Yeah, you know that the new discussion system is totally broken on IE6. Of course, I knew this six months ago when I elected not to test it, and since then they have fixed nothing.
What's with the duplo-block-sized titles, do we suddenly have armies of babies and old people reading the site?
And to stay on-topic: my stepfather was working for the census while they considered this transition, and it was the most painful decision they had to make in all his years working there. Digitizing something as flexible as paper meant that you actually HURT efficiency of data collection. Think about it: with paper, you can easily correct mistakes, skip questions (or go in a different order). Most important: with the computer, you're SOL if you drop the computer or the battery dies, or the software crashes.
And while digital data collection reduces costs at the back-end (the data is already digitized), the fact is that collecting the data is the most expensive part of the census process, and any increase in costs there erased the gains at the back-end.
On an eight bit display, each pixel is made of three sub pixels, each with one of 256 possible values. These displays only have 766 different colours (3 x 256 = 768, but that counts the colour black three times, so there are only 766 colours). The 16.7 million colours are just created by dithering.
You're not using the data properly. The three colors of an LCD pixel are blended by our eyes to create 2^24 distinctive shades, the same way your eye blends the red. green and blue phosphors of a CRT pixel. For your information, we include three different "blacks" because they are three different colors:
Say you set R to "black" and G and B to other values. You now have zero Red component; this is how you remove colors. If you make all three components black, then you have true black.
By having 6 bits of range versus 8 bits, this means that you will lose FOUR DISTINCT colors for each single color you can create in the 8-bit range.
Say the numbers 0 - 255 (8-bit) map to a color range. Now say you map the numbers 0 - 63 (6-bit) over that same range. That means 63 and 255 represent the same color, and 62 and 251 represent the same color...but what about the colors 252, 253 and 254? These are all represented by one value, now 63, unless you do some dithering.
Dithering methods: typically, to "recreate" more colors, 6-bit displays will cycle between the two "closest" colors to create a color outside it's rendering range. Thus, in order for our 6-bit display to create the color value "253," it alternates between displaying the color "63" (maps to 255) and the color "62" (maps to 251). You use similar methods to create the 8-bit colors "254" and "252", using different patterns.
What you end up with is a marginal image, because contantly changing pixels means that the image is blurry and loses contrast. This is why most 6-bit displays look like crap in darker scenes. You also don't end up with quite the color fidelity of a true 8-bit display, because it's just not possible to create, even with good dithering techniques.
One of the problems you have is that gate volume is approaching thousands of atoms. This is a problem because certain regions need to be doped in order to make the silicon do it's job.
What is the problem, you may ask? Well, just look at the Wikipedia entry you linked. Even doped silicon is still %99.999999 pure.
So, you have your gate that is thousands of atoms in volume, and dopant concentration that is in the 1 million to 1 billion ratio...so what is the likelyhood that your gate is going to contain that one dopant atom you need in the lattice for best performance? Without that dopant, the performance of your gate suffers, and may not even work at all.
I didn't understand who has a population density 5 to 10 times larger than Brazil
You do. So does the UK, and Germany. Specifically, 10 times that of Brazil.
Italy: 197 people per square km, Gini = 36 UK: 246 people per square km, Gini = 34 Germany: 230 people per square km, Gini = 28.3 United States: 31 people per square km, Gini = 47.0 Brazil: 22 people per square km, Gini = 56.6
I understand that poverty can often sadly lead to crime.
And just to give you an idea of how poor the poorest people in Brazil are, I've provided Gini numbers for each country. These numbers are a measure between the poorest and richest residents of a country, where higher numbers indicate a larger divide. You will notice that in Europe, where the poor are not all that poor and the population density is high, crime is not a problem. In the US, things get worse, but are not completely out-of-hand. In Brazil, the situation is a nightmare compared to Europe.
About Germany I think people should always remember which horrible things the nazism did but should stop blaming german people that actually live in Germany and have nothing to do with it, because let's face it now it's history and most of the people of that historic period are dead or very old now.
The rest of the world stopped blaming the Germans years ago, including myself. The only people blaming the Germans are the Germans themselves. I mean, really, what other free democracy in the world makes it illegal to form a group or have a webpage promoting hate speech or more specifically, neo-Nazis? They're completely stuck in the past.
So tell me why all this doesn't happen in Europe...
Because, like rail mass-transit, you get more police effectiveness for your dollar when you have a population density FIVE TO TEN TIMES LARGER than Brazil. Also, the people of Brazil aren't fat and happy like the west, because a good chunk of them live in huts below the poverty line, so you get worse gang violence than you'll ever see in the first-world. Police can only do so much when the population is restless and spread-out.
And let me make one thing clear: ultra-strict gun regulation is not a common theme in Europe. This is mostly restricted to the UK (nanny police state, complete with closed-circuit TV to watch the fun), and Germany (yes, the entire country is still embarassed about Hitler, even today). In France, Switzerand, Finland, you name it, gun ownership is high; a lot of EU countries even allow handguns for home defense/target practice use.
The Super NES had a 16-bit CPU with ~65,000 colors. But PC gamers were already upgraded to 16 million colors.
First: SNES only has a 32K color palette (5 bits RGB = 15 bits). VGA has a 262K color palette (6 bits RGB = 18 bits), and SVGA has a palette of millions of colors, as you stated.
Second: the number of on-screen colors is not the same as the palette. The number of colors on-screen was the same for VGA and the SNES, which was 256, and these could be selected from the palette. Most SVGA games also stuck to 8-bit color, because it was more efficient.
Where the PC had the advantage was higher resolutions, with early SVGA games pushing 640x480, and later games offering 1024x768 or higher. Higher resolutions easily made up for the low number of colors on-screen, because this allowed game designers to use dithering.
The console is typically one generation (4-5 years) behind the PC gaming world, becuase using older technology helps keep console prices low (~$300).
I'd disagree with your definition for a PC "generation." In the PC world, things move fast. The N64, for example, had parity with midrange PCs when it was released, mostly because there were no affordable 3D graphics accelerators for PC at the time. Within a year, PC graphics eclipsed the best the N64 could muster. Another example: the Xbox featured GeForce3 graphics, which were only a year old at the time of the system's release.
So sure, if you decide to slap the "PC" label onto everything, then yeah, the PC market is doing fine. Meanwhile, I don't think nVidia is going to have a strong season selling top-end video cards to only the people who bought Crysis
Nvidia doesn't make any money selling these high-end cards. These high-end cards exist for only one reason: inexpensive viral advertising. If your top-end card gets the most 3dmarks, it tends to affect buying decisions for purchases across the board.
The viral advertising works two ways: first, you get all these wonderful reviews that remind you that Nvidia exists; the reviewers show you incredible numbers, then as an aside tell you that the "real deals" (i.e. 9600GT, 8800GT) can be found for much less. Then you get the second effect, where hardcore enthusiasts read these articles and then preach the gospel to the masses via forums/blogs.
I mean it when I say that Nvidia doesn't make any money on ultra high-end; the largest segment for their earnings comes from the $200-250 segment, which has been populated by the venerable 8800 GT for the last six months. The 8800GT alone literally stole %5 of the desktop graphics market from AMD last quarter, even up-against the competitively-priced 3870.
PC graphics cards are only expensive if you insist on top-end. Example: you can get more graphics power than RSX in the PS3 for a paltry $100 today, and by next year the same power will be available in an entry-level $50 card. The graphics core in the PS3 used to be sold as the 7800 GTX, and sold for $500 only three years ago; now you can get MORE power in the $100 8600 GTS.
A better comparison would be the Sega Master System to the NES. The SMS was far superior hardware-wise, yet still barely made a dent in the marketplace.
Far superior? Not really. It just bugs me that this gets passed around all the time.
Same general CPU power (the Z80 is clocked twice as fast, but has more wait states per op than the 6502). Same resolution (256 x 224). Same maximum number of sprites (64, but the NES has flicker if you have > 8 on one scanline).
The sound on the Japanese Mark IV was better, but the first-generation FM synth didn't sound all that impressive. Really, the next-generation of FM synth in the Genesis was far better.
In the USA, they dropped the FM synth, so the NES had far better sound (with an extra PSG channel), and a MUCH stronger noise channel (anyone who has actually heard a real Master System knows the noise channel sounded weak and pathetic).
For graphics, yes, the Master System featured 32 colors onscreen without tricks, whereas the NES could only muster 25. However, most games at the time could make little use of all those fancy colors because they had pathetically small storage space. In the end, very few games took advangate of the impressive 32-color limit; we remember those few games as standouts, but the fact that there were so few of these games is one reason the console sold poorly.
The reason the SMS sold poorly in the US was because it was late to market compared to the NES, and didn't really have solid advantages. I only knew one person who bought an SMS while growing up, and he only bought it because his brother (and I) already had an NES.
We did the same thing when DVDs were first being ripped. We didn't have cheap DVDs, but we did have cheap CD-Rs, and lots of free time. With the pioneering Divx codec, attempting to put a full movie on a 1-2 CD-Rs wasn't all that crazy. By the time DVDs were cheap, you could actually create a fairly good copy of the movie on a single CD, or a pristine copy on two CDs.
Today, we have single-layer DVDs that are dirt-cheap, and dual-layer DVDs that aren't all that expensive. The single-layer disc is more than enough to store a 720p h.264 rip with reasonable quality, and the dual-layer disc can hold a 1080p rip using h.264 with only marginal quality losses.
The human eye's 'refresh rate' is around 60 Hz... If you think you're being distracted by flicker from 100Hz, you're only fooling yourself.
No, I'm not fooling myself, you just can't see what I see. You are the one fooling yourself; if you can't see it, it does not exist.
Let me give you an example: I run my CRT at 85 Hz. I can tell the difference between 60 Hz (tons of flicker), 75 Hz (marginal flicker) and 85 Hz (no noticeable flicker, except on pure white screen). Why do modern CRTs have flicker problems at the same refresh rates televisions have used for years? Because they have lower-persistence phosphors, which makes the difference between a TV and monitor at 60 Hz visible to a subset of all users (i.e. me, and the other complainers).
Why are LEDs a problem? Because their persistence is even lower than CRT phosphors. Also, the LED taillight is an array of PIERCING-BRIGHT point-source lights at a single wavelength. Bright red lights have always bothered me, but red LED light arrays and red lasers are even more annoying. Combined with the flicker, it can give me a headache.
What realy pisses me off is that large-scale automakers like Honda are starting to put the damn things on every car.
Why don't you do yourself and your brother a huge favor and buy him a couple 1GB sticks for his birthday? Unless you live in the Amazon or some such technical wasteland, memory is dirt-cheap these days.
Than you don't have to ask tough questio s like this; instead, you can just install the service pack and get on with your life.
DYNAMIC POWER = FREQUENCY * CAPACITIVE LOAD * VOLTAGE^2
The above ignores leakage, but as another poster mentioned, that is not related to frequency. Leakage actually scales LINEARLY with the device voltage.
Adding more cores DOES increase power linearly. but the frequency^3 comment is completely off-base. The worst offender is actually voltage, which adds quadratic dynamic power and linear leakage power. As you raise the frequency, power consumption can increase even more than just linearly, because you need higher voltages to drive higher frequencies.
Example: say you had a 2.0 GHz, 1.0v processor, which used 20w dynamic power. Now, let's say you want double the performance (4.0 GHz), but that requires 1.2v to drive it. This would give you:
20w * (4.0 GHz / 2.0 GHz) * (1.2v^2 / 1.0v^2) = 57.6w. You increased performance 2x, but had to increase dynamic power by almost 3x!
Thus, the designers of processors usually target a sweet-spot with a good combination of low voltage and marginal frequency. Adding more cores allows them to use the same parameters for each core, which yields (potentially) a 1 to 1 performance to power increase.
Gosh, I hope it is more effective, because in my implementations I actually saw a slowdown instead of an advantage. Even then I'm generally not happy with hyper-threading. The OS & Applications simply don't see the difference between two real cores and a hyperthreading core. If I run another thread on a hyperthreading core, I'll slowdown the other thread. This might not always be what you want to see happening. IMHO, the advantage should be over 10/20% for a desktop processor to even consider hyperthreading, and even then I want back that BIOS option so that disables hyperthreading again.
SMT, or Hyperthreading, is intended to reduce the overhead of task switching. Instead of clearing out the pipeline and and switching tasks at the OS level, the processor automatically detects a wait state on TASK_A and switches to TASK_B in a single clock cycle. There is no magical performance improvement, the goal is to reduce the overhead of task switching.
This was intended to hide the horrible I/O latency of the Pentium 4, and it did a decent job of it (on the server). On the desktop, it caused slowdown because HyperThreading added processing overhead, and most applications were not multithreaded OR I/O limited. If you didn't have multi-threaded code, you actualy got lower performace due to the overhead.
Well, now Intel sees two things happening in the market: one, Sun is making inroads with their Niagra architecture, using simple cores with SMT to produce high performnace on I/O-bound tasks. Intel sees this as an opportunity to reintroduce SMT to compete with Sun in the low-power, high-I/O market. On the desktop side, games are using more resources than ever (finally I/O bound), and some games can take advantage of as many as 4 threads. Having zero switching time between those I/O-hungry game threads could be impressive.
No, it certainly won't be all that impressive for the desktop, but at the same time it won't be the performance detriment it once was. The server market will LOVE it, and that's what really pleases Intel.
Can count all the major improvements - that you know/can think of, sure you've demonstrated that. What about all the improvements that occur outside of the processor instruction set? Such as improvements to branch prediction
The last two times branch prediction on x86 saw a HUGE improvement were with chips like the Pentium (which introduced a 2-entry branch history table), and the P6 and the K6, which introduced much more powerful branch history implementations. Everything beyod the P6/K6 generation has been minor tweaks; when you get above the most basic implementation of the branch history table, your performance improvements are tiny for the general-use case. Even additions like the loop predictor on the Pentium M don't add much, when you realize that most loops have lots of iterations.
Improvements to the pipeline so that branch prediction misses aren't so expensive
This would be great, except that most CPU pipelines ELONGATE as you add more features. Even the saga of the P6 -> Core2 has added more pipeline stages. And when you add more stages, you need better prediction, but there is only so much you can do: the Pentium 4 featured one of the most advanced branch predictors ever (but still only slightly better than previous), plus hyperthreading, all in an attempt to reign-in the cost of clearing the pipeline from branch misses (and other I/O). It didn't work.
(as well as additional instructions such as condition move that can reduce need for branching)?
Does not exist. You can execute both branches (at the expense of memory overhead) like Itanium does, or you can make a prediction (like all modern x86 processors), but only the programmer can actually remove branches.
Improved out-of-order instruction execution,
Which, as I highlighted, hasn't really improved in years. The Athlon took one step above the P6/K6 by adding a third floating-point unit, but kept the 3-wide decoders. The Pentium 4 added an ungodly number of rename registers, but that's mostly to handle the long pipeline and HyperThreading, and it didn't really improve performance per clock (see above).
The Core2 and the Phenom are the the first processors to actually improve out-of-order execution in years, with a combination of wider decode/retire path (Core2), support for packed SSE instructions (both), two 128-bit SSE units (both), and support for out-of-order loads/stores (both). The wider SSE registers double the potential performance for SSE2+ code, and the 4-wide decode path on the Core2 allows you to keep the integer pipes full while the SSE pipes work their magic.
fast loading/saving of contexts useful for multitasking, and of course, hyperthreading (okay just kiddin on the last one;-)
I grouped these together, because they are essentially the same. Hyperthreading allows fast task switching without involving the operating system, because it makes one processor behave like two.
Don't kid, HyperThreading isn't a bad idea, it just had no place in the (largely) single-threaded univese of the PC. Intel added it to bandage the P4's poor I/O and branching peformance, but it wasn't enough to make the difference (plus it slowed-down single-threaded apps). On I/O limited processors (like Sun's Niagra line), SMT is a wonderful thing.
improvements to the MMU such as global and large pages mean less TLB misses, memory prefetching, greater memory bandwidth,
Large pages, memory prefetch, larger caches - they all add their little bits, but nothing extrodinary. On the Pentium 4, for example, the move from Willamette to Northwood (double the cache, plus prefetch) yielded a %5-7 performance improvement, and nobody knows whether the cache or the prefetch were more important. Those are small potatos.
And as for memory bandwidth, it is not a question: it will improve, or processors will starve. But the amount of input from the processor side is startlingly low: only four times since the introduction of the 8088 have processor manufacturers widended the memory bus (16-bit 286, 32-bit 386, 64-bit P5, 128-bit P4/Athlon), and only with the 486 did we actually introduce on-chip cache.
Clock frequency is not an indicative of CPU performance. For example, the Core 2 chips, despite generally operating at a lower frequency than the Pentium 4's outperform them significantly.
But massive instruction per clock improvements do not happen very often in the x86 chip industry. In fact, I can count all the major improvements for the last 15 years on one hand:
1993: Intel Pentium Pro (approximately 2 INT, 2 FP operations per clock, best case) introduces real time instruction rescheduling to the x86 world. The design can decode 3 instructions per clock. Yes, I am disregarding the Pentum, because you got NO performance improvement without an optimizing compiler.
1997: MMX increases maximum number of integer instructions to 8 per cycle. But, because of the 64-bit data size, you really see little improvement unless using 16-bit or 8-bit types.
1998/1999: 3DNOW! and SSE double the potential throughput for 32-bit floating point, again not all that impressive.
2001: the Pentium 4 actually REDUCES performance per clock, with a single instruction decoder, and heavy reliance on trace cache to make up for this. SSE2 gives the potential to increase FP thoroughput to 4 instructions per clock, per SSE unit, but a half-assed implementation by both Intel and AMD means nothing changes.
2006/2007: the addition of more decode units on the Core 2, packed SSE instructions for both the Core 2 and the Phenom, and TWO 128-bit SIMD units means we see the first improvements in instructions per clock in years.
Also the interface for uploading songs was conected to the _parallel_ (LPT) port of the comp. It was pretty unstable. The filesystem was also not FAT12/16/32 based so it was a hassle to get the songs on the player a few yeras after when it was hard coming by Win98 (for which the software was written).
This is the major reason why I held-off buying an mp3 player in the dark ages of portable players. The idea of spending hundreds on a device with a proprietary dongle cable that used a slow bus (parallel or serial), and depended on some random music transfer software did not appeal to me.
Creative Labs was the first manufacturer to get my money because they released the Muvo, a 128MB wonder. Not only did the Muvo use the USB interface for faster transfer speeds, it required no dongle because the USB port was built-in to the player. The battery case detached from the player, revealing a full-sized USB connector. This ability, combined with the UMS disk mode, made it easy to transfer files, and made it feasable to use as a thumbdrive.
It goes one step further than this: deletion trolls today can take advantage of several tools that Wikipedia has made available to them, including listings of all newly-created pages, and changes to pages. This makes deletion easy, whereas finding older pages that are not up to standards is a little more difficult.
Moreover: there are so many scripted and "in-the-know" things on Wikipedia today, I can't make a "standards-compliant" page to save my life. For example: how the fuck do you put in a standard references block, and refer to those references in-article using superscript links? I've looked at several other pages to figure out how this works, but the code seems to be hidden. All the strange codes you can use to mark pages, like say "this article needs expansion," are not really easy to find. Compared to the ease of creating my first Wikipedia page, creating a page today that will susvive the hounds is near impossible.
I guess it's not surprising that I stopped contributing to Wikipedia recently.
One small problem, AMD's contract with Intel states that they can't outsource more than 20% of their chip production.
I've seen this line dropped a million times on Slashdot, and I have NEVER seen anyone back it up with proof. This is turning into the ultimate geek internet rumor, because everyone parrots it verbatim without checking the facts.
I say, put up a link, or shut the hell up. Unless you have a news article or a copy of AMD's contracts, you have proof of nothing.
The fact is, you don't even need a CPU license to make an IA-compatible chip (just look at the Crusoe). Even NexGen, Centaur and Cyrix made chips for years without licenses, although recently they have all obtained cross-licensing agreements (out-of-court settlements)
The license simply makes doing business easier because cross-licensing means Intel won't litigate you to death. So, even if your post is true, AMD is a particularly strong opponent, and any litigation by Intel will raise the monopoly flags all over. If this actually happens, the eventual result will be another cross-licensing agreement, except with AMD having much more leverage than before (with Intel under the monopoly pressure).
Although Via is not overpriced any longer, they used to charge hundreds of dollars for their Mini-ITX boards, and thus they keep their hard-won repuation. Believe me, those skinflints deserve it.
They used to charge OVER $300 just for passive-cooled Mini-ITX boards with processors slower than 1GHz. That's not competitively priced with ANY other platform. Just as an example, the (now $60) C7 1.5GHz board you linked went for over $200 on introduction last year! And the worst part: the prices never went down, not even with new product introductions, because Via was the only game in town.
Yes, Intel delivered more performance, at about the same power envelope, for less than half what Via was charging. Overnight, Intel changed a market that Via had literally let stagnate for the last 5 years: Via had product delays, overhyped improvements, and overpriced products, while Intel simply made a board that delivered all-of-the-above. Now THAT is a market force that can move mountains!
And now Intel wants to push power consumption and price even lower than Via ever dared, with performance that will probably be competitive with the C7. I say, bring it on!
By the looks of things Isaiah will wipe the floor with Atom if intel doesnt bury Via with branding power. Isaiah's out of order execution will offer much better performance than Atom's in-order execution....and more power consumption, from the looks of it. VIA estimates the 65nm Isaiah will have the same power envelope as with their 90nm C7 (10-20w), which is not unreasonable. However, this is an order of magnitude higher than the Atom, which is sipping power at 0.5-3w. The only thing VIA has that could compete with this power-wise would be a 65nm C7, and I have no doubt Intel can design an in-order chip that can outperform the crappy C7.
That's an interesting idea, but it has some complications.
Two things:
I don't think we could stretch a Wifi network nintety miles without violating FCC rules. Twenty miles with a high-gain antenna is pushing it. You could, of course, solve this by getting an FCC exception (call it "Project Free Cuba," they'll eat it up).
Nintey miles means you have line-of-sight issues, and high frequencies like Wifi don't tend to duct/refract as much as low frequencies. You'll need a tower to solve this problem.
A nice alternative would be a boat anchored off the shore of Cuba in international waters, which lets you get around the whole FCC thing and use a much shorter tower. All you need is a satellite uplink to make the prefect relay. Unfortunately, the running cost of a ship relay is a lot more than the cost of a land-based relay, plus you'd be a nice fat target for Cuba's airforce:)
But whenever someone brakes the 3-car-space decreases (I assume) to a 2-car or 1-car space. Shouldn't this extra space compensate for the slowdown, so that the cars at the back don't have to overcorrect?
From your comments, I think you're assuming that drivers are competent, and keep three (or more) car-lengths between themselves and the next car. With competent drivers and plenty of reaction time buffer, you can avoid this problem.
To see how this problem can actually occur, I'm describing a worst-case scenario where you have cars riding bumper-to-bumper at high-speed. While this is obviously a tinderbox waiting to explode, it is not an uncommon sight on American roadways. When a driver suddenly forces their way into a lane, or hits the brakes to make a sudden lane change, it can start a wave. With little reaction room or warning, each imprefect driver is forced to overcorrect, which creates SOME additional space between each car, but at the expense of reducing the speed for every link in the chain.
Consider this:
Car 1 slows down by (A) Car 2 (tailgaiting) slows down by (A + B), creating a little extra space to avoid an accident with Car 1. Car 3 (tailgaiting) slows down by (A + 2B), creating a little extra space to avoid an accident with Car 2.
You can see where this is going. Since the wave creation happens one motorist at a time, all it takes is one savvy person to break the chain with a sizeable buffer and constant speed. But you'd be surprised just how rare individuals with this understanding are, and further, aggressive drivers make life painful for these people by merging into buffer spaces.
Given a smoothly flowing, full capacity freeway (as in everyone's going at 70mph and there's a max of 3-car spacing)... I've always wondered how it's possible for a slowdown "shockwave" to ripple back and cause traffic to come to a complete stop. I mean, if the traffic in front doesn't stop, why should the traffic at the back do?! Does anyone know?
The ripple effect is a magnification of a small change in speed form the lead car. The small change gets magnified because (1) the entire line of cars cannot see the lead car and (2) people are not perfect machines. People typically ride other driver's asses, and then stomp the brakes whenever they see the lights go off in front of them. Thus, the result of a "small" correction from a driver in-front propogates back and becomes larger with every car in the chain.
Think about it: if you are driving at 65 mph with cars packed tight, then if the lead car brakes even slightly, every car in the link needs to OVERCORRECT to avoid an accident (say 1 mph correction) with the car in front of them. The second car will be going 1 mph slower than the lead, and the third card will be going 2 mph slower than the lead, etc. Given this, then all you need is 65 cars to guarantee the car at the rear comes to a complete stop.
Once you get beyond a certain maximum capacity, it simply happens on its own, because the cars are too close to each other. If humans were machines, and could stop simultaneously, then cars could safely drive with no space in-between, and this wouldn't happen.
When was the last time you bought a microwave, however? Probably the last time your old microwave broke. There's no reason to upgrade your microwave until it stops microwaving- but that's just not true of computer hardware and never has been. If computers stop failing, and stop becoming faster because everyone wants them cheap and cheap means the same in mass quantity, which is basically what commodities are, it will slow the adoption of equipment which by its nature is only undergoing radical, rapid generational change because people keep buying it.
And yet, it is still possible to purchase microwaves like this crappy $50 number, but there is still a sizeable enough market for high-end microwaves to produce innovative products like this one.
I disagree with your premise entirely, because it is wrong. Just like high-end PCs and microwaves, there is a high-end, cutting-edge market for almost every piece of technology. Sony is only scared because the commodization of computers means the high-end market will probably stop growing, and possibly shrink...but it's not going away, no matter how loudly Sony whines.
Slashdot Mirror
Yeah, you know that the new discussion system is totally broken on IE6. Of course, I knew this six months ago when I elected not to test it, and since then they have fixed nothing.
What's with the duplo-block-sized titles, do we suddenly have armies of babies and old people reading the site?
And to stay on-topic: my stepfather was working for the census while they considered this transition, and it was the most painful decision they had to make in all his years working there. Digitizing something as flexible as paper meant that you actually HURT efficiency of data collection. Think about it: with paper, you can easily correct mistakes, skip questions (or go in a different order). Most important: with the computer, you're SOL if you drop the computer or the battery dies, or the software crashes.
And while digital data collection reduces costs at the back-end (the data is already digitized), the fact is that collecting the data is the most expensive part of the census process, and any increase in costs there erased the gains at the back-end.
On an eight bit display, each pixel is made of three sub pixels, each with one of 256 possible values. These displays only have 766 different colours (3 x 256 = 768, but that counts the colour black three times, so there are only 766 colours). The 16.7 million colours are just created by dithering.
You're not using the data properly. The three colors of an LCD pixel are blended by our eyes to create 2^24 distinctive shades, the same way your eye blends the red. green and blue phosphors of a CRT pixel. For your information, we include three different "blacks" because they are three different colors:
Say you set R to "black" and G and B to other values. You now have zero Red component; this is how you remove colors. If you make all three components black, then you have true black.
By having 6 bits of range versus 8 bits, this means that you will lose FOUR DISTINCT colors for each single color you can create in the 8-bit range.
Say the numbers 0 - 255 (8-bit) map to a color range. Now say you map the numbers 0 - 63 (6-bit) over that same range. That means 63 and 255 represent the same color, and 62 and 251 represent the same color...but what about the colors 252, 253 and 254? These are all represented by one value, now 63, unless you do some dithering.
Dithering methods: typically, to "recreate" more colors, 6-bit displays will cycle between the two "closest" colors to create a color outside it's rendering range. Thus, in order for our 6-bit display to create the color value "253," it alternates between displaying the color "63" (maps to 255) and the color "62" (maps to 251). You use similar methods to create the 8-bit colors "254" and "252", using different patterns.
What you end up with is a marginal image, because contantly changing pixels means that the image is blurry and loses contrast. This is why most 6-bit displays look like crap in darker scenes. You also don't end up with quite the color fidelity of a true 8-bit display, because it's just not possible to create, even with good dithering techniques.
One of the problems you have is that gate volume is approaching thousands of atoms. This is a problem because certain regions need to be doped in order to make the silicon do it's job.
What is the problem, you may ask? Well, just look at the Wikipedia entry you linked. Even doped silicon is still %99.999999 pure.
So, you have your gate that is thousands of atoms in volume, and dopant concentration that is in the 1 million to 1 billion ratio...so what is the likelyhood that your gate is going to contain that one dopant atom you need in the lattice for best performance? Without that dopant, the performance of your gate suffers, and may not even work at all.
I didn't understand who has a population density 5 to 10 times larger than Brazil
You do. So does the UK, and Germany. Specifically, 10 times that of Brazil.
Italy: 197 people per square km, Gini = 36
UK: 246 people per square km, Gini = 34
Germany: 230 people per square km, Gini = 28.3
United States: 31 people per square km, Gini = 47.0
Brazil: 22 people per square km, Gini = 56.6
I understand that poverty can often sadly lead to crime.
And just to give you an idea of how poor the poorest people in Brazil are, I've provided Gini numbers for each country. These numbers are a measure between the poorest and richest residents of a country, where higher numbers indicate a larger divide. You will notice that in Europe, where the poor are not all that poor and the population density is high, crime is not a problem. In the US, things get worse, but are not completely out-of-hand. In Brazil, the situation is a nightmare compared to Europe.
About Germany I think people should always remember which horrible things the nazism did but should stop blaming german people that actually live in Germany and have nothing to do with it, because let's face it now it's history and most of the people of that historic period are dead or very old now.
The rest of the world stopped blaming the Germans years ago, including myself. The only people blaming the Germans are the Germans themselves. I mean, really, what other free democracy in the world makes it illegal to form a group or have a webpage promoting hate speech or more specifically, neo-Nazis? They're completely stuck in the past.
So tell me why all this doesn't happen in Europe...
Because, like rail mass-transit, you get more police effectiveness for your dollar when you have a population density FIVE TO TEN TIMES LARGER than Brazil. Also, the people of Brazil aren't fat and happy like the west, because a good chunk of them live in huts below the poverty line, so you get worse gang violence than you'll ever see in the first-world. Police can only do so much when the population is restless and spread-out.
And let me make one thing clear: ultra-strict gun regulation is not a common theme in Europe. This is mostly restricted to the UK (nanny police state, complete with closed-circuit TV to watch the fun), and Germany (yes, the entire country is still embarassed about Hitler, even today). In France, Switzerand, Finland, you name it, gun ownership is high; a lot of EU countries even allow handguns for home defense/target practice use.
The Super NES had a 16-bit CPU with ~65,000 colors. But PC gamers were already upgraded to 16 million colors.
First: SNES only has a 32K color palette (5 bits RGB = 15 bits). VGA has a 262K color palette (6 bits RGB = 18 bits), and SVGA has a palette of millions of colors, as you stated.
Second: the number of on-screen colors is not the same as the palette. The number of colors on-screen was the same for VGA and the SNES, which was 256, and these could be selected from the palette. Most SVGA games also stuck to 8-bit color, because it was more efficient.
Where the PC had the advantage was higher resolutions, with early SVGA games pushing 640x480, and later games offering 1024x768 or higher. Higher resolutions easily made up for the low number of colors on-screen, because this allowed game designers to use dithering.
The console is typically one generation (4-5 years) behind the PC gaming world, becuase using older technology helps keep console prices low (~$300).
I'd disagree with your definition for a PC "generation." In the PC world, things move fast. The N64, for example, had parity with midrange PCs when it was released, mostly because there were no affordable 3D graphics accelerators for PC at the time. Within a year, PC graphics eclipsed the best the N64 could muster. Another example: the Xbox featured GeForce3 graphics, which were only a year old at the time of the system's release.
So sure, if you decide to slap the "PC" label onto everything, then yeah, the PC market is doing fine. Meanwhile, I don't think nVidia is going to have a strong season selling top-end video cards to only the people who bought Crysis
Nvidia doesn't make any money selling these high-end cards. These high-end cards exist for only one reason: inexpensive viral advertising. If your top-end card gets the most 3dmarks, it tends to affect buying decisions for purchases across the board.
The viral advertising works two ways: first, you get all these wonderful reviews that remind you that Nvidia exists; the reviewers show you incredible numbers, then as an aside tell you that the "real deals" (i.e. 9600GT, 8800GT) can be found for much less. Then you get the second effect, where hardcore enthusiasts read these articles and then preach the gospel to the masses via forums/blogs.
I mean it when I say that Nvidia doesn't make any money on ultra high-end; the largest segment for their earnings comes from the $200-250 segment, which has been populated by the venerable 8800 GT for the last six months. The 8800GT alone literally stole %5 of the desktop graphics market from AMD last quarter, even up-against the competitively-priced 3870.
PC graphics cards are only expensive if you insist on top-end. Example: you can get more graphics power than RSX in the PS3 for a paltry $100 today, and by next year the same power will be available in an entry-level $50 card. The graphics core in the PS3 used to be sold as the 7800 GTX, and sold for $500 only three years ago; now you can get MORE power in the $100 8600 GTS.
A better comparison would be the Sega Master System to the NES. The SMS was far superior hardware-wise, yet still barely made a dent in the marketplace.
Far superior? Not really. It just bugs me that this gets passed around all the time.
Same general CPU power (the Z80 is clocked twice as fast, but has more wait states per op than the 6502).
Same resolution (256 x 224).
Same maximum number of sprites (64, but the NES has flicker if you have > 8 on one scanline).
The sound on the Japanese Mark IV was better, but the first-generation FM synth didn't sound all that impressive. Really, the next-generation of FM synth in the Genesis was far better.
In the USA, they dropped the FM synth, so the NES had far better sound (with an extra PSG channel), and a MUCH stronger noise channel (anyone who has actually heard a real Master System knows the noise channel sounded weak and pathetic).
For graphics, yes, the Master System featured 32 colors onscreen without tricks, whereas the NES could only muster 25. However, most games at the time could make little use of all those fancy colors because they had pathetically small storage space. In the end, very few games took advangate of the impressive 32-color limit; we remember those few games as standouts, but the fact that there were so few of these games is one reason the console sold poorly.
The reason the SMS sold poorly in the US was because it was late to market compared to the NES, and didn't really have solid advantages. I only knew one person who bought an SMS while growing up, and he only bought it because his brother (and I) already had an NES.
We did the same thing when DVDs were first being ripped. We didn't have cheap DVDs, but we did have cheap CD-Rs, and lots of free time. With the pioneering Divx codec, attempting to put a full movie on a 1-2 CD-Rs wasn't all that crazy. By the time DVDs were cheap, you could actually create a fairly good copy of the movie on a single CD, or a pristine copy on two CDs.
Today, we have single-layer DVDs that are dirt-cheap, and dual-layer DVDs that aren't all that expensive. The single-layer disc is more than enough to store a 720p h.264 rip with reasonable quality, and the dual-layer disc can hold a 1080p rip using h.264 with only marginal quality losses.
The human eye's 'refresh rate' is around 60 Hz... If you think you're being distracted by flicker from 100Hz, you're only fooling yourself.
No, I'm not fooling myself, you just can't see what I see. You are the one fooling yourself; if you can't see it, it does not exist.
Let me give you an example: I run my CRT at 85 Hz. I can tell the difference between 60 Hz (tons of flicker), 75 Hz (marginal flicker) and 85 Hz (no noticeable flicker, except on pure white screen). Why do modern CRTs have flicker problems at the same refresh rates televisions have used for years? Because they have lower-persistence phosphors, which makes the difference between a TV and monitor at 60 Hz visible to a subset of all users (i.e. me, and the other complainers).
Why are LEDs a problem? Because their persistence is even lower than CRT phosphors. Also, the LED taillight is an array of PIERCING-BRIGHT point-source lights at a single wavelength. Bright red lights have always bothered me, but red LED light arrays and red lasers are even more annoying. Combined with the flicker, it can give me a headache.
What realy pisses me off is that large-scale automakers like Honda are starting to put the damn things on every car.
Why don't you do yourself and your brother a huge favor and buy him a couple 1GB sticks for his birthday? Unless you live in the Amazon or some such technical wasteland, memory is dirt-cheap these days.
Than you don't have to ask tough questio s like this; instead, you can just install the service pack and get on with your life.
not likely, being a new architecture and all.
But you see, Core and Core2 are two completely different architectures. This would make Core3 a fitting name for the next big thing.
Core is simply a Pentium M with a streamlined FP unit. The SSE units are still 64-bit and there are still only three instruction decoders.
Core2 adds the two 128-bit SSE registers, adds a fourth decoder, and a mess of other optimizations. This is certainly an architectural change.
NOT TRUE, PLEASE MOD DOWN PARENT.
DYNAMIC POWER = FREQUENCY * CAPACITIVE LOAD * VOLTAGE^2
The above ignores leakage, but as another poster mentioned, that is not related to frequency. Leakage actually scales LINEARLY with the device voltage.
Adding more cores DOES increase power linearly. but the frequency^3 comment is completely off-base. The worst offender is actually voltage, which adds quadratic dynamic power and linear leakage power. As you raise the frequency, power consumption can increase even more than just linearly, because you need higher voltages to drive higher frequencies.
Example: say you had a 2.0 GHz, 1.0v processor, which used 20w dynamic power. Now, let's say you want double the performance (4.0 GHz), but that requires 1.2v to drive it. This would give you:
20w * (4.0 GHz / 2.0 GHz) * (1.2v^2 / 1.0v^2) = 57.6w. You increased performance 2x, but had to increase dynamic power by almost 3x!
Thus, the designers of processors usually target a sweet-spot with a good combination of low voltage and marginal frequency. Adding more cores allows them to use the same parameters for each core, which yields (potentially) a 1 to 1 performance to power increase.
Gosh, I hope it is more effective, because in my implementations I actually saw a slowdown instead of an advantage. Even then I'm generally not happy with hyper-threading. The OS & Applications simply don't see the difference between two real cores and a hyperthreading core. If I run another thread on a hyperthreading core, I'll slowdown the other thread. This might not always be what you want to see happening. IMHO, the advantage should be over 10/20% for a desktop processor to even consider hyperthreading, and even then I want back that BIOS option so that disables hyperthreading again.
SMT, or Hyperthreading, is intended to reduce the overhead of task switching. Instead of clearing out the pipeline and and switching tasks at the OS level, the processor automatically detects a wait state on TASK_A and switches to TASK_B in a single clock cycle. There is no magical performance improvement, the goal is to reduce the overhead of task switching.
This was intended to hide the horrible I/O latency of the Pentium 4, and it did a decent job of it (on the server). On the desktop, it caused slowdown because HyperThreading added processing overhead, and most applications were not multithreaded OR I/O limited. If you didn't have multi-threaded code, you actualy got lower performace due to the overhead.
Well, now Intel sees two things happening in the market: one, Sun is making inroads with their Niagra architecture, using simple cores with SMT to produce high performnace on I/O-bound tasks. Intel sees this as an opportunity to reintroduce SMT to compete with Sun in the low-power, high-I/O market. On the desktop side, games are using more resources than ever (finally I/O bound), and some games can take advantage of as many as 4 threads. Having zero switching time between those I/O-hungry game threads could be impressive.
No, it certainly won't be all that impressive for the desktop, but at the same time it won't be the performance detriment it once was. The server market will LOVE it, and that's what really pleases Intel.
Can count all the major improvements - that you know/can think of, sure you've demonstrated that. What about all the improvements that occur outside of the processor instruction set? Such as improvements to branch prediction
;-)
The last two times branch prediction on x86 saw a HUGE improvement were with chips like the Pentium (which introduced a 2-entry branch history table), and the P6 and the K6, which introduced much more powerful branch history implementations. Everything beyod the P6/K6 generation has been minor tweaks; when you get above the most basic implementation of the branch history table, your performance improvements are tiny for the general-use case. Even additions like the loop predictor on the Pentium M don't add much, when you realize that most loops have lots of iterations.
Improvements to the pipeline so that branch prediction misses aren't so expensive
This would be great, except that most CPU pipelines ELONGATE as you add more features. Even the saga of the P6 -> Core2 has added more pipeline stages. And when you add more stages, you need better prediction, but there is only so much you can do: the Pentium 4 featured one of the most advanced branch predictors ever (but still only slightly better than previous), plus hyperthreading, all in an attempt to reign-in the cost of clearing the pipeline from branch misses (and other I/O). It didn't work.
(as well as additional instructions such as condition move that can reduce need for branching)?
Does not exist. You can execute both branches (at the expense of memory overhead) like Itanium does, or you can make a prediction (like all modern x86 processors), but only the programmer can actually remove branches.
Improved out-of-order instruction execution,
Which, as I highlighted, hasn't really improved in years. The Athlon took one step above the P6/K6 by adding a third floating-point unit, but kept the 3-wide decoders. The Pentium 4 added an ungodly number of rename registers, but that's mostly to handle the long pipeline and HyperThreading, and it didn't really improve performance per clock (see above).
The Core2 and the Phenom are the the first processors to actually improve out-of-order execution in years, with a combination of wider decode/retire path (Core2), support for packed SSE instructions (both), two 128-bit SSE units (both), and support for out-of-order loads/stores (both). The wider SSE registers double the potential performance for SSE2+ code, and the 4-wide decode path on the Core2 allows you to keep the integer pipes full while the SSE pipes work their magic.
fast loading/saving of contexts useful for multitasking, and of course, hyperthreading (okay just kiddin on the last one
I grouped these together, because they are essentially the same. Hyperthreading allows fast task switching without involving the operating system, because it makes one processor behave like two.
Don't kid, HyperThreading isn't a bad idea, it just had no place in the (largely) single-threaded univese of the PC. Intel added it to bandage the P4's poor I/O and branching peformance, but it wasn't enough to make the difference (plus it slowed-down single-threaded apps). On I/O limited processors (like Sun's Niagra line), SMT is a wonderful thing.
improvements to the MMU such as global and large pages mean less TLB misses, memory prefetching, greater memory bandwidth,
Large pages, memory prefetch, larger caches - they all add their little bits, but nothing extrodinary. On the Pentium 4, for example, the move from Willamette to Northwood (double the cache, plus prefetch) yielded a %5-7 performance improvement, and nobody knows whether the cache or the prefetch were more important. Those are small potatos.
And as for memory bandwidth, it is not a question: it will improve, or processors will starve. But the amount of input from the processor side is startlingly low: only four times since the introduction of the 8088 have processor manufacturers widended the memory bus (16-bit 286, 32-bit 386, 64-bit P5, 128-bit P4/Athlon), and only with the 486 did we actually introduce on-chip cache.
Clock frequency is not an indicative of CPU performance. For example, the Core 2 chips, despite generally operating at a lower frequency than the Pentium 4's outperform them significantly.
But massive instruction per clock improvements do not happen very often in the x86 chip industry. In fact, I can count all the major improvements for the last 15 years on one hand:
1993: Intel Pentium Pro (approximately 2 INT, 2 FP operations per clock, best case) introduces real time instruction rescheduling to the x86 world. The design can decode 3 instructions per clock. Yes, I am disregarding the Pentum, because you got NO performance improvement without an optimizing compiler.
1997: MMX increases maximum number of integer instructions to 8 per cycle. But, because of the 64-bit data size, you really see little improvement unless using 16-bit or 8-bit types.
1998/1999: 3DNOW! and SSE double the potential throughput for 32-bit floating point, again not all that impressive.
2001: the Pentium 4 actually REDUCES performance per clock, with a single instruction decoder, and heavy reliance on trace cache to make up for this. SSE2 gives the potential to increase FP thoroughput to 4 instructions per clock, per SSE unit, but a half-assed implementation by both Intel and AMD means nothing changes.
2006/2007: the addition of more decode units on the Core 2, packed SSE instructions for both the Core 2 and the Phenom, and TWO 128-bit SIMD units means we see the first improvements in instructions per clock in years.
Also the interface for uploading songs was conected to the _parallel_ (LPT) port of the comp. It was pretty unstable. The filesystem was also not FAT12/16/32 based so it was a hassle to get the songs on the player a few yeras after when it was hard coming by Win98 (for which the software was written).
This is the major reason why I held-off buying an mp3 player in the dark ages of portable players. The idea of spending hundreds on a device with a proprietary dongle cable that used a slow bus (parallel or serial), and depended on some random music transfer software did not appeal to me.
Creative Labs was the first manufacturer to get my money because they released the Muvo, a 128MB wonder. Not only did the Muvo use the USB interface for faster transfer speeds, it required no dongle because the USB port was built-in to the player. The battery case detached from the player, revealing a full-sized USB connector. This ability, combined with the UMS disk mode, made it easy to transfer files, and made it feasable to use as a thumbdrive.
It goes one step further than this: deletion trolls today can take advantage of several tools that Wikipedia has made available to them, including listings of all newly-created pages, and changes to pages. This makes deletion easy, whereas finding older pages that are not up to standards is a little more difficult.
Moreover: there are so many scripted and "in-the-know" things on Wikipedia today, I can't make a "standards-compliant" page to save my life. For example: how the fuck do you put in a standard references block, and refer to those references in-article using superscript links? I've looked at several other pages to figure out how this works, but the code seems to be hidden. All the strange codes you can use to mark pages, like say "this article needs expansion," are not really easy to find. Compared to the ease of creating my first Wikipedia page, creating a page today that will susvive the hounds is near impossible.
I guess it's not surprising that I stopped contributing to Wikipedia recently.
One small problem, AMD's contract with Intel states that they can't outsource more than 20% of their chip production.
I've seen this line dropped a million times on Slashdot, and I have NEVER seen anyone back it up with proof. This is turning into the ultimate geek internet rumor, because everyone parrots it verbatim without checking the facts.
I say, put up a link, or shut the hell up. Unless you have a news article or a copy of AMD's contracts, you have proof of nothing.
The fact is, you don't even need a CPU license to make an IA-compatible chip (just look at the Crusoe). Even NexGen, Centaur and Cyrix made chips for years without licenses, although recently they have all obtained cross-licensing agreements (out-of-court settlements)
The license simply makes doing business easier because cross-licensing means Intel won't litigate you to death. So, even if your post is true, AMD is a particularly strong opponent, and any litigation by Intel will raise the monopoly flags all over. If this actually happens, the eventual result will be another cross-licensing agreement, except with AMD having much more leverage than before (with Intel under the monopoly pressure).
Although Via is not overpriced any longer, they used to charge hundreds of dollars for their Mini-ITX boards, and thus they keep their hard-won repuation. Believe me, those skinflints deserve it.
They used to charge OVER $300 just for passive-cooled Mini-ITX boards with processors slower than 1GHz. That's not competitively priced with ANY other platform. Just as an example, the (now $60) C7 1.5GHz board you linked went for over $200 on introduction last year! And the worst part: the prices never went down, not even with new product introductions, because Via was the only game in town.
So, what has changed?
Intel recently offered the D201GLY Mini-ITX with integrated Celeron for 80 bucks, and overnight the overpriced Via Mini-ITX market crashed.
Yes, Intel delivered more performance, at about the same power envelope, for less than half what Via was charging. Overnight, Intel changed a market that Via had literally let stagnate for the last 5 years: Via had product delays, overhyped improvements, and overpriced products, while Intel simply made a board that delivered all-of-the-above. Now THAT is a market force that can move mountains!
And now Intel wants to push power consumption and price even lower than Via ever dared, with performance that will probably be competitive with the C7. I say, bring it on!
By the looks of things Isaiah will wipe the floor with Atom if intel doesnt bury Via with branding power. Isaiah's out of order execution will offer much better performance than Atom's in-order execution. ...and more power consumption, from the looks of it. VIA estimates the 65nm Isaiah will have the same power envelope as with their 90nm C7 (10-20w), which is not unreasonable. However, this is an order of magnitude higher than the Atom, which is sipping power at 0.5-3w. The only thing VIA has that could compete with this power-wise would be a 65nm C7, and I have no doubt Intel can design an in-order chip that can outperform the crappy C7.
Next, please.
That's an interesting idea, but it has some complications.
:)
Two things:
I don't think we could stretch a Wifi network nintety miles without violating FCC rules. Twenty miles with a high-gain antenna is pushing it. You could, of course, solve this by getting an FCC exception (call it "Project Free Cuba," they'll eat it up).
Nintey miles means you have line-of-sight issues, and high frequencies like Wifi don't tend to duct/refract as much as low frequencies. You'll need a tower to solve this problem.
A nice alternative would be a boat anchored off the shore of Cuba in international waters, which lets you get around the whole FCC thing and use a much shorter tower. All you need is a satellite uplink to make the prefect relay. Unfortunately, the running cost of a ship relay is a lot more than the cost of a land-based relay, plus you'd be a nice fat target for Cuba's airforce
But whenever someone brakes the 3-car-space decreases (I assume) to a 2-car or 1-car space. Shouldn't this extra space compensate for the slowdown, so that the cars at the back don't have to overcorrect?
From your comments, I think you're assuming that drivers are competent, and keep three (or more) car-lengths between themselves and the next car. With competent drivers and plenty of reaction time buffer, you can avoid this problem.
To see how this problem can actually occur, I'm describing a worst-case scenario where you have cars riding bumper-to-bumper at high-speed. While this is obviously a tinderbox waiting to explode, it is not an uncommon sight on American roadways. When a driver suddenly forces their way into a lane, or hits the brakes to make a sudden lane change, it can start a wave. With little reaction room or warning, each imprefect driver is forced to overcorrect, which creates SOME additional space between each car, but at the expense of reducing the speed for every link in the chain.
Consider this:
Car 1 slows down by (A)
Car 2 (tailgaiting) slows down by (A + B), creating a little extra space to avoid an accident with Car 1.
Car 3 (tailgaiting) slows down by (A + 2B), creating a little extra space to avoid an accident with Car 2.
You can see where this is going. Since the wave creation happens one motorist at a time, all it takes is one savvy person to break the chain with a sizeable buffer and constant speed. But you'd be surprised just how rare individuals with this understanding are, and further, aggressive drivers make life painful for these people by merging into buffer spaces.
Given a smoothly flowing, full capacity freeway (as in everyone's going at 70mph and there's a max of 3-car spacing)... I've always wondered how it's possible for a slowdown "shockwave" to ripple back and cause traffic to come to a complete stop. I mean, if the traffic in front doesn't stop, why should the traffic at the back do?! Does anyone know?
The ripple effect is a magnification of a small change in speed form the lead car. The small change gets magnified because (1) the entire line of cars cannot see the lead car and (2) people are not perfect machines. People typically ride other driver's asses, and then stomp the brakes whenever they see the lights go off in front of them. Thus, the result of a "small" correction from a driver in-front propogates back and becomes larger with every car in the chain.
Think about it: if you are driving at 65 mph with cars packed tight, then if the lead car brakes even slightly, every car in the link needs to OVERCORRECT to avoid an accident (say 1 mph correction) with the car in front of them. The second car will be going 1 mph slower than the lead, and the third card will be going 2 mph slower than the lead, etc. Given this, then all you need is 65 cars to guarantee the car at the rear comes to a complete stop.
Once you get beyond a certain maximum capacity, it simply happens on its own, because the cars are too close to each other. If humans were machines, and could stop simultaneously, then cars could safely drive with no space in-between, and this wouldn't happen.
When was the last time you bought a microwave, however? Probably the last time your old microwave broke. There's no reason to upgrade your microwave until it stops microwaving- but that's just not true of computer hardware and never has been. If computers stop failing, and stop becoming faster because everyone wants them cheap and cheap means the same in mass quantity, which is basically what commodities are, it will slow the adoption of equipment which by its nature is only undergoing radical, rapid generational change because people keep buying it.
And yet, it is still possible to purchase microwaves like this crappy $50 number, but there is still a sizeable enough market for high-end microwaves to produce innovative products like this one.
I disagree with your premise entirely, because it is wrong. Just like high-end PCs and microwaves, there is a high-end, cutting-edge market for almost every piece of technology. Sony is only scared because the commodization of computers means the high-end market will probably stop growing, and possibly shrink...but it's not going away, no matter how loudly Sony whines.