would seek not benefit of something this fast.
Linux is like a teepee. It has no windows, no gates, and there's an Apache inside.
If, as the Dr. Dobbs article says, "the free lunch is over", then the only sensible thing to do is make do with what we have now. For goshssakes, people, the computers we have now are already insanely over-powered. How many more gigahertz do we need my life already?
--
What short sigs we have -
One hundred and twenty chars!
Too short for haiku.
The difference between Intel and AMD's cpu architecture yields similar performance but at very different clock speeds(AMD's 3200+ runs at 2.2GHz). Other aspects of PC performance continue to improve, so as long as the trend is towards greater overall system performance, clock speed matters less. And greater parallelism is a good way to achieve this.
After reading the previous article about the laser on a chip. I think that we will continue to see sped increases.
Everytime we come upon the "maximum" capabilities, some smart guys and gals figure out a way to get beyond it.
I've been hearing for 10 years that computers can't possibly get much faster. Memory can't possibly get any smaller.
I say "bah!" to the nay sayers. People will always find a way. A guaranteed way to be labeled an idiot in history is to claim something is impossible.
Wasn't there some british lord in the 30's who publically claimed that it was impossible to fly?
Wasn't the maximum possible speed to travel 100mph?
then the speed of sound was impossible to overcome.
then, you can only get to space with billions and billions of dollars.
or, a computer to do "that" would be the size on the empire state building and require all the power of niagra falls just to keep it cool.
The impossible is for weenies.
----- If communism is a system where the government owns business, what do you call a system where business owns govern
Moores Law is an exponential law; it just cant go on doubling exponentially, there has to be a ceiling at which the technology being used reaches its peak.
Think of it like accelerating your car. 2mph, 4mph, 8mph, 16mph, 32mph... you are increasing your speed exponentially, but even that 5.7 V10 is going to max out at some point. IANAM, but I'm sure there is a name for this.
Also from TFA, this hasn't happened yet, chip designers are just being more intelligent about how they boost speed (not just cycles, but cache, multi-threading etc).
However, I'm sure that within the next few years silicon will have given us all it can... long live grid computing
Go pick it up in your flying car.
Lodragan Draoidh
The more you explain it, the more I don't understand it. - Mark Twain
It's really interesting that dual prossesors haven't caught on in mainstream computing. This isn't to say that there are mobos out there that are avalable to the public that support two or more prossesors, but it's really a niche market, people who generally understand computers looking for a super-fast new computer at a certain point have to start looking at dual prossesor solutions. Not to mention that those consumers(let's call them geeks), when they have their dual prossesor system, there are very fiew apps that actually take advantage of the dual prossesor capabilities(photoshop, final cut, etc). The rest of them just end up being run on seperate prossesors, which is still faster, if ALL apps where made for both single and dual prossesors, then Joe Gamer and the Geek and eventually Fred Email will want a dual prossesor computer, because the two prossesors can act as one(as I said, if the apps are coded right), therefore making faster systems.
For >95% of users, I see no need to have computers faster than 2Ghz. Maybe I'm getting old... oh, and music these kids listen to....
When they get off the silicon and hop onto those nice diamond wafers (there is an article in wired), then we will see faster processing.
The main problem - our largest producer (Intel) said they would not stop utilizing silicon until they made more money from it...We know that the industry likes to stagger upgrades. Instead of giving us the latest and greatest - they give us everything in between in nice "slow" steps so we spend more money. Personally, I wouldn't mind seeing the jumps of 1ghz at a time. This year 2.0 ghz, next year 3.0, following year 4.0, etc...and then eventually increase it further so its 5ghz at a time, etc. et al.
I mod down so you can mod up. Your welcome.
When there's no free ride, programmers will have to compete with each other on who can squeeze that last bit of performance out of existing hardware. So you can kinda sorta predict the revival of the performance-conscious programming.
Judging from these pictures of the Intel retail boxed heatsink for the Pentium 4 560J (3.6 GHz), by the time we get 10 GHz PCs, the hovercar problem will take care of itself.
Raw speed isn't measured in Megahertz anymore. Actually, it never really depended on MHZ, it was always MFLOPS. For years, and finally getting due recognition, AMD has destroyed Intel despite having a slower mhz core. MFLOPS was and is the key.
Brooklyn.
"Based on decades of growth in CPU speeds, Santa was supposed to drop off my 10 Ghz PC a few weeks back, but all I got was this lousy 2 Ghz dual processor box"
Santa was unable to deliver your 10Ghz system this year for the following reasons:
1) Santa's Flying Car has not arrived
2) Santa could not use his sleigh because it failed the new FCC saftey requirements for subobital ships (something about flaming reindeer poo falling from the sky).
3) The OS for the new 10Ghz computer is Duke Nukem Forever which isn't currently available - maybe next year or decade.
My code won't perform adequately until processors have 1GB L2 cache.
And for that matter, where's my Mr. Fusion, Hovercar conversion, Jaws 17 and perfected weather service? Aren't those supposed to be done by 2015?
All semiconductor manufacture processes have defects, and when a defect occurs it ruins the chip.
Typically defects occur at random so there will be X defects per cm^2 (in this case X should be 1).
So the bigger you make the individual chip dies, the more likely the case that there will be a defect in one of them. Hence larger chips have far lower yield than smaller chips.
This is the main reason that very few digital cameras have sensors the size of 35mm film... since you'll probably end up with a yeild of less than 0.5.
Just click here.. and send me your CC number, name and billing address ill get it shipped right out to you.
Free shipping if you act in 24hours..
But wait.. theres more..
---- Booth was a patriot ----
Why should a car have more than a 50hp motor ?
Why put more than a 25hp motor in a motorcycle ?
Why should anybody want to get a plasma flatscreen TV?
Hell, what good is color TV?
----- If communism is a system where the government owns business, what do you call a system where business owns govern
Man, I'm just gonna redirect tinyurl.com to localhost, fuck this troll shit.
Long live Moore's Law
(Since I don't know any better, I'm going to keep betting on Moore's Law and the inexorable genius of the hardware engineers.)
Does that mean that I have to learn how to code properly?
The fallacy here is that the clock speed has to keep doubling. Moore's law says that the number of transistors on a chip doubles each 18 month period, and we're still pretty close to that.
;^)
Intel has just caved on the speed doubling in particular, by knocking the clock speed off their product designations, mainly because the Pentium M chips were running significantly faster than the same-speed P4's. AMD's Athlons have been 'fudging' their numbers by having the product number match not their clock speed, but that of the roughly equivalent P4 chip.
Meanwhile, cache sizes are up, instruction pipes are up, hyperthreading has been here a while, multi-core chips are coming down the pike... we're still getting speed gains, just not in raw clocks.
At the same time, the Amiga philosphy of offloading to other processors is truth, with more transistors on the high-end graphics processors than there are on the CPUs!
I hate to say it, but what do you think you need 10GHz for anyway? Unless you've got a REALLY fat pipe, there's a limit on how much pr0n you can process
The high-end machines do make good foot-warmers in cold climes.
Design for Use, not Construction!
Double channels have to have twice the transistors, right? If that's true then Moore's law lives on. Think of it as quad-processor machine in one cpu case....that's 4 times as many transistors as 1 CPU. Or am I missing something here?
Broken Hearts are for Assholes. - Frank Zappa
Slashdot can't imitate Hunter S. Thompson. He can actually write!
I fully agree that we *could* see significant speed increases. However, the benefit for AMD and Intel in boosting speeds just isn't there - Intel delayed their 4Ghz offering not because they couldn't bin some chips into a 4ghz bucket, but because the return for hitting higher clock speeds just isn't there - there's not enough of a market, they need to get the return out of existing investments before moving on, and there are few applications that demand that kind of horsepower.
Whenever the government "picks winners" rather than letting nature pick winners, the technologists and therefore technology loses.
(Now that Cray is dead, according to the supercomputing FAQ, "The CCC intellectual property was purchased for a mere $250 thousand by Dasu, LLC - a corporation set up and (AFAIK) wholly owned by Mr. Hub Finkelstein, a Texas oilman. He's owned this stuff for five years and hasn't done anything with it.")
Secondly, as I've discussed before both operating system and database programming are awaiting the development of relations, most likely via the predicate calculus, as a foundation for software. Both are essentially parallel processing foundations for software.
This feeds into quantum computing quite nicely as well, as relations are not just inherently parallel, but are parallel in such a way that they precisely model quantum software.
Seastead this.
Moore made a prediction about how the number of transistors in computer chips change over time. By no standard is his prediction a law. It is called a law because it has held true for the most part over time and it makes a prediction about the progress of technology. Furthermore, Moore's Law really has nothing to with the speed of processors. So arguing about whether we should or shouldn't have 10 GHz processors according to Moore's Law is pointless.
Do I hafta say it?
You know it....
Like how AMD has been able to make instructions take fewer clock cycles to run.
If an operation can execute in 3 instead of 4 clock cycles the gain in speed is obvious.
How both Intel and AMD have built branch prediction.
We'll probaly start to see chips come out with common compiler generated routines built in (hardware peephole optimization anyone?).
We may see some speed increases occur as the transistor count goes down from engineers finding more efficient ways to layout a chip.
Where there is a will (profit), there is a way.
----- If communism is a system where the government owns business, what do you call a system where business owns govern
There is no corollary (deduction or inference from axioms) about the change in price. There is an assumption of it.
Well, I have been here for a while already, but I'm not politically correct
Lesbian Nazi Hookers Abducted by UFOs and Forced Into Weight Loss Programs - -all next week on Town Talk.
...as we've been saying for, oh, at least the last 20 years, which is about the time I was writing up my Ph.D. thesis on concurrent languages and hardware.
As far as I can see (being slightly out of the language/computer design area these days), concurrent machines and languages aren't taking off for the same reasons they didn't take off in the 1980's:
There's more than a handful of generalisations there, but in short: Moore's Law means that nobody is going to buy a highly concurrent computer when consumer PC's are still getting faster, and the people who really need high parallelism (modellers and the like) have their own special-purpose toys to work with.
Actually, Moore's law states that transistor channel length will halve every 18 months.
But thanks for playing!
".. with most of their traditional approaches ..."
So, we find non-traditional approaches. So, yes your V10 gives out at 200mph. But, then some guy comes up with an SR71 Blackbird. Then some other guy builds a rocket. Then.....
It may be aero-dynamically impossible for a bumble-bee to fly but, its not physically impossible.
----- If communism is a system where the government owns business, what do you call a system where business owns govern
I like 'em slow. That way I can go get a cup of coffee, talk to my co-workers, etc., and just tell the boss "It's still booting" or "I'm running a large spreadsheet".
Remember the old days when we didn't expect a computer to be as fast as lightning?
Slow down and listen to the sound of the files being saved to floppy.
HexaByte - he's a square and a half!
I find that few people - even tech people - really appreciate just how fast 3GHz is.
To put it into perspective, consider that at that clock frequency, light travels about 4 inches in one clock cycle.
Mind boggling.
garethw
Start researching how to build better computers and start a company. You have some options to explore -
/ 9/1/2/1
l l save yourself a lot of time by not modeling nonlinear processes with linear equations and infinite corrections under 10^-13 m (ultraviolet divergences).
1) DNA/Molecular computers
2) Atomic switches
http://www.physicsweb.org/articles/news
3) Betacomputation (Switches made from neutrons and protons that can be on/off by adding/removing electrons bound inside of the hadronic structure)
This makes for good power supply too http://www.betavoltaic.com/
4) Positron/electron photon exchange
(Yes Virginia, antimatter/matter changes the phase of absorbed photons)
5) Integrated silicon/optic chips
6) Black holes (See Sci Am Dec 2004)
Also for all of you aspiring scientists out there do yourself a favor and join the present by reading about nonlinear/nonunitary mechanics
http://www.i-b-r.org/ir00018.htm
You'
Posthuman since 2001.
There is no surprise here people. IMHO there hasn't been any new technology in the past 50 years, for the consumer. Sure everything is computerized, runs fast, does a crap load of things...But the basic make up is still the same. Cars are still the same, Planes are still the same, houses are still the same...washing machines, blenders, The basic make up of just about everything is still the same as it was for your parents or grandparents. There nothing new here, hasn't been for awhile... Nano looks promising, cloning looks good...But it's so far off, I won't see anything come from them in my life time.
I might not be up on all things new, but I havn't seen anything Differnet in a very long time.
What we need here is a new composites, new minerals....Something different.
Captain Scarlet............out
It's left blank because I have nothing to say to you punks!
You have dollars? Where can I get me somma those? Guess it's just another thing that I predicted I'd have 5 times more than what actually happened.
And to think, that apple's CPUs are nearly at the same 'number speed' in the mhz race now!
Who would'a ever thought to see that happen?
Tibbon
tibbon.com
Moore's Second Law:
Every 18 months there will be twice as many stories saying Moore's law can't continue.
Moore's Third Law:
In 18 months, with hindsight given to whatever the next advance is that they missed, they will look twice as stupid.
authoring a DVD in less than an 4 hours from the dv-avi source?
my own CGI production in my lifetime?
every day http://en.wikipedia.org/wiki/Special:Random
Sounds like they suckered you and gave you one of them Apple boxes Instead of a new Intel P4 system... i got dual 3.6ghz procs for christmas which i overclocked to 4ghz each giving me 8ghz of pure power! So i dont know about you turkeys, but im close to 10ghz over here...
No sig today.
Now that the only way up is through more CPU's, I think it's time that the inherently multithreaded BeOS makes a big comeback.
Yay BeOS!
I mean please let us give -10 "junk" "yesterday's news" "no sane human being interested" "FUD" "poster is an analphabet [i.e. can't read the news, just post]" whatever.
First, most of said "bugs" were fixed the day before last blue moon. Second, there are "bugs" mentioned which are the fault of 1). windows 2). user (which means they can't freaking configure the darn Thunderbird to store the mails where they know it's most safe). This last one is as saying an applciation is faulty and buggy because it has a default store directory option which it uses if nothing else is specified. Come on. Whatever. To the malformed URL "bug": you can do that on every and each browser, huw is that a sw bug. Get lost.
Get a life, post news. Junk we have a'plenty already.
I am putting myself to the fullest possible use, which is all I can think that any conscious entity can ever hope to do.
The problem was Intel's P4 design that didn't take heat into consideration. It hit 3GHz and aside from bus speed bumps, cache bumps, and an IPC lowering redesign (Prescott) that has so far allowed it to attain 3.8GHz. When did it hit 3GHz? Was it 2 years ago? That's pretty abysmal in my opinion.
Of course, AMD has been slowed down as well, having not done many speed bumps in the past year, aside from a 2.6GHz boutique release. However AMD will probably get 2.8 and 3.0GHz processors out of the door this year.
Of course, for Apple, they've gone from 1.2GHz G4s to 2.5GHz G5s. The slow-down was earlier for Apple, and they've got over it, and hit another one. Still, those processors are from different companies.
Dual core is a logical progression, when you can do it and there is a real benefit to doing it (OS dual support, application multi-threading, etc). It can keep the power consumption lower (you don't need to push the processor to the maximum), and it spreads the heat sources out, instead of having one extremely hot core, you have two merely hot cores to cool.
The article wibbles on about GHz, but what matters is performance. Whilst benchmarks aren't ideal, I'm sure a graph of Spec Rate performance on single processor (where a processor can have multiple cores, of course) and multiprocessor has been improving steadily over that time. 2 years ago Athlon XPs were pushing 900 in SpecFP and stuff, yet high end Athlon 64s now push 1700+. Systems are getting faster! It is just the value of GHz as a useful system performance indicator is no more, and hasn't been for around 5 years to be honest. People are waking up to that.
I appologize to you.
:)
A bit difficult to understand was your grammar.
----- If communism is a system where the government owns business, what do you call a system where business owns govern
There is one law in computer programming that is even more certain than Moore's Law: Over time, the user is going to do less work for the computer and the computer is going to do more work for the user.
Remember back when users had to wait in line in front of a terminal to run their punchcards through the mainframe? Back then, human time was cheap and computer time expensive. Nowadays the user's time is paramount.
Multithreaded programming breaks this law: It is hard to do multithreaded programming- Humans just don't think that way very well. To do it in a way that an arbitrary program (i.e. not a ray tracer) can see consistent performance gains in a multi-CPU environment is almost PhD-level hard. Making single-threaded software is already a major undertaking and anyone thinking that, in general, they should start designing all their programs as fundamentally concurrent programs is going to fall behind their competition due to other factors (security, features, etc.).
Instead, the only way concurrent programming is going to play a major role for the majority of software, I believe, is at the compiler and OS levels: The OS and compiler designers are going to have to do their utmost to transform single-threaded software to perform optimally in a multi-CPU environment- These folks are going to have to take up the slack that the slow CPUs clockspeeds are causing in terms of limiting the speed of Software- Concurrent programming at the application-level is only going to play a minor role in this, in my opinion.
...can really spoil your conclusions. Examine Figure 1 in the article. Log-log graph, with a two-kinked trendline. The article uses this figure to justify the statement that
"Around the beginning of 2003, you'll note a disturbing sharp turn in the previous trend toward ever-faster CPU clock speeds."
However, these trendlines are meaningless; they are plainly wrong. For a start, all the data points are to one side of the trendline. That's a bit glaring, even if you're not a statistician. More interestingly, if you just consider the points, rather than the trendlines, you'll find a single trendline, linear in log-log space, that fits the data, up to the present day. It has a similar gradient to the supposed 1980-1994 trendline. During the 1995-2005 period the points were clustered slightly above this line, but the effect is that of a big technological push - like a region of compressed time on the graph, just what you'd expect with the amount of money pushed into the industry about this time. Otherwise, pseudo-Moore's Law is on track. Now, in the development labs, there's another story. Exponential growth is already hitting up against a wall, and we must turn to QIP technologies to sort that out.
That's not true at all. At a mere 2GHz, light can only travel 15cm (6in) through free space in one cycle -- hardly a long distance. Add in modulation and switching delays, and you really can't ignore the board-level latency even with optical interconnect. On the other hand, even on-chip communication takes multiple clock cycles these days, so maybe it wouldn't be that much worse..?
The processor is only as good as the software it runs. In other words, even if you have a fast processor, it's not going to really help unless the programs you run, are actually efficient.
Maybe a bad analogy, but kind of like driving a car that can go 200 miles per hour, on a roadway that is 80 miles per hour.
Why would anyone throw out the language/runtime with built-in multithreading support when system processors are becoming increasingly parallel?
Economics will favor hardware acceleration
I believe the software industry is still in its infancy and humanity's physical pursuits (hardware) will always outpace our intellectual pursuits (software). It remains simply too expensive to fix performance by assuming the software industry will deliver ubiquitous quality. Too many amateurs and those learning on the job in this industry (myself included) and that is not going to change anytime soon.
The complaining about limits in hardware and such generally comes from Pentium IV owners (myself included). Not all future architectures will be as dismal. Even for Intel. These multi-channel future baselines will be good. Reminds me of the Amiga days. When developers could assume multiple channels of DMA without penalty and concurrent access to memory while custom chips were displaying graphics and playing sounds, they were freed to do amazing things with the software.
Computability at the Planck Scale
http://www.arxiv.org/pdf/gr-qc/0412076
Posthuman since 2001.
Faster processors MEAN NOTHING!
Your computer is as fast as the slowest hardware component that is used. Additionally, if the hardware component doesn't provide enough capacity, then it causes slowness.
Traditionally, the order of slowness is (slowest first):
1. Network
2. Disk
3. Main Memory
4. CPU
(I'm not counting peripherals)
CPU is last. Concentrate on more memory and really fast disk. This gives you more for your money.
Now, use (or write) really good software. That would be non-Microsoft. Put Linux on on your box and double the speed right away.
Read the fine graph. Look at the data points and ignore the author's blue lines for a minute. There is nothing in the data that suggest a hard limit has been reached. OK... the rate of change slowed down a little in 2002. That's all.
Today's CPUs sport some more powerful instructions, and they perform optimizations that range from the pedestrian to the exotic, including pipelining, branch prediction, executing multiple instructions in the same clock cycle(s), and even reordering the instruction stream for out-of-order execution.
Calling any of those "optimizations" is a stretch, in my opinion, unless a V8 engine (or maybe two 4-cylinder engines with a shared gas tank and transmission) is an "optimization" of an inline 4.
Note that some of what I just called "optimizations" are actually far more than optimizations, in that they can change the meaning of programs and cause visible effects that can break reasonable programmer expectations.
Not on any "normal" CPU (see explanation below).
But in recent years they have been willing to pursue aggressive optimizations just to wring yet more speed out of each cycle, even knowing full well that these aggressive rearrangements could endanger the semantics of your code.
They won't affect the semantics of code unless the CPU has a design problem (again, see below).
Two noteworthy examples in this respect are write reordering and read reordering: Allowing a processor to reorder write operations has consequences that are so surprising, and break so many programmer expectations, that the feature generally has to be turned off because it's too difficult for programmers to reason correctly about the meaning of their programs in the presence of arbitrary write reordering. Reordering read operations can also yield surprising visible effects, but that is more commonly left enabled anyway because it isn't quite as hard on programmers, and the demands for performance cause designers of operating systems and operating environments to compromise and choose models that place a greater burden on programmers because that is viewed as a lesser evil than giving up the optimization opportunities.
Actually, you reorder them inside the processor, and then do a check to make sure your reordering did not affect the functionality before actually committing to memory (some instruction sets allow memory access reordering, and have explicit instructions to specify that certain operations must happen before others). As it is right now, your top-of-the-line x86 chip is going appear to execute a program exactly the way it would if it only did one operation at a time, and everything in order. A HUGE amount of work goes in to implementing "precise interrupts" - making sure that at any point, if you interrupt the CPU, all instructions before the current PC (program counter) have executed to completion, and no subsequent instructions have started execution).
Basically, the way instruction reordering works is that instructions are fetched from memory in order, and their dependencies are evaluated. They're given tags (unique IDs inside the CPU), and their tags are added, in order, to a queue at the end of the pipeline. Then, they flow through the actual execution units out of order. As they execute, they update a "speculative state" - results are sometimes stored in an entirely separate register file called the "future file". An instruction can only affect "architectural state" (programmer-visible state) when it is the oldest instruction in the queue at the end of the pipeline. If an instruction has an exception (divide by zero, software breakpoint, interrupt), all subsequent instructions are squashed - you can do this because the queue is added to in order, and the machine looks just like an in-order machine would.
It's worth noting that various researchers have looked into the optimal pipeline depth - basically, due to various sources of overhead (flip flops cannot store values instantaneously, there is clock skew on a chip, and other higher-level factors like data dependencies), they conclude that 6-8 gate delays
My server
One thing to notice in ddj-diagram is that the x-axis is logarithmic. Jump from 1 MHz to 1 Ghz in speed equals to jump from 1GHz to 2 Ghz. I think we are doing just fine.
One thing to consider is that electrical signals in chips have maximum speed of 2/3 of light speed.
Homework: how many centimeters (or inches) electrical signal travels in one clock cycle when processor is running 1Ghz and 10Ghz.
Yes! it is kind of hard to design chips when bits in other side of the bus are different form opposite side.
Dyslexics have more fnu.
According to Microsoft, an average Longhorn system will need to have a 4-6GHz CPU. But if when Longhorn arrives, 4GHz CPUs are high-end parts and 6GHz CPUs don't exist, well...I don't predict good things for Microsoft. Longhorn in 2007, anyone? Or maybe 2008...
For science, yes perhaps, but the everyday user is not doing that many floating point calculations. Even if they are, they won't need them done at the speed that a scientific application would.
Flops are more important for scientific applications, which is why you usually see all the large modern supercomputers measured using this unit of measurement; universities and research groups are the targeted audience.
For the normal user, floating point calculations aren't as important, so hertz is a decent standard for rating a CPU's speed. MIPS (millions of instructions per second) is also a halfdecent measurement system for the general person, but it can only be used to compare CPUs of the same architechture.
Beware he who would deny you access to information, for in his heart he dreams himself your master. -Anonymous
http://curbstone.com/_pigs.htm
"Everything that can be invented has been invented."
Charles H. Duell, Commissioner, U.S. Office of Patents, 1899
"I think there's a world market for about five computers."
Thomas J. Watson, Chairman of the Board, IBM
"The bomb will never go off. I speak as an expert in explosives."
Admiral William Leahy, US Atomic Bomb Project
"This 'telephone' has too many shortcomings to be seriously considered as a means of communication. The device is inherently of no value to us."
Western Union internal memo, 1876
"Heavier-than-air flying machines are impossible."
Lord Kelvin, president, Royal Society, 1895
"Professor Goddard does not know the relation between action and reaction and the need to have something better than a vacuum against which to react. He seems to lack the basic knowledge ladled out daily in high schools."
1921 New York Times editorial about Robert Goddard's revolutionary rocket work
"Man shall never reach the moon, for such a quantity of gunpowder would be needed as to gravely injure the crew."
Children's Encyclopaedia, 1926
"Man will never reach the moon regardless of all future scientific advances."
Dr. Lee De Forest, inventor of the vacuum tube, father of television and owner of over 300 patents
----- If communism is a system where the government owns business, what do you call a system where business owns govern
This is not 10 GHZ, but due to lack of hard drive (who needs hard drive slowing things down if you have 256 GB of RAM?) and 4 GHZ CPU, this is probably going to be ten times as fast as today's desktops.
Don't let the horrible site design alarm you, this is a real thing, and was displayed at this year's CES. Right now, it's gonna cost about $14K, but will drop in the near future.
Well maybe we don't have 10G, but we do have 64bit!!!
But that concurrency might not be in the main CPU. It will be in the graphics processor. Graphics processors don't have the von Neumann bottleneck - with enough hardware, you could have one processor per pixel without changing the application level programming model. Shader languages, from Renderman on, are explicitly one program per pixel.
The Playstation 2 is the world's largest selling non von Neumann machine. The PS2 is two very capable vector processors, an underpowered MIPS CPU, and a frame buffer. Its vector processors are wierd; they're nothing like a standard CPU. Yet, with difficulty, programmers are doing work in them. Not just graphics, either - physical simulation and planning.
The game development community is already dealing aggressively with concurrency. Pretty well, too.
... and will always be! ;-) I think I first read this qoute sometimes in late 80s/early 90s, and it is still true. You know why? Ever looked at power dissipation specks of even the simplest GaAs chips? You would not want to build a processor out of those, Cray tried with Cray 4 and failed... ;-(
;-)
superconductors is the way to go for highest speeds/most concentrated processing power, due to extremely small power dissipation and extremely high clock frequencies (60 GHz for logic is relatively easy right now), but the problem is that after someone invests $3B in a modern semiconductor fab they do NOT want to build a $30M top-of the line superconductor fab to compete with it. IBM would be a good candidate for this, but they got burned on superconductor computer project back in 80s and would not touch it with 10 foot pole now, though both logic and fab has changed dramatically since then.
Disclosure: on my day job I do design III-V chips, and I used to design superconductor chips up until recently, now trying to push that technology forward is more of a night job for me...
Paul B.
When I will be able to visualize this fractal http://http//www.superliminal.com/fractals/bbrot/b brot.htm in real time, I will be satisfied with the processors speed.
I hate to say it, but what do you think you need 10GHz for anyway? Unless you've got a REALLY fat pipe, there's a limit on how much pr0n you can process ;^)
Photorealistic (or at least much better than the current high-end) rendering in real-time, I have some database apps that do a whole lot of number crunching, I have plenty of large projects that take 20 minutes to compile on a 3.06 P4 - CPU speed is the bottleneck on all of these.
A 10 Ghz CPU would probably bring with it 2GHz+ BUS and RAM.
You can never have too fast a CPU or GPU, too much RAM or too much HDD space.
A multicore CPU is great, but no substitute for raw speed. It's like comparing a bullet train to a fleet of honda escorts. The cars can move the same group of 1000 people, but the train does it so much faster and more efficiently.
I don't need no instructions to know how to rock!!!!
'nuf said.
Which processor outperforms which:
1a)486-25SX
1b)486-25DX
2a)PIII - 450
2b)G4 - 450
3a)G3 - 300
3b)Playstation 2 - 300
Moral of the story : there are far, far more important performance measurements than clock frequency. If you think otherwise, you might as well slap a VTEC sticker on your case.
P.S. As other's have pointed out, Moors law has nothing to do cpu frequency.
SO you want to make it more... etheral? Dude, it's called ENGLISH. Quit using bloglish.
the proper term is probably "signal," but anyway, depending on your dielectric and how you set up your lines, your speed varied but almost all of them is a fraction of c. IIRC coax is ~25cm / 1nS, FR4 (PCB material) is ~18cm / 1nS, and internal wiring in chips are actually slower, which means clock distribution is a very big deal within that 1.x sq.cm of silicon.
My life in the land of the rising sun.
I am tempted to agree but there is still a high demand for a "yet faster" machine. Its just not with infinite price. So, the second part of the engineering task is to keep the production costs inline while attempting to meet the demand.
It generally appears that if you can get the price of a high end chip below ~$400 its going to sell.
The companies are pretty good about getting the prices down once production is up. Compare a 1ghz chip today with one 2 years ago. You can even pick up a knock-off one now (microtel? I forget the name).
----- If communism is a system where the government owns business, what do you call a system where business owns govern
While the idea of having a superfast processor is a pretty exciting thing in itself, the question that arises is whether we really need it? For one, most of the time the processor is idle or uses only a little %age of its power for the very simple reason that we need only a little power of the processor for tasks like web surfing, etc. Games requires a bit more though. I can only see the utility of a super processor in high-end animations or computer modelling, etc and not every John and Jenny do that, I presume.
This was spewed from Intel in 2002:
"First, by switching to the Pentium 4 architecture, Intel can drastically boost the clock speed. The old server Xeon topped out at 1.4GHz. The new one debuts at 1.8GHz, 2GHz and 2.2GHz, and will eventually pass 10GHz, she said."
http://news.com.com/2100-1001-843879.html
I can't find the exact quote and article, but another Intel exec/rep stated that this goal would be achieved by 2006.
Well, it's 2005, the P4 has topped out at 3.6ghz and has been discontinued because Intel has determined that the P4 arcitecture is streached to the limit.
Bottom line is that we should be expecting a 10ghz processor soon because Intel brazenly stated that they would produce one. Whenever they do make these statements the AP drools over the story, stock prices jump and I'm sure investors get excited.
Instead, their next gen processor is a 2ghz Pentuim M dothan. Intel should be ashamed of themselves for lying to the public and should be investigated for inflating their stock value though fictional claims about their processor technology.
I started with a 4.77mhz PC XT computer (Leading Edge) and years and years of PC Magazine subscriptions. They were always predicting the end of Moore's law. And they were always wrong.
The algorithms chosen *are* important, and in some cases you shouldn't simply reach into the API toolbox and use the third-party solutions. There is no substitute for knowing how to write your own sort routines, specialized linked lists, and binary trees.
However, this leads to people going too far in that direction as well, which is just as problematic. People re-invent the wheel for no better reason than that their own inflated egos make them believe they can do everything better. Or because the available routines differ in some completely insignificant respect from what they'd prefer.
The result: Lots of programs duplicating effort, creating bloat. You have Gnome, KDE, Mozilla and OpenOffice all with their own set of widgets and APIs. It's fragmentation and duplication of effort for no good reason, and so the casual Linux (in this case) user ends up having to store all four of them in memory.
I think that the Not Invented Here syndrome is a far more important factor in creating 'bloat' than people chosing inefficient algorithms.
> but there are times when you have to have low
> latency and there's no substitute for smallness
> then; light just isn't that fast!
True, but your thinking is just SO 2D! What we need is to make the L2 cache into a sphere and paint the CPU on the surface. Then you put the clock in the center, guaranteeing synchronization at every point. When are chip manufacturers going to leave Flatland?
Resistance and Capacitance becomes a problem at the higher switching rates and the longer distances. Capacitance over long pairs of wires creates cross talk between the two (one reason why power lines are croseed every now and then is to create some induction to offset this). Another is that the longer distances increase the resistance on the wires. This causes the recieved voltage to be lower and can sometimes be to low to trigger a voltage high reception. This can be offset by making the wires wider but is still a problem. A third is the combination of Resistance and Capacitance call an RC constant. This problem is basically that when you go from high to low or low to high signal propogation takes longer. Causing the recieving end to wait more until it can make sure that the incoming signal is either high or low.
Take a basic AC electronics course and you should learn about most of these. Either that or a Digital Design (of Integrated Circuits) course.
Fly me to the moon Let me sing among those stars Let me see what spring is like On jupiter and mars
Santa is creating a beowulf cluster of the 10Ghz machines up north. As it turns out he is a computer Geek too.
The OS would do all the dirty work of breaking up your application into pieces that can run concurrently for you.
and yet the Itanium is still stuggling. The only sin greater than underestimating the technology of the future is to overestimate it.
Your CPU is not doing anything else, at least do something.
My main system will make many of you laugh, but here it is:
Beige Mac G3 tower, upgraded to 733MHz G3(ZIF), overclocked system bus to 100MHz, 768MB Ram (PC100), Original 24x Apple CD-Rom, HP 12x6x32 CD-RW Drive. WDC 20GB EIDE, Maxtor 10GB EIDE, Belkin 4 port PCI USB card, Adaptec PCI Ultra SCSI card, 6x IBM 10000rpm Ultra-2 Mil-Spec 1GB drives (RAID), 3DFX Voodooo 5/5500 PCI Graphics Card, various scanners, printers, monitors (2), Wacom Pads.
So there it is, the manufacture date on my CPU is 1997. So what do I do with this system? Graphic design, large format printing (Epson Stylus 10000), Digital art (Painter/Photoshop/Illustrator), Play the SIMS (all available expansion packs), Play Quake 3, etc etc... Probably most of what the average person does with a machine and more.
Sure my Quake framerates never break 60, and I have to actually WAIT for a gausian blur on a 200MB image, and most annoyingly there is no support for my VooDoo 5 under OS-X, so I still run 9.2.1 most of the time. But the point here is that with well written software there are VERY few people who would actualy bennefit from a 10GHz processor. My computer does all I need it to, and I demand a fair ammount from it. Right now my only real reason to upgrade is either for more HD space or to be able to run OS-X with full Quartz acceleration.
Please, I Laugh every time I hear some fellow MacHead whine about being stuck at 3.2GHz, I mean really, GET OVER IT!
P.S. My linux machine is a hacked EyeOpener running Debian/Knoppix. It sits on my end table to surf the web and check e-mail. It has 128MB ram, a 4GB HD, and does everything it needs to.
A Call For A New Slashdot Moderation Level!
Folks (like me ;-) who spend their careers developing server-side software are going to (like me ;-) read this article and go, "Look, bozo, not every problem has concurrency!" or, "We're already doing that the best we can!".
Mainframe developers (I'm not one) have wrestled with load balancing issues for, like, eons. Sometimes parallel processors help. Sometimes, the problems are some other system (e.g.: database I/O). I don't think it was fair to act like even most software developers have been depending on Moore's Law. Anyone running on the edge of their performance bandwidth eventually looks for ways to work on parts of their problem concurrently.
Mark
Dude, that is what Intel was doing until AMD came along and forced them to get into this "keeping up with the Joneses" routine.
I can't decide whether to put a smiley face on this or not. I was being sarcastic, but for all we know it might be partially true!
My beliefs do not require that you agree with them.
You do realize who the author of the article is? Herb Sutter, he works on the C++ compiler team at Microsoft and IS THE CHAIR OF THE ISO C++ STANDARDS COMMITTEE.
Keep in mind chips these days have all the transistors "facing" up so they can be mated to a substrate that removes heat. Creating a sandwich of components is probably not a good idea unless you completely change the manufacturing process, using 3d "pathways" to conduct heat and signals in and out of the core. Quite a bit more difficult to model and produce.
THIS THING CAN TURN ON A DIME, MACROSSZERO STYLE ALSO FUCK BETA, ~NYORON
Where is it? I see lots of software that states it works with a hyperthreading capable operating system, in the Win XP world for instance, but I haven't seen any examples of software that clearly states that it is hyperthreading capable.
Is this a case of the emperors new clothes or is my limited exposure to XP software making me look foolish?
I did a google search and haven't found any software that clearly states it is hyperthreading capable other than XP itself. Anyone have a list?
Andy
The Amiga philosphy of offloading to other processors is truth
:(
:)
Interestingly, this predated the Amiga. Jay Minor's (father of the Amiga) project prior to the Amiga was the Atari 400/800, which used the same type of hardware design. Certainly the 800 had other design limitations which killed it in the end, but that aspect of it was way ahead of its time. I sure wish Jay was still around today; it'd be fascinating to see what he'd come up with by now.
I miss my Guru Meditation Errors!
Now go fuck yourself faggot.
The win32 stuff you see in the list might get dropped even though it's already written since it's not clear I will have long term resources to support it. Something about the 100's of dollars required to replace all your win32 software when moving to the current win32 or win64 platform.
Remember, growth of computing power is exponential and when comparing trends it often makes sense to switch to a logarithmic scale. log(10GHz/4GHz) isn't too much of a deviation from the trend.
Doesn't it make you feel good to know that our freedoms are protected by politicans, lawyers and journalists.
An extreme example would be the human brain - quite "slow" - but gets a lot done each "clock cycle"
If you crave the challenge of making tight, efficient code, sometimes with very little under you but the bare chip itself, then embedded systems might be the place for you.
cue the grumpy old man voice: "Why back in my day, we didn't have 64-bit multi-core chips with gigabytes of memory to waste, no sir, we had to write in assembly code for 8-bit processors, and WE LOVED IT!"
-paul
Pistol caliber is like religion: everyone has their favourite, and theirs is the only right choice.
I don't think this article said what you think it said.
- or -
Truly, your intellect is dizzying.
Here you go, here's your 100ghz Pentium, cooled to 40 degrees Kelvin. Yes, we know the case looks like a chest freezer and sounds like a horny sperm whale, but clock speed's what it's all about. Yes, we realize that your electricity bill will quadruple and that they'll have to build another diesel-burning power plant to keep all these fine machines, but on the plus side, you've got a place to keep your beer!
The world's burning. Moped Jesus spotted on I50. Details at 11.
My laptop is a Transmeta 700 with 300MB RAM. XP on it is impossibly slow. 98 is acceptable. Linux is fast and responsive, with a lot of services on, kde 3.3. My wife has a desktop Athlon 450 with the same amount of RAM. it's slower than my laptop, even with 98 and nothing else running.
It's better to be the foot on the boot than the face on the pavement. ~~ tkx Kadin2048
Herb Sutter, the author of the article, is a voice worth respecting. He's chair of the C++ standards committee, and although he now works for Microsoft, he isn't actually evil. He's Canadian, for a start. He is also actively steering the Microsoft C++ compiler towards standards compliance, and he even managed to get a room full of C & C++ programmers to give a spontaneous round of applause to the guy that implemented partial specialisation in MSVC++. That was at the ACCU conference in 2003.
Dont click that link.
Freedom or George Bush
Hub Finkelstein is (was, actually - he died last year) the landlord for my place of business.
"Stupid! Stupid stupid stupid stupid! I touched the hot wire right there - I'm an idiot!"
if we havee optical for sound on some audiophiles stereo rig, why not in a 3000 dollar computer??
No, seriously.
Light - and information - can travel no faster than 3x10^11 cm/s.
Your typical processor chip is what, a couple of inches across? Let's round that off to 5 cm for ease of calculation. And do a thought experiment...
Even if calculations are performed instantaneously, there's still the little matter of the result - information - travelling to the next portion of the chip for the next calculation. If you're using the entire chip to do calculations, that means you have to wait for the data to go from one side of the chip to the other. In other words, the information needs to travel a distance of 5 cm (again, assuming calculations take no time at all).
Thus, the upper limit on processor speed in Hz is the time it takes for information to physically cross the processor... in this case, 3 x 10^11 cm/s divided by 5 cm or 6 x 10^10 Hz... that's 6 GHz for those of you keeping count at home.
Obviously, calculations take slightly more time than 0, so that slows the processor down even more... which means that a 4 GHz processor is probably pushing the upper physical limits of a 5 cm processor.
The challenge now is obvious... to get any significant increase in speed, we must find a way to make the processor physically smaller (thus decreasing the distance that information must travel). And that's no easy task! For instance, to increase speeds from 4 GHz to 10 GHz (as the submitter asked for) requires shrinking the processor from 5 cm in size to 2 cm in size.
Obviously, the faster speed you seek, the smaller you have to compact the processor... the hypothetical 100 GHz processor can be no more than 0.3 cm in size based on the speed of light alone (again, assuming an ideal processor that performs calculations instantly)!
*That's* why processor power is slowing down a bit compared to the old axe of "doubling every 18 months" - because we're running up the speed of light and we can't break it... and we can't shrink component sizes by half every 18 months, either!
--AC
?Andy giveth, and Bill taketh away.?
That's only half right, because you don't have to let Bill take away. KDE3 runs well on a 233MHz PII and 64MB of RAM, almost a whole order of magnitude less of hardware than it takes to make XP happy. The picture is more drastic when you consider the virus load most XP setups must endure. You need a 2GHz processor just to keep running while your computer serves out spam and kiddie porn.
The changes Dr. Dobbs so wants are already happening in free software. There's a reason things like Arts can play music, games and system noises all at the same time while software on M$ has trouble sharing sound devices. If Beowulf is not a 10 year head start on concurrency, I don't know what is.
Quoth SVDave:
I don't predict good things for Microsoft. Longhorn in 2007, anyone?
Perhaps old Billy should put his money into processor development instead of SCO and FUD.
Friends don't help friends install M$ junk.
This is really one of the best articles I've read in a long time. My compliments to the author.
My favourite bit:
I will speculate that today's single-threaded applications as actually used in the field could actually see a performance boost for most users by going to a dual-core chip, not because the extra core is actually doing anything useful, but because it is running the adware and spyware that infest many users' systems and are otherwise slowing down the single CPU that user has today. I leave it up to you to decide whether adding a CPU to run your spyware is the best solution to that problem.
Yes. This is exactly what I was thinking as I read that part of the article.
My first thought was that the author had confused the optimization settings of his compiler with the in chip reordering that the processor was doing.
# (/.);;
- : float -> float -> float =
As a side note to the above, it's also why "three-dimensional" computer chips are of such interest - because you can fit more processors into the three-dimensional space that light can cross in a given amount of time, thus giving you the potential to do more calculations - or alternatively, being able to fit the same amount of processors in a smaller radius than in two dimensions (and not having to simply "shrink down" a two-dimensional circuit board). But again, the size limit comes into play because information can only move so fast.
--AC (the same as above)
1) Make impact by thinking out of the box
2) Leverage paradigm shift
3) ????
4) Profit!!!
I'm not speaking of bleeding-edge applications, I speaking of how long it takes a menu to appear after you hit START, or how long it takes to open an editor for a simple text file.
About a dozen years ago, a PC hacker laughed at my Unix box because there was a noticeable lag between touching a key and getting a response. He can laugh no longer. Unix systems aren't any better, but PCs have become dogs.
But there is hope. If we put nearly as much intellectual energy into making software more efficient, perhaps we could actually harness the couple of orders-of-magnitude improvement in processor power that has occurred in the last dozen years.
The processor designers have spent their energy making a well defined set of operations (with minor enhancements) run faster. The software designers have instead spent their energies making the core operations slower in support of more and more (dis?)functionality.
In vacuum (and approx also air) this is 3e8m/s. To compare other material parameters an epsilon_r and mu_r are defined, relative to those of vacuum.
Most other materials have a mu_r=1 (except ferrites etc), but an eps_r which can be very different. For silicon this eps_r is about 12, therefore the speed of light/electrical signals in silicon is sqrt(12)=3.5 times lower than in air.
When the distance traveled becomes of the order of magnetude of the wavelength things can become problematic. For a 3GHz processor this is about 3.10^8/3.10^9/3.5=0.03m, or 3 cms. This effect will soon start to play. Another problem would be race conditions, i.e. different signals should moreless take equally long to reach their destination or you get into trouble.
All of this a bit intuitive, hope anyone will correct possible mistakes, I am no expert in semiconductor technology...
Z
How do you deposit another fresh layer of uncorrupted substrate on top of a processed layer?
?
Chemical vapor deposition? It's not as easy as it sounds.
What about thermal expansion/contraction, thermal effects on timing, IR drop of a sandwich layer of substrate-oxide-metal-oxide-substrate-oxide-metal
How do you analyze process defects on the lower layers?
And a host of other properties.
If you want to do 3D, just make alot of chips and stack them together. That's the easiest way right now.
The problem with that guy is he is allways talking about applications.
We all know a dual core CPU is not really twice as fast than a single core one. We all also know that an application is not necessarily twice as fast if it runs on a dual core or dual ship system.
But thats not the point.
Most of us run a system where ALLWAYS more than one thread is ready to be executed. More than one thread from different applications of course.
The overall system can easily be nearly twice as fast (199%) with a dual core chip. As soon as the chip internal communication restrictions hit it from being truly twice as fast.
E.G. my mail programming waking up and accessing my mail box while I listen to a song on iTunes and type this posting can easy run on the second core without slowing down the other core besides concurring about the memory bandwith.
angel'o'sphere
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
I can't believe a respectable publication like Dr. Dobbs is citing Wikipedia as a source in this article...
If processor technology does indeed plateau, perhaps there would be hope for what the parent suggests. However, if we were to treat the English language and the brain as an example of code and fairly static processor power, we could use the post as an example of a simple application (more like a shell script than emacs, but an example nonetheless).
Consider the following paraphrase:
That's a character savings of approximately 25%. Obviously, there is good reason for the lack of optimization in the original parent post. It would not be worth the time and effort to optimise such a trivial bit of English code. If it were a manuscript for a techical manual (more like emacs than like a shell script), more optimizations would like be done.
Programming parallels language usage. There will always be some who optimise to the Nth degree. In programming, these people sometimes make demos and kernels and clean, nice libraries; in language, they tend to be poets and lawyers and grammar nazis. There will be others who write large quantities of clean, but not hyper-optimized code. These are the application programmers and general writers. Then there will be the great, unwashed masses. Outside of programming, they either don't write more than an occasional postcard, or they write terrible fanfic. In programming, the postcards are quick-and-dirty shell scripts and utilities, and the terrible fanfic includes most of Microsoft (ever notice how they're always the main character in their software?).
Anyway, the analogy is strained and breaks down easily, but it fun to just sit back and take a thought at it once in a while. It may be that we will get more efficient coders out of a slower processing environment, but then again, it could just mean that we end up with a lot more fanfic-quality applications that make the well-written code stand out as an enjoyable and welcome group.
Can anyone honestly say that todays top of the line computer is not enough (in practical terms)? Computer development usually lags behind software, but we have reached the point in the last three years where processors outpace the demands of software. The top of the line consumer level computer today (say, an Alienware ALX?) is roughly equivalent to a few hundred Pentium 133's from just a decade ago. Most apps today claim to run on hardware as slow as a PII-233, and most games will run on a 1 Ghz Athlon with 512 mb of RAM.
Concurrency is just another buzzword. What we are going to find now that we've hit a glass ceiling on processor clockss is that we need to look at how our needs are changing. You don't need a super fast processor to run a PVR, play MP3's, run your cars engine, or to manage your to-do list. Any decent computer nerd can automate his entire house with an old windows machine. In terms of using computers to improve the standard of living, keep us entertained, and to advance civilization, we're nowhere near the limits just because we topped out on absolute processor clock speed. There are still factors which determine a processors power, and more importantly the processor itself is only one small part of the overall computing devices which we utilize.
In the future, as our needs grow and change, so will the manner in which we implement technology. I'm willing to bet that 20 years from now nobody will care how fast their computer's clock can cycle.
http://sipipe.com
Transfer 20-25Gb/s from chip to chip easily, and 1-Gb/s per channel (wire/copper).
Modesty is one of life's greatest attributes
Your comment makes me thing you've done embedded software development. I'm currently turning up some new hardware, and I'm disgusted by the software IDE. The "hello world" demo program compiled, with full optimization, to just over 65k of code. Yelchkk. I ended up writing a complete console monitor program, complete with initialization, in assembly in just under 4k bytes.
...
The current mentality of "abstraction from the hardware layer" allows software folks to write code without knowledge of the underlying hardware, but the penalties are code bloat and an ever increasing appetite for cpu cycles. So while the cpu speeds keep increasing, the code performance stays about the same as more layers of abstraction are spackled on top. I have complete faith that someone will build an abstraction layer that allows a GHz SMP machine to run legacy single-thread code with the perfomance of a Pentium 2 366. And then the complaining about "we need better hardware" will continue
I've told myself, that once they hit Dual 4.0ghz ones, i'll sell/trash/retire my 1.25DP MDD and go get a fulled loaded 4.0DP.
And maybe a 30" screen to boot!
Tibbon
tibbon.com
Krusty: So he's proactive, huh?
Lady: Oh, God, yes. We're talking about a totally outrageous paradigm.
Meyer: Excuse me, but "proactive" and "paradigm"? Aren't these just
buzzwords that dumb people use to sound important?
[backpedaling] Not that I'm accusing you of anything like that.
[pause] I'm fired, aren't I?
Myers: Oh, yes.
"No fair, you changed the outcome by measuring it!" - Professor Hubert J. Farnsworth
My wife has an antique Compaq 486 DX2/66 with 8MB RAM (I bought it to her when she entered her office as a DA, ten years ago). It runs 95 + Works 95 (It originally ran Win 3.11 but I upgraded it). No browser, no dial-up network. It's barely acceptable. It is in her office now, but I intend to get it back and make it run some LTSP client.
It's better to be the foot on the boot than the face on the pavement. ~~ tkx Kadin2048
They're working on it, and progress will be made as a result, but the difficulty finding such is the reason CPU speed increases have slowed. I expect that there may be one more breakthrough in my lifetime allowing another factor of 10 Moore's law style increase in single CPU speed with an 18-30 month doubling period. One. After that, we need to go *MP-- and probably before that, too.
And, kook that I am ("Energy Profit Ratio!" "Hubbert Peak!" "Club of Rome Report!"), I suspect the energy demands of the increased speed CPUs may render them uneconomical outside of the most limited high-end demanding server applications.
//Information does not want to be free; it wants to breed.
The problem is electricity. If you have an electrical junction, current must flow through it as if at a stop light--one pathway goes while all other wait. This is the cause of 99.999% of waiting for stuff in a computer to happen. Light, on the other hand, does not have to wait for other light. Light beams simply pass right on through each other and keep on keepin' on. Furthermore, two light beams can interfere on a sensor in all sorts of complicated ways to convey all sorts of complicated information--one light pulse can convey one bit of information, but two can convey 4 or more, and so on.
Optical computing, when it finally gets here, will provide so much computing power it will not be consumable, even by Bill.
but have you considered the following argument: shut up.
Also, the idle time may be technically correct, however most modern processors use pipelining to overcome this. Pipelining is sort of like an assembly line for instructions in that instructions are queued in the processor. Say it takes 3 clock cycles to get the data for the instruction. A simple way to make the processor more efficient is to check what data an instruction needs when it enters the instruction queue, and hope that the data is loaded when the instruction gets executed a few cycles later. While that instruction is waiting, three or more instructions that are ahead in the same queue are executed. A pipeline stall occurs when something goes wrong with this process; either the data takes longer to load then was expected, or the queue empties. There was an artile on the Ars Technica web site a while back that explained pipeling in detail and better then I can.
Analogue lines aren't like DS-0 lines, which have a seperate control channel, the control is "bit robbed" from the signal. They take out 8kbps for signaling, giving 56k effective for encoding. That's why with ISDN there is talk of B and D channels. For BRI ISDN you get 2 64k (DS-0) B (bearer) channels that actually carry the signal. There is then a 16k D (data) channel that carries the information on how to route the B channels.
That's also why IDSL is 144k. The total bandwidth of an ISDN line is 144k, but 16k is used for circut switching data. DSL is point-to-point, so that's unnecessary and the D channel's bandwidth can be used for signal.
So 56k is as good as it will ever get for single analogue modems. I suppose, in theory, this could be changed in the future, I suppose, but I find that rather unlikely given that any new technology is likely to be digital end to end.
Another example of why processor speed isn't everything: my home system is a 1.1 gigahertz monster, designed as a server. Not the fastest system around, but close enough. But it doesn't really play AM's Alice very well, and if I want to run GTA3 at all, I'd need to spend a few bucks on a better video card. I'd sort of like to play Sims 2, but the necessary video card would cost more than I spent for the whole system. I'll never want to play it that badly.
As long as I can remember, people have obsessed about processor speed, not realizing how little effect this has on most apps. (The big bottleneck most people face is RAM; retail economics result in most systems being shipped without enough.) Nowadays, nobody needs more than a couple hundred megahertz, unless they're running something really graphics intensive. And even there, the specialized hardware between your monitor and the rest of the system does most of the work -- I doubt that Sims 2 would run any better on your hypothetical 10 gighertz system.
Fortunately, if you convert your current processor speed to binary you'll find you have that 10 ghz machine you are looking for
Seeing as the cell is going to (supposedly) be capable of 2 Teraflops as a 64-bit core chip, I'd say we're not too far off.
By the way, I don't know what I'm talking about.
How fast can you flip bits without blinking?
The confluence of physics and information theory flows form the central maxim of quantum mechanics:at bottom, nature is discrete. A physical system can be described usinfg a FINITE number of bits. Each particle in the system acts like the logic gate of a computer.Its spin "axis" can point in one of two directions, therby encoding a bit, and can flip over, thereby performing a simple computational operation.
The system is also discrete in time. It takes a minimumn amount of time to flip a bit. The exact amount is given by the theorem named after two pioneers of the physics of information processing, Norman Margolus of MIT and Lev Levitin of BU. This theorem is related ot the Heisenber uncertainty principle, which describes the inherent trad-offs in measuring physical quantities, such as position and momentum or time and energy.
The theorem says that the time it takes to flip a bit, t, depends on the amount of energy you apply, E.
The more energy you apply,the shorter the time can be. Mathematically, the rule is
t>= h/4E
where h is the Planck's constant, the main parameter of quantum theory.
For example, one type of experimental quantum computer stores bits on protons and uses magnetic fields to flip them. The operations take place in the minimum time allowed by the Margolus-Levitin theorem.
From this theorem, a huge variety of conclusions can be drawn, from the limits on the geometry of spacetime to the computational capacity of the universe as a whole.
Quantum mechanics predicts that spactime is discrete. Distance and intervals cannot be measured to INFINITE precision; on small scales, spacetime is bubbly and foamy. The maximum ammount of information (matter) that can be put into a region of space depends on how small the bits are, and the CANNOT be smaller than foamy cells.
Physicists have long assumed that the size of this cells is the Planck lenght (lp) of 10E-35 meter, which is the distance at wich both quantum fluctuations and gravitational effects are important. If so, the foamy nature of spacetime will always be too minisucle to observe.
Seth Lloy and Y.Jack NG
excerpt taken from Scientific Amercian Nov 2004
Black Hole Computers.
- these are not the droids you are looking for -
Cute signature, but incorrect. Apache Indians are a member of the "Southwestern Tribes" (A term applied by white scholars, not the tribes themselves), and the Southwestern tribes did not have enough skins to create TeePees (TeePees have a lot of wasted space, as far as construction material are concerned). The Apache lived in homes called WikiUps. They were dome shaped (like an igloo) and usually covered with bark, sicks, mud, and sometimes hides.
It just drives me nuts when people (not nec. parent poster) have this Indian/TeePee connection. The Native American cultures are just like the "Native European" cultures. Yes the are all "white", but you have german, english, russian, etc. Not all "red" indian cultures are the same.
Sorry, I'll step off my soap box now :-)
I'm still running a P3@500Mhz (with Suse Linux) and it works great. Sure it's not my everyday production machine, but for most things 'normal' users do, it would do just fine.
I once did a calculation of how quickly my PC could add up floating point 1.1 a billion times, and using Moore's Law realized that we'd hit Plank Time for the communication of information in about 150 years.
I accidentally ran over your 10GHz PC with my flying car.
You see? You see? Your stupid minds! Stupid! Stupid!
All you pale faces look the same to me so you all must be the same.
20 years ago my 1MHz C64 could do some pretty neat things and was adequate for most tasks. Now machines are more than 3000X faster, have more than 1000x memory, and something like 1000000x storage. Have we really gained much?
I cut a picture out of one of the industry magazines a year or two ago. It showed the prototype of the machines they are building that will allow building the next generation chips...and yes...it mentioned speeds over 10ghz.
If the machines to build the chips aren't even built yet, we'll have to wait a while for the next step.
I've done a lot of real-time embedded system programming with multiple tasks, interrupt service routines, etc. I've yet to use a compiler/language that made any attempt at making concurrent programming easier for the programmer. If I need something, I have to do it myself. It's sort of like doing OOP in C.
Mea navis aericumbens anguillis abundat
After all, Moore's Law predicts exponential growth, and clearly exponential growth can't continue forever before we reach hard physical limits; light isn't getting any faster. The growth must eventually slow down and even end
If we measure the speed of light in terms of feet per second as opposed to the more commonly known kilometers per second, the multiplier will yield a number which is greater (ignoring units). This will be good enough for the people in marketing and the salesmen at the major electronics outlets.
Similarly the onus will no longer be on the programmers or engineers to produce faster, more efficient, or more capable products. Businessmen, investors, and executives being what they are they will simply pressure the marketing department to produce more glitz and glam. Was the onus ever really on them? That's arguable. The size of programs has increased to the point where even a simple word processor wouldn't fit on a hard drive of 10 years ago, yet are the word processors of today helping us to write better letters? For all the extra horsepower and storage space we still do (mainly): e-mail, surf web, print pictures, play games. The only people who have a significantly improved computing experience are those who no longer have to wait 18 hours for the graphical rendering, mathematical modeling, or compilation to finish.
Intel has already shown us the way of advertising. Those three blue guys who are always doing some really neat stuff like taking off with rocket jetpacks or surfing around on anti-gravity boards. What does that have to do with processor capability, quality, or reliability? Nothing. At least automakers genuinely show a car in use doing something that a car does. When's the last time your processor caused you to physically fly to work?
fast as fast can be. you'll never catch me.
If desktop machines move towards concurrency, it could really give Apple a significant performance advantage.
Apple's performance problems truly began when Motorola proved incapable of getting the G4's clockspeeds to scale. They compensated by making the Powermac a largely dual processor line. This move was, of course, laughed at because of the absolutely terrible software support. At the time, Photoshop was really the only app that could use that processor, not even OS 9 really worked with it. This was the classic desktop dual-processor chicken-and-egg problem. No one wrote software, because no one had the hardware. Apple hinged the product's success on people being willing to buy the machine knowing that the second processor wouldn't do much until OS X finally shipped.
So now we find ourselves 5-6 years later.
The chicken-and-egg problem has largely solved itself. One part magically appeared, and given enough time, the other followed. An incredible amount of desktop software on OS X is multi-threaded (from OSS to shareware to commerical offerings), and demonstrates significant performance improvements when that other core is available. Most desktop software on other platforms (X11 and Windows apps), are still largely sequential, and it's going to take years for that to change. Apple's consumers are going to see a much larger, and more immediate, benefit as it becomes economical to grant the Powermac a couple more cores, and move dual core down to the consumer lines. The software is already there.
Both major CPU makers have stopped using the clock spped as the selling point of their products for a very good reason. Raw clock speed is the lest effective way to judge preformance. A system is only as fast as its slowest part. With Athlon FX systems getting close to 2GHz FSB speeds, on chip memory controllers, and beefy L2 caches, it's easy to see just how much preformance you can get out a of a chip if you give it what it wants: low latency, high bandwith, and memory to spare.
How about CPUs that can modify their own hardware-functionality on the fly, and very fast?
I know that programmable-hardware devices are already well-established, but what I am considering is something that can do this on the fly, and repeatedly.
Question is, how much performance improvement would this give?
"Slashdot - News and Chat Sites Deviant". (Click "homepage" link above for details).
The real lessons of computer era are not Moores' law of transistor density but Gate's law of cyber-guzzling by dense users which I now coin: In the long run, marketing pull always gets ahead of engineering push...just look at DEC. When I joined DEC in 77, those old programming proverbs were taped like gospel to the cubicles where the compiler writers worked. They seem forgotten but I found them here and here. In that year, nearly any system or periperal DEC dreamed up found willing scientific and engineering customers in labs. By '84, the earliest PC's were all the buzz but DEC's offerings in that area were ignored by the market. The imagination and genius of engineers and scientists is generally sufficient to awe the public initially. But if the technology has gratifying uses and economic benefits, markets absorb magic and make it an ordinary commodity and finally a staple for which improvement is always wanted.
And by the way, since when is this topic news? I hope it isn't just because DDJ mentioned it. I don't feel like dredging up pointers to the bezillion pages on the matter but there has been academic and industry handwringing about the inevitable limitations of transistor size and speed for a decade. OK, one URL, thats all you get! read pg 62 for consice four year old description of the issue [and how carbon nanotubes are going to save us]. The predictions of the exact day when progress stops were always a bit vague and hedged with hopeful notes about gallinium and going to 3-d circuits...all that is really happening is that, having seen the wall a long ways off, chip makers aren't going to smack into it head on with an abrupt cesation of speed increases but veer off in new directions and so only slow down the increases.
SLASHDOT: news for people who can't concentrate on work or have no life at all and got tired of yelling back at the TV.
Lying? How so? This is a prediction. How many of us know exactly what will happen in the future? You didn't *really* think they knew the future did you? They just had a roadmap based on their expectations of the process technology. It didn't pan out, so they changed the roadmap. Everyone else did too. AMD has delayed the hell out of their technology scaling, higher frequency and dual core procs to the point where Intel will beat them to a dual-core release, but I don't see you complaining about that, do I? Transmeta is pulling out of the IA-compatible CPU business, but you aren't crying foul there either.
The quote from the marketing rep you included does not have a date on it, and I'm sure the roadmap you remember seeing had an appropriate disclaimer on it, if it were even published by Intel (Intel doesn't generally publish these officially).
Furthermore, you can rest assured that some day there will still be 10GHz chips sold.
Haven't you ever noticed this clause at the bottom of similar PR? It's a shame that people actually need to cover their ass with legalese like this, really.
This Business Outlook contains certain forward-looking statements that are subject to known and unknown risks and uncertainties that could cause actual results to differ materially from those expressed or implied by such statements. Such risks and uncertainties include, but are not limited to, the Risk Factors noted in the Business Outlook as well as in the Earnings Releases, Business Update press releases and Intel's filings with the Securities and Exchange Commission on, e.g., Form 10-K and Form 10-Q.
> Ball Semiconductor
:) /me does the happy dance.
Hah! I knew it!
That company also have discovered some even more interesting advantages of doing spherical circuits. It turns out that the balls can be made without a cleanroom, which is obviously an enormous expenditure. They also cite lower environmental cost, due to less silicon going to waste. And making an inductor coil on the surface for communication is way cool!
At 3 GHz, light travels 4 inches in a clock cycle. If you have a 2 inch processor, including the delay due to not using light in a vacuum, you simply can't have clock synchronization throughout the chip. That doesn't mean that processors can't go faster, but, when they do go faster, there won't be a single clock whose speed you can report.
What are the consequences on Longhorn release? IIRC Microsoft planned their capacity requirements according to the projected capacity of hardware at the time Longhorn will be released (2006), not what makes sense with current hardware. IIRC Longhorn specs are for something like 4-6 Ghz CPU. If the hardware is not there, will there be a wrench thrown into Microsoft plans?
Object Oriented Programming worked because we had speed to spare, but thats over. Of course concurrency will be more important now, but the code will also have to be optimized. Goodbye C++, java, welcome back C.
I misread the title of the article. I was wondering if the poster was having some huge nostalgic crisis because he lost some old 286/386 or was stolen.
printf($randomline(sigs.txt) \n "-- "$randomline(authors.txt));
-- myself
"Unless you've got a REALLY fat pipe, there's a limit on how much pr0n you can process"
Amateur!!!
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
Hmm. Makes me think you should be running stuff like this on a different platform entirely. A Sun V490 would probably rip through that in seconds.
"People who do stupid things with hazardous materials often die." -- Jim Davidson on alt.folklore.urban
Hogwash! Write first, optimize later . . . what are the chances that I can write a better sorting algorithm than one included in a standard library that was written by some who studied sorting algorithms [and thoroughly tested the code].
Or more conretely:
"We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil." - Donald Knuth
I would argue that its still important to know HOW to write these algorithms, so you know what the cost/benefits of different types of sorts are, but there is no virtue in reinventing a working wheel.
I can't believe this got posted as a story...seems like trolling to me.
As you can see if you want concurrency and high optimisation you want Ada.
cool. interesting to know.
question - does this mean you're indian/native-american ?
For God's sake, please stop the business-speak!
But then how are we supposed to leverage our synergies going forward to create a win-win situation? You are generating negative ROI in this incumbent conversation, and have become a cromulent addition to the team. You will be capsized^W rightsized immediately.
I want to drag this out as long as possible. Bring me my protractor.
Many good points by other posters, but the core reason is simple. We are too scared to do it.
Great ideas have been around for a fair while how to speed things up by orders of magnitude. Examples? Anybody remember John Backus and F.P.??? Remember Prolog?
The truth of the matter is that great ideas for ways around the performance limitations we are experiencing have existed for decades, and we just haven't got the GUTS to make the jump.
I dream sometimes of having the courage to throw away my imperative mind, and program in a world of logic (PROLOG) or make parallelism my mantra (F.P). I haven't had the guts yet, I always fall back to LDA #$0F, STA $D20F. have you had the guts to do it? If you have then please followup.
We need to hear from people who made the leap from the imperative world of constructs we are all so familiar with and immersed themselves in those other worlds.
--tarp
You can find it in the trunk of your flying car, right next to your alcohol-powered laptop.
"Whoever would overthrow the liberty of a nation must begin by subduing the freeness of speech."--Benjamin Franklin
Ok, I'm sure I'm missing lots here, but one of the reasons we like silicon is because SiO2 is an insulator, it's a solid, and it's insoluable. We use SiO2 very heavily in CMOS electronics, upon which all of our "fast CPUs".
Well, I may be remembering this incorrectly, but I think one of the barriers for GaAs, besides being a more expensive material, is that when you oxidize GaAs, you get something which is water-soluable.
Kinda kills the idea of doing CMOS in GaAs which throws us back into the 70's when people still used TTL for VLSI.
i was merelay talking about bandwidth...
and btw i guess you know better then:
Physicist Anthony F.J. Levi of University of Southern California
Michael Morse, Photonic Researcher at Intel
Mark T. Bohr Director of process Architecture at Intel
and W. Watt Gibbs, author of the very interesting article i got my information from in the November Issue of Scientific American.
And btw those are just a couple of names taken in the two first paragraphs or so from the article.
Don't sell me faster hardware... sell me more efficient software and peripherals.
Speak truth to power.
Actually they have proactivly developed new thingys ;)
I'd like to note that the average 3Ghz PC can do MORE than the eqivalant of a 10Ghz 5Mhz 8086. Don't forget that it's not just your CPU doing math now days, it's that fancy $400 super-computer rivaling video card you've got too.
TeePees Wikiups, TeePees Wikiups, sounds to me like you are two tents.
Sounds great. Use your CPU and recycle the heat as a major power source for a household. Large offices may sell their surplus. A Brave New World is waiting!
daim
Thread Level Parallelism
Tim Bray concurrently covered a simmlar topic in Software in the TLP Era and offers some strategies to deal with the coming MultiCore chips.
"I will speculate that today's single-threaded applications as actually used in the field could actually see a performance boost for most users by going to a dual-core chip, not because the extra core is actually doing anything useful, but because it is running the adware and spyware that infest many users' systems and are otherwise slowing down the single CPU that user has today. I leave it up to you to decide whether adding a CPU to run your spyware is the best solution to that problem."
Sounds like users dont really need faster processors, just adequete spyware protection & removal. But I cant wait for a dual core AMD-64 system.
We have the best government that money can buy.
How does this stuff compare to Burton Smiths MTA/Tera stuff?
> If you were to surround them with a hot spherical
> shell, then they would become HOTTER than the average temperature of that shell
Right, but if you can keep the shell temperature low, you would then also set a limit on the core temperature. If liquid cooling can keep a flat chip at 30C, the center of a likewise cooled sphere should be well within its normal operating temperature range.
Even more important than bad (or less optimized = more cost effective) programming is the fact that the size of the data we want to process keeps getting bigger. Examples where this is obvious include things like computer games, meteorology, data mining, physical simulation, etc. There is still plenty of need for CPU power because each advance can be met by a desire to work on bigger/harder problems. But these bigger problems don't occur anymore in word processing, web browsing, or doing business forms, so computer gaming is the most activity where more processor can be useful.
One the main points of TFA is that Moore's Law has continued to be true while clock speed has stalled. Actually you don't even have to RTFA, you can just look at the pretty graph on there.
Back when Vitesse finally started getting decent yields on its communication components I bought shares that eventually doubled several times. I admit that a lot of less substantial technical achievments have made equally substantial profits in the market of the mid 90's. I don't know squat about stock markets and the only thing that kept VTSS from skining me alive later was my wife's advice to sell half the shares any time it split or duoubled. In 96, the typical stock analyst didn't know squat about GaAs so I had a temporary advantage.
Just wanted to note that a few GaAs promises have been kept. [looks like I should invest in companies that make good heat sinks and fans?]
SLASHDOT: news for people who can't concentrate on work or have no life at all and got tired of yelling back at the TV.
What you're missing is that the main source of increased speed for silicon was decreased feature size. This geometric scaling of speed is pretty independent of the material. What is not independent of the material is the electron mobility and capacitance.
Seastead this.
As the cold war ended there was a shift away from such performance-based purchases to idiotic programs where really really "smart" guys would divine the next wave in wonderfulness and have people with Ivy League and press connections hang around the halls of power schmoozing it up a real whole lot so they could get funding in advance of technical accomplishment.
Seastead this.
It may be 2005 but my P3-450Mhz still packs a punch for my wife.
:)
She's blissfully unaware that she's now using the 450Mhz instead of the super duper AMD 3500+.
From browsing the internet, to playing Solitaire, to reading email, even this 450Mhz from 1999 works like a charm.
Forget the processor, today the bus speeds are like 3 times faster.
This makes us wonder why do you really need a 10GHz PC ? Perhaps you can lower your heating bill using this as a heater
Not enough to count for anything. But Native American history is a hobby, and I am a Indian Hobbiest (check out www.powwows.com) it is heavy with the java/multimedia/etc but if you can get past that a lot of good info on pow wows and both Native and hobbiest.
Since when is this topic news? I hope it isn't just because DDJ mentioned it. I don't feel like dredging up pointers to the bezillion pages on the matter but there has been academic and industry handwringing about the inevitable limitations of transistor size and speed for a decade. OK, one URL, thats all you get! [rice.edu] read pg 62 for consice four year old description of the issue [and how carbon nanotubes are going to save us]. The predictions of the exact day when progress stops were always a bit vague and hedged with hopeful notes about gallinium and going to 3-d circuits...all that is really happening is that, having seen the wall a long ways off, chip makers aren't going to smack into it head on with an abrupt cesation of speed increases but veer off in new directions and so only slow down the increases.
/. interest...but you could have had this conversation anytime in the last 5 years!
at 500 comments in 2 hours, there is clearly healthy
SLASHDOT: news for people who can't concentrate on work or have no life at all and got tired of yelling back at the TV.
there will always be money to be made in the market of technology surpression. its why we still run our cars on gas.
1001100 1100101 1100001 1110110 1100101 1001101 1111001 1000010 1101001 1110100 1110011 1000001 1101100 1101111 110111
I find that excessive.
CGI is something I'd like to dabble in for future projects, I know it's a resource hog.
every day http://en.wikipedia.org/wiki/Special:Random
And "just" not in the same magnitudes (over time) as we saw from those previous clock speed gains. Which is the point of the article. Doubling cache or going multi-core has much smaller returns than doubling the clock speed for most apps.
I hate to say it, but what do you think you need 10GHz for anyway? Unless you've got a REALLY fat pipe, there's a limit on how much pr0n you can process ;^)
The same thing you needed 1 GHz for when 300 MHz was state-of-the-art. And don't ever ask that again or you'll have to turn in your geek card.
The differance is that MultiCore is going mainstream. Intel, ADM, IBM, Sun all have chips we can buy in the next year. Not as powerfully as Trea, but a lot cheaper.
My point would be that - regardless of why its happening - there is a major change it hardware happening, Like the change from Intel 16 bit to 32 bit, which took several years, but much more difficult. (we are also going 64 bit is servers and desktops to address all that chaep ram, but that is also most a non-story.)
The industry has been talking about Parallelism for a long, long, time. Looks like we a now starting down that path. Its going to be hard, but it is rewarding. Which is part of Herb Sutter's and Tim Bray's message. The OS's, tools, and applications that do this right are going have a big advantage in the server and desk/laptop market with all that highly scablible hardware.
Could we do a multi-processor system that splits the tasks according to their horsepower needs? The OS splits the tasks the processor that can handle it, an no more. Multilevel, Trickle UP CPU power. Say, for sake of argument, set up arbitrary CPU usage levels
relegate the slow, stupid stuff to a lesser processor, like Notepad, some sniffers... look at your Task Manager processes (or equivalents) for stuff you could be running on a 80286 and shove them to the CPU slums. This level keeps the lights on, controls the heat. It's the oil and the water pump on your Ferrari.
Runs the OS functions themselves, if they can. File transfer, TCP/IP, simple Multimedia like MP3 & CD, Virus, Spyware detection... This level replaces the Pentium ][ in your kid's room running Limewire. It's the Stereo on your Ferrari
The GUI, the heavy multimedia, like video, CD burning, the bloated web rendering... this level makes the UI responsive. It's the suspension on the Ferrari
The Engine of your Ferrari! The monster throughput... The Doom VII, the Celestia: Andromeda, SETI:Alpha 6, Climate Prediction, and of course the rendering of the perfect sig-other on Maya. When yer not using this part, shut it down, despin the fan and listen to the sound of silence.
The offshoot of this is that there can be a Level 5, if one were to plug into a cluster. From there, you can plug into BIGGER clusters until you either reach Blue Gene/X or you ARE part of the cluster.
Why not?
I hate to say it, but what do you think you need 10GHz for anyway?
When I bought a 160GB drive a couple of years ago, my linux kernel only saw 130GB of it. I laughed, since it didn't matter. I wasn't going to use 130GB, either.
Right? Right?
Well, wrong.
I bought a DV camcorder, and started pulling in all of our old tapes, even my wedding tape. Now I have 250GB or so of video on the LAN and I'm adding more HD space.
What will I use a 10GHz processor for? I'll find out when I get one. Probably something with video...
Do you have ESP?
"A machine that can do realtime interactive full motion video".
Only in a very limited way. The absolutely latest "realtime interactive full-motion video" - Halflife 2 - is just beginning to do optical transparency and realistic physics in a reasonable way. It's only good enough to fool the eye for a few seconds, if you pay attention to shadows and reflections, but it's the first system that's managed to provide anything like photorealistic rendering.
Luckily raytracing is very parallelisable through brute-force techniques (give each GPU a full copy of the polygon database and a segment of the screen to render, and Nx as many processors will be able to give you Nx the framerate... up until pushing out the polygons starts taking a significant part of the CPU), and luminosity (shadowcasting is a really crude kind of luminosity calculation) and the physics engine can be run separately from the final engine...
So "realtime interactive full-motion video" is potentially possible, but we're not there yet.
GPUs are good enough that a realtime 3d window manager is possible. It's a much simpler problem than FMV, and Apple's beginning to play with some of the features in Quartz Extreme and Exposé, but until someone actually produces something like my fantasy 3dWM we won't see much demand for it.
You can now have even room temperature super-conductors, but these are (I think) not 3D superconductors, but 2D supercuducors (surface superconducing). They do not work for any application with any power demands (eg. magnets or power distribution), but then the purpose of a CPU is not to trasfer or translate power.
They are called compound semiconductors. Intel is already funding research in III-V devices.
:D
Yes, I know this was lame on the font of all the business speak
Why wouldnt programmers add more to their software? Its kind of like developing and using a 10,000 horsepower engine, but never pushing the throttle above 10%.
The performance gains should initially be about the same as having a true dual-CPU system (only the system will be cheaper because the motherboard doesn't have to have two sockets and associated "glue" chippery), which means something less than double the speed even in the ideal case, and just like today it will boost reasonably well-written multi-threaded applications. Not single-threaded ones.
That's OK, for most people dropping back to 1 GHz and getting half a dozen cores would give them a lot nicer computer than pushing the CPU speed to 4 GHz on a single chip. Why? because there's gobs of concurrency on your desktop already no matter what your processor... no mater what your OS, there's half a dozen processes (tasks, whatever) that each take turns using a lot of the CPU and occasionally stepping on each others toes and making your computer feel slow.
For games, increasingly, the bottleneck is the video card. There's still a lot of straight-line code in most games, but things like the physics engine, the rendering engine, and the user interface... these parts of the system can be made to have relatively narrow communication channels between each other. Each can then become a separate thread, or even a separate program, running on a separate processor. Overall a 6x1GHz system would still be able to provide a better experience than a 1x4GHz one, even for games.
This won't make the benchmark boys happy, but who cares? They haven't been happy in a long time anyway.
In my office at home, I am running AMD 3000+ for my XP box and AMD 2200+ for my SuSe 9.1 box. If you look at the S.M.A.R.T. number off of the MOBO sensors, the temps are scary. Running one "in" fan and one "out" fan on each box, in addition to multi-speed power supply fans, the 8' * 10' room gets rather warm, too. As the density of chips increases, and the clock frequency increases, we are getting to the point where liquid-based cooling is necessary on gamer machines. Is this where we all want to go? The short-term answer is dual-core at lower clock speeds. Assuming, with MP factors that 1+1 = 1.9 or so, a multi-tasking system can give a pretty good illusion of being a honkin' single image. The other short-term answer is multiple, slower CPUs for the same reason; spread the heat dissipation across more real estate on the MOBO to keep from frying things. Remember - in the world of electronics - Heat IS NOT your freind :^)
My wife doesn't listen to me either...
This is exactly the premise that BeOS tried to sell a few years ago.
Maybe they were just a little bit ahead of their time.
war is the best time for technology, however it has its draw backs, like death.....
That my NES emulators aren't challenging my brand new P4, or utilizing my gig of ram to peak performance??
Of blankness, I know nothing.
Well I guess you can't call it a Law anymore if we are hitting this road block.
Moore's Idea? Moore's Philosophy?
Perhaps some day, we'll have processors dedicated to physics-- a PPU? Perhaps it would model a simple subset of physics: rigid body collisions or rigid jointed bodies.
If you've read this far you'll realize what a tweerp you are & all others when you realize the counter to this agr.. is on the same slash page as the initial listing...l ogy/articles /2005/01/05/intel_researchers_build_laser_on_chip? mode=PF
http://www.boston.com/business/techno
Maybe we finally have a concrete wake-up call for America to stop relying on the fact that it is America and embrace innovation. Innovation creates jobs, innovation builds an economy, and innovation ultimately saves lives.
Actuallly, I generally find that the more pr0n I process, the fatter my pipe gets.
I hate to say it, but what do you think you need 10GHz for anyway? Unless you've got a REALLY fat pipe, there's a limit on how much pr0n you can process
Oh yes, how we all love to "process pr0n" with our "REALLY fat pipe".
So it has water cooling... so what? .. and "lying out his ass" is a bit strong. Jobs will have said that because IBM told him that they hoped to get there. They, and Intel, found the next step in development of their cpu's harder than they anticipated.
Isn't the speed of PC's supposed to double every 18 months? Someone needs to get out a periodic table and find something better than silicon. Slashdot comments, however, seem to double every 30 seconds. ~T http://www.ModLife.Net
ModLife.Net - If it ain't modded, what's the point?
As for the water cooling part, you clearly didn't read the message. Water cooling on the G5 is obviously a stopgap measure because of the insane heat levels. They wouldn't voluntarily start using water cooling. Nobody would.
I suspect Apple will demonstrate a 3.0 GHz G5 within a month or two, but it'll also be water cooled and won't find its way to consumer products until the end of this year, volume shipments much later.
I'm sorry, the G5 isn't a miracle you'd have hoped for. My money is on the Athlon 64, which can be clocked well past 3.0 GHz (maybe 3.5) on air at 90 nm process technology and dual stress liner strained silicon. These chips are due to be shipped in the first quarter of 2005, though AMD will probably play it safe and not release insanely fast chips even though they have the chance to beat Intel repeatedly with a large stick.
You typed 10Ghz right? I look up 10 and it means 10. The capital G is a metric prefix for one billion "Giga", and the abbreviation for hazard is hz. That's what you meant isn't it? Surely you didn't trip over one of my pet peeves, ignorant of the work of German Scientist Heinrich Hertz (who spelled his last name with a capital H) and accidently (or foolishly) droped the capitalization!!! Next thing, you would replace the metric prefix M (for mega, representative of million), with m (for milli, representative of 1/1000). I saw a lot of computers for sale in stores advertised as 900mHz. 900 millihertz, or worse 900mhz --900 millihazard, whatever that is. When I was studying engineering, people would occasionally get this wrong (and thus the whole answer was wrong). I was told that even if you didn't give a sniff about the engineering "I only do business, not of that yukky engineering stuff", that legal business contracts with incorrect metric prefixes are legally binding as stated. If you order 200 1800 mHz computers for $400,000, they can sell you 200 old IBM XT's, since they excede the contract by running at 2.77 MHz, and all you should have expected was 1.8 Hz --1.8 cycles each and every second. It's wouldn't be their fault that you couldn't tell the thickness of a dime -1mm- from 621 miles -1Mm - either.
All that talk about how great OOP is and nobody challenges it? I have not seen any good evidence that OOP improves the maintanence of complex software. Surveys don't show any lower failure rate for OOP software. Ed Yourdon did a survey of such in the 90's.
Table-ized A.I.
Article doesn't say when this was written, but we're already beyond 3.4 ghz. You can but a 3.6ghz on Pricewatch right now and Tom's Hardware announced a 3.8 ghz being released way back in November 2004, so I don't know why the author is saying 3.4ghz is the top of the line currently and Intel will release a 3.73 ghz in 2005.
When was this written, early 2004?
I do have to laugh a little at it: "This article will appear in Dr. Dobb's Journal, 30(3), March 2005. A much briefer version under the title "The Concurrency Revolution" will appear in C/C++ Users Journal, 23(2), February 2005."
With the author saying "3.4 is out now and 3.73ghz early 2005" while everyone's upgrading to a 3.8 that's been out since 2004 it's gonna make that article look very old and inaccurate.
my karma will be here long after I'm gone
I think what has started to happen is that because CPU's are running pretty fast nowadays, they are now starting to upgrade the rest of the computer to keep up with the CPU.
As I said in another post here, I wrote these things are now happening to desktop computers:
1. The development of faster motherboard interconnects with improved chipsets and things like HyperTransport and its competitors.
2. The wide availability of PC3200 (DDR-400) DDR-SDRAM system RAM, with even faster RAM coming over the next 18-24 months.
3. The development of AGP 8x and new PCI Express connections for graphics cards with 3-D processing ability that would be the domain of ultra-expensive workstations only a few years ago.
4. The development of ATA-100/133 IDE, Serial ATA and soon Serial ATA-II IDE, and UltraSCSI 160/320 interfaces and 10,000+ RPM drives with 8 to 16 MB on-drive memory caches for very fast hard disk access. Even optical disk drives are benefiting from these faster interfaces.
5. The very wide availability of 100Base-T Ethernet connections on most motherboards, plus some motherboards now sport 1000Base-T Gigabit Ethernet connections.
6. The near-universal availability of USB 2.0 connections and increasing use of IEEE-1394 connections to external devices, which makes the use of external disk drives to back up data and connect to digital camcorders possible.
These improvements in the rest of the computer means you don't need the fastest CPU to get much-improved performance over desktop machines of a few years ago.
mmm. Warmed my lap.. and did a little bit. 0% idle.
Mangee: Father to Lliam
Yes, maybe we have reached this limit for the moment, but I don't think the decision to just put more cores on a chip will fit well with customers. I wrote a small article about this a while back on this page Basically concurrency is too hard for the average developer to use in any single project. It opens the floodgates to a mighty source of errors. No customer will want that. The Industry obviously has got to invest in R&D for a more acceptable solution.
I'm posting really late on this article so no one will read this but the line of best fit for those points is shockingly out! If i was to draw it then the 10000ghz would surface in 2008
My journal entry here contains a nice little article on why Moore can show his law into his pipe and smoke it.
:-) I like to think so.
Also, Smoking Gnu in Going Postal... is this GNU??
#hostfile 0.0.0.0 primidi.com 0.0.0.0 www.primidi.com 0.0.0.0 radio.weblogs.com
Flying cars
Duke Nukem Forever
Windows now crashing...
Tibbon
tibbon.com