See Sega v Accolade (re: Sega Genesis) or Paradise (Hercules?) v IBM (re: VGA clones). Also the inkjet refill wars.
Copyright is meant to protect creative works. If you use a string as a key, it becomes functional, not creative and is not eligible for copyright protection.
And since it appears you are decompressing and compressing to the same directory, and thus same volume, you'll likely be I/O bound anyway, specifically seek bound.
Multiple processors are nice, but are not nearly as good as a single faster core. Of course, if the price is right, I'll take it anyway.
They succeded better for two reasons: 1. The directors worked around the bad dialogue a bit better. 2. Alec Guiness. 3. By far the biggest, Harrison Ford. Without Harrison Ford there would have been no episode V, let alone VI,I,II,III. He made the character work, he made his dialogue work. He knew the character better than Lucus. He ad-libbed the "I know" response to Leia's "I Love You".
The other actors and their dialogue varied. Hamill was a great farm boy, a mediocre Jedi. Fisher was terrible all around. But Harrison Ford glued it together and made it work.
You can't go and say "if..." and say that the G5 has a performance advantage based upon a chip that doesn't exist. Part of the reason Intel has a performance advantage is that they have superior process technology and are more willing it use it on their CPUs.
And besides, what until you see what 65nm does for Intel...
Anyway, Intel has been faster than PowerPC since DAY ONE. When Apple announced 60/66/80 MHz PowerPC 601s (40MHz FSB tops), Intel already had had 66MHz Pentiums for a year. Two months later, Intel had 90 and 100 MHz Pentiums with a 66MHz FSB.
And beyond that, Pentium outperformed the 601 in everything but floating point. Check Microprocessor Reports' report on it. 601 was superscalar with the ability to (sometimes) execute an integer, floating point and a branch instruction all at once. Pentium had the ability to often execute two integer instructions at once plus a branch (sort of). Microprocessor Reports correctly showed that Pentium was able to execute multiple instructions per clock far more often than a PPC 601 could. And as noted above, it was also running at a higher clock rate.
Despite being CISC, Microprocessor Reports took the uncomfortable stance that the Pentium was a better chip than the vanguard of RISC's entry into the mass market.
PPC's next step was the miserable 604, which at 120 and 132MHz was choked by its still awful 40 and 44MHz FSBs. Pentium was already humming along along at 133MHz with a 66FSB. And it was only going to get worse. Pentium Pro had already been out for a few months, at 200MHz and an on-package 1:1 L2 cache.
Around this time, Apple released machines using the 603ev, which at 275MHz marked the last time PowerPC was faster (in MHz) than Intel's offerings. It still couldn't match up in real-world performance, as Pentium Pro and followons were capable of significant parallelism, and the 603ev did virtually none.
Much later, Apple released the PowerMac 9600/250, 300 and 350, with 50MHZ FSBs and a still slow 1:2 off-chip cache. By this time, Pentium II was at 400MHz with a 100MHz FSB.
Apple was in a deep hole now, because the entire 604 line was killed when its next chip, the overly complex 614 was killed. After some scrambling, the 613 took off. Apple named this the G3. It was a much simpler, but it was the best PPC had to offer, and helped Apple make up some of the huge gap in performance between them and Intel.
Apple's absolutely terrible memory latencies and bandwidths held them back through all the G3 and G4 days. This perhaps reached a peak when Apple released G4s that used DDR, despite the fact that the 133FSB on the G4 meant it couldn't get any more performance from DDR than from SDRAM. Meanwhile, over on Intel, Intel had 800MHz (effective) FSB processors, with dual-channel DDR RAM that provided 6.4GB/s theoretical memory bandwidth, compared to an Apple G4's 1.04GB/s. Yes, the gap in performance was as large as it sounds.
The next time Apple would even come close in performance was the early days of the G5, which with dual processors and a very fast and tricky bus architecture was capable of beating a single fast P4 in general performance at times. It also was capable of beating the P4 handily on performance, at least until SSE2 took off and the gap closed a bit.
And now the new frontier is notebooks. Intel's fastest Pentium-Ms are capable of providing 2/3rds of the performance of their fastest desktop processors. Apple's laptops meanwhile with their G4s are only capable of perhaps 1/3rd the performance of a single G5 chip.
Apple/PPC simply started out behind and never caught up. It's a wonder Apple was able to hold out this long.
File systems, starting with Apple's HFS in 1986 or so use database-type structures to store info about files. The directory a file is in is only a field in the database. So it doesn't change anything about the data structures if you store all the files in one directory or in many.
However, if you do try to iterate that directory it will take forever to do so. But in theory that isn't going to happen, as directories are no longer organizational strategies at that point.
Virtually none of the Mac OS X toolbox will work in a 64-bit app. So you can't make a real 64-bit app. You basically can make 64-bit command line tools, and that's it.
Apple's own docs say if you want to make a 64-bit app, you have to put all the 64-bitness in a separate task and message back and forth to it. Because of this, there aren't any 64-bit apps to speak of.
You do have to understand also that this is the guidelines for making a universal app. How universal would it be if it only ran on 64-bit processors? All current Intel Pentium-M offerings are 32-bit. The point of this announcement is to get a body of apps for x86 Mac OS X so that people can buy machines like that when they come out will have something to run.
The point isn't to encourage developers to make a lot of apps that will only run on a subset of these x86 machines.
That time might come later, there's nothing that says Apple can't release a 64-bit ABI as soon as they get their 32-bit house in order. It honestly would put this x86 platform then on up on the G5s, since as mentioned above, making a 64-bit app for the G5 Mac OS is very difficult.
I agree how this processor is unsuitable for a Mac.
But an in-order core is okay in a console, because all the memory latencies and pipeline latencies are fixed for the 5 year run of the console. If you get your compiler to get it right, it'll run like a top.
The in-order becomes a problem (as you alude to) when you try to make follow-on chips and machines that have different characteristics. This wouldn't happen in a console.
I remember how MIPS didn't have memory latency interlocks (in fact, it is was in their name!) in the R2000. But they quickly found out they had to add them in the R3000, when they sped up clock speeds and the latencies which had been staticly compiled into their code were not sufficient anymore.
You say they had the best battery life before Pentium M came out. And then you say they were very slow.
So knowing these two things, how do you make the leap to best MIPS/Watt? Your laptop would have to be some combination of faster and longer battery life to win. Yet you say it was very slow. Would a comparably slow Intel machine have as long battery life?
Intel's current offerings destroy that laptop in MIPS/Watt. Intel's P3 mobiles released right after the first Transmetas bested the Transmetas significantly. And the Pentium M obliterated it. And now the ultra low volt Pentium Ms?
You're greatly mistaken.
I do agree Transmeta perhaps lit a fire under Intel to make more power-efficient chips. But they ceased to be competitive on power-efficiency or MIPS a long time ago. I won't be sorry to see them go.
There's no way to have gone to E3 this year and have concluded otherwise.
Every PC game maker who did an FPS was showing it on Xbox. All the other PC game makers were making some kind of massive multiplayer online game.
If you subtract that stuff, there is only a tiny amount of PC game development still going on.
This is going to suck for originality, since dev kits for consoles cost a fortune. So small houses can't buy dev kits without publisher backing, and publishers won't greenlight original (risky) games.
But that's the way it is going.
I have a feeling Nintendo is going to jump on this and try to make their console the home to small-time developers with original games. They've said as much, but I think they'll back it up with a cheap development system too.
Remember back when E3 had one hall with a portion of console games? Then remember when it had the PC room and the console room? Now it's all the console room. Even Blizzard was concentrating on consoles (except for one MMO game).
It was an example, not a case study.
If you think that this isn't a problem, why are there multiple parallel +3.3V and +5V wires from your power supply to your motherboard? Over only 18 inches?
Whatever drop I have right now, if you extend the cable to 5 feet instead of 18 inches, and go from multiple parallel wires to a single wire for each voltage (you can't have a 1" bundle across the floor), you'll end up losing 15x as much power (assuming about 3x farther and 5x parallel wires right now) as you do with the power supply in the case.
And, as an added note, often your power supply is running a re-regulating switching power supply. These things have massive input currents, well north of 5A.
I'd love to see smaller supplies. But look at the current situation. My 6800 Ultra video card requires TWO hard drive power connectors into it. Why? The vast majority of power supplies have only one circuit of power for +12V each of and +5V (and virtually all when the card was made, as the number is rising). Yet the instructions say you need to put plugs from two different strands into the card.
Why? Because the drop in those power cables is too large for the currents they draw. They need parallel wiring to halve the drop. So what is going to happen if the power supply is now 3x farther away? I'll need 6 wires.
Big portions of the power supply must remain in the case right now. And likely forever.
Where you mount an antenna on your roof pointed at a particular location. You can get multiple megabits from this, and cheaper than from a cell-phone provider.
Here's an example. http://www.sonic.net/sales/broadband/broadlink/ind ex.shtml You can't use them unless you're in Santa Rosa, but there may be companies in your area. Basically go to dslreports.com and check for wireless DSL.
http://www.broadbandreports.com/search
UHD? I'm supposed to pay extra to watch shows in HD? They only air like 4 original shows a week.
Universal already has UPN, and UPN has HD. Heck, UPN just lost their front-runner Sci-Fi show.
It'd be perfect if this were simulcasted in HD on UPN so I could see it for free instead of getting soaked for more money.
Laptops actually only have a portion of the power supply outside the case. They convert to 28V, for example. Converting to a low voltage like 3.3V outside the case is inefficient, for the same reason that houses don't use 10V power.
If you have a device that needs 330W, it can take 12A at 28V or 100A at 3.3V.
Now, let's consider the resistance in the wires. If they are 0.01 ohm/foot, and you have 1 foot of wire, then you will lose 0.01*12 or 0.12V if you use 28V , or you will lose 0.01*100 or 1V at the 3.3V input. Now, at 28V, you lost 0.5% of your power. At 3.3V you lost 30%. And this is over one foot. If the wire is 3 foot, it gets far worse.
So it is smart to run high voltage as much of the way as possible, and the low voltage runs should be short. Again, this is the same to your house.
Laptops get away with this because they use less power, less power is less power lost. And as I said, they don't put the power supply outside the machine, only part of the power supply.
It might be smart to put the rectification (AC->DC conversion) outside the case, because it doesn't reduce the voltage and it does generate heat. But it really won't save you much.
On a final note, connectors have a voltage drop of their own, so it really isn't smart to put extra connectors in the low voltage outputs from the power supply. New power supplies are doing this, because it sells units, but it will cost you in the end in lost efficiency and lower reliability (connectors can break).
It works today, because you haven't seen it much before. IE's box "are you sure you want to install/download this?" used to work before, when it was new.
But it becomes part of the process after a while. You click yes automatically. It's just fatigue. You can't remain vigilant all the time.
Outsourcing makes these kind of measurements useless. IBM could contract out all their work, and thus they would have less than 5% the employees they have right now. And they would seem to be 20x more efficient, even though it wouldn't change anything.
Return on working capital is a better measure. Google will still destroy IBM there, due to being in markets with a lower cost of entry.
Slashdot Mirror
See Sega v Accolade (re: Sega Genesis) or Paradise (Hercules?) v IBM (re: VGA clones). Also the inkjet refill wars.
Copyright is meant to protect creative works. If you use a string as a key, it becomes functional, not creative and is not eligible for copyright protection.
Well, not unless I/O is free on your platform.
And since it appears you are decompressing and compressing to the same directory, and thus same volume, you'll likely be I/O bound anyway, specifically seek bound.
Multiple processors are nice, but are not nearly as good as a single faster core. Of course, if the price is right, I'll take it anyway.
Yep, Eps IV-VI had bad acting and dialogue.
They succeded better for two reasons:
1. The directors worked around the bad dialogue a bit better.
2. Alec Guiness.
3. By far the biggest, Harrison Ford. Without Harrison Ford there would have been no episode V, let alone VI,I,II,III. He made the character work, he made his dialogue work. He knew the character better than Lucus. He ad-libbed the "I know" response to Leia's "I Love You".
The other actors and their dialogue varied. Hamill was a great farm boy, a mediocre Jedi. Fisher was terrible all around. But Harrison Ford glued it together and made it work.
...It also was capable of beating the P4 handily on FLOATING-POINT performance, at least until SSE2 took off and the gap closed a bit...
Sorry about that.
This is just speculation, and where he isn't already provably wrong, there's no particular reason to thing he's right either.
You can't go and say "if..." and say that the G5 has a performance advantage based upon a chip that doesn't exist. Part of the reason Intel has a performance advantage is that they have superior process technology and are more willing it use it on their CPUs.
And besides, what until you see what 65nm does for Intel...
Anyway, Intel has been faster than PowerPC since DAY ONE. When Apple announced 60/66/80 MHz PowerPC 601s (40MHz FSB tops), Intel already had had 66MHz Pentiums for a year. Two months later, Intel had 90 and 100 MHz Pentiums with a 66MHz FSB.
And beyond that, Pentium outperformed the 601 in everything but floating point. Check Microprocessor Reports' report on it. 601 was superscalar with the ability to (sometimes) execute an integer, floating point and a branch instruction all at once. Pentium had the ability to often execute two integer instructions at once plus a branch (sort of). Microprocessor Reports correctly showed that Pentium was able to execute multiple instructions per clock far more often than a PPC 601 could. And as noted above, it was also running at a higher clock rate.
Despite being CISC, Microprocessor Reports took the uncomfortable stance that the Pentium was a better chip than the vanguard of RISC's entry into the mass market.
PPC's next step was the miserable 604, which at 120 and 132MHz was choked by its still awful 40 and 44MHz FSBs. Pentium was already humming along along at 133MHz with a 66FSB. And it was only going to get worse. Pentium Pro had already been out for a few months, at 200MHz and an on-package 1:1 L2 cache.
Around this time, Apple released machines using the 603ev, which at 275MHz marked the last time PowerPC was faster (in MHz) than Intel's offerings. It still couldn't match up in real-world performance, as Pentium Pro and followons were capable of significant parallelism, and the 603ev did virtually none.
Much later, Apple released the PowerMac 9600/250, 300 and 350, with 50MHZ FSBs and a still slow 1:2 off-chip cache. By this time, Pentium II was at 400MHz with a 100MHz FSB.
Apple was in a deep hole now, because the entire 604 line was killed when its next chip, the overly complex 614 was killed. After some scrambling, the 613 took off. Apple named this the G3. It was a much simpler, but it was the best PPC had to offer, and helped Apple make up some of the huge gap in performance between them and Intel.
Apple's absolutely terrible memory latencies and bandwidths held them back through all the G3 and G4 days. This perhaps reached a peak when Apple released G4s that used DDR, despite the fact that the 133FSB on the G4 meant it couldn't get any more performance from DDR than from SDRAM. Meanwhile, over on Intel, Intel had 800MHz (effective) FSB processors, with dual-channel DDR RAM that provided 6.4GB/s theoretical memory bandwidth, compared to an Apple G4's 1.04GB/s. Yes, the gap in performance was as large as it sounds.
The next time Apple would even come close in performance was the early days of the G5, which with dual processors and a very fast and tricky bus architecture was capable of beating a single fast P4 in general performance at times. It also was capable of beating the P4 handily on performance, at least until SSE2 took off and the gap closed a bit.
And now the new frontier is notebooks. Intel's fastest Pentium-Ms are capable of providing 2/3rds of the performance of their fastest desktop processors. Apple's laptops meanwhile with their G4s are only capable of perhaps 1/3rd the performance of a single G5 chip.
Apple/PPC simply started out behind and never caught up. It's a wonder Apple was able to hold out this long.
File systems, starting with Apple's HFS in 1986 or so use database-type structures to store info about files. The directory a file is in is only a field in the database. So it doesn't change anything about the data structures if you store all the files in one directory or in many.
However, if you do try to iterate that directory it will take forever to do so. But in theory that isn't going to happen, as directories are no longer organizational strategies at that point.
HFS stores all file data in B-trees.
Virtually none of the Mac OS X toolbox will work in a 64-bit app. So you can't make a real 64-bit app. You basically can make 64-bit command line tools, and that's it.
Apple's own docs say if you want to make a 64-bit app, you have to put all the 64-bitness in a separate task and message back and forth to it. Because of this, there aren't any 64-bit apps to speak of.
You do have to understand also that this is the guidelines for making a universal app. How universal would it be if it only ran on 64-bit processors? All current Intel Pentium-M offerings are 32-bit. The point of this announcement is to get a body of apps for x86 Mac OS X so that people can buy machines like that when they come out will have something to run.
The point isn't to encourage developers to make a lot of apps that will only run on a subset of these x86 machines.
That time might come later, there's nothing that says Apple can't release a 64-bit ABI as soon as they get their 32-bit house in order. It honestly would put this x86 platform then on up on the G5s, since as mentioned above, making a 64-bit app for the G5 Mac OS is very difficult.
Wow, news for nerds? This isn't news, it's been known for months.
d -run-at-2-5-GHz-056.shtml
Just search for Intel Yonah. Jeez.
http://news.softpedia.com/news/Intel-s-Yonah-coul
I agree how this processor is unsuitable for a Mac.
But an in-order core is okay in a console, because all the memory latencies and pipeline latencies are fixed for the 5 year run of the console. If you get your compiler to get it right, it'll run like a top.
The in-order becomes a problem (as you alude to) when you try to make follow-on chips and machines that have different characteristics. This wouldn't happen in a console.
I remember how MIPS didn't have memory latency interlocks (in fact, it is was in their name!) in the R2000. But they quickly found out they had to add them in the R3000, when they sped up clock speeds and the latencies which had been staticly compiled into their code were not sufficient anymore.
You say they had the best battery life before Pentium M came out. And then you say they were very slow.
So knowing these two things, how do you make the leap to best MIPS/Watt? Your laptop would have to be some combination of faster and longer battery life to win. Yet you say it was very slow. Would a comparably slow Intel machine have as long battery life?
Intel's current offerings destroy that laptop in MIPS/Watt. Intel's P3 mobiles released right after the first Transmetas bested the Transmetas significantly. And the Pentium M obliterated it. And now the ultra low volt Pentium Ms?
You're greatly mistaken.
I do agree Transmeta perhaps lit a fire under Intel to make more power-efficient chips. But they ceased to be competitive on power-efficiency or MIPS a long time ago. I won't be sorry to see them go.
They don't. They couldn't even beat Intel on MIPS/Watt, and ARM has between 20 and 100 times the MIPS/Watt that Intel does.
There's no way Transmeta was the best in any technical measure.
There's no way to have gone to E3 this year and have concluded otherwise. Every PC game maker who did an FPS was showing it on Xbox. All the other PC game makers were making some kind of massive multiplayer online game. If you subtract that stuff, there is only a tiny amount of PC game development still going on. This is going to suck for originality, since dev kits for consoles cost a fortune. So small houses can't buy dev kits without publisher backing, and publishers won't greenlight original (risky) games. But that's the way it is going. I have a feeling Nintendo is going to jump on this and try to make their console the home to small-time developers with original games. They've said as much, but I think they'll back it up with a cheap development system too. Remember back when E3 had one hall with a portion of console games? Then remember when it had the PC room and the console room? Now it's all the console room. Even Blizzard was concentrating on consoles (except for one MMO game).
It was an example, not a case study. If you think that this isn't a problem, why are there multiple parallel +3.3V and +5V wires from your power supply to your motherboard? Over only 18 inches? Whatever drop I have right now, if you extend the cable to 5 feet instead of 18 inches, and go from multiple parallel wires to a single wire for each voltage (you can't have a 1" bundle across the floor), you'll end up losing 15x as much power (assuming about 3x farther and 5x parallel wires right now) as you do with the power supply in the case. And, as an added note, often your power supply is running a re-regulating switching power supply. These things have massive input currents, well north of 5A. I'd love to see smaller supplies. But look at the current situation. My 6800 Ultra video card requires TWO hard drive power connectors into it. Why? The vast majority of power supplies have only one circuit of power for +12V each of and +5V (and virtually all when the card was made, as the number is rising). Yet the instructions say you need to put plugs from two different strands into the card. Why? Because the drop in those power cables is too large for the currents they draw. They need parallel wiring to halve the drop. So what is going to happen if the power supply is now 3x farther away? I'll need 6 wires. Big portions of the power supply must remain in the case right now. And likely forever.
Where you mount an antenna on your roof pointed at a particular location. You can get multiple megabits from this, and cheaper than from a cell-phone provider. Here's an example. http://www.sonic.net/sales/broadband/broadlink/ind ex.shtml You can't use them unless you're in Santa Rosa, but there may be companies in your area. Basically go to dslreports.com and check for wireless DSL.
http://www.broadbandreports.com/search
UHD? I'm supposed to pay extra to watch shows in HD? They only air like 4 original shows a week. Universal already has UPN, and UPN has HD. Heck, UPN just lost their front-runner Sci-Fi show. It'd be perfect if this were simulcasted in HD on UPN so I could see it for free instead of getting soaked for more money.
Laptops actually only have a portion of the power supply outside the case. They convert to 28V, for example. Converting to a low voltage like 3.3V outside the case is inefficient, for the same reason that houses don't use 10V power. If you have a device that needs 330W, it can take 12A at 28V or 100A at 3.3V. Now, let's consider the resistance in the wires. If they are 0.01 ohm/foot, and you have 1 foot of wire, then you will lose 0.01*12 or 0.12V if you use 28V , or you will lose 0.01*100 or 1V at the 3.3V input. Now, at 28V, you lost 0.5% of your power. At 3.3V you lost 30%. And this is over one foot. If the wire is 3 foot, it gets far worse. So it is smart to run high voltage as much of the way as possible, and the low voltage runs should be short. Again, this is the same to your house. Laptops get away with this because they use less power, less power is less power lost. And as I said, they don't put the power supply outside the machine, only part of the power supply. It might be smart to put the rectification (AC->DC conversion) outside the case, because it doesn't reduce the voltage and it does generate heat. But it really won't save you much. On a final note, connectors have a voltage drop of their own, so it really isn't smart to put extra connectors in the low voltage outputs from the power supply. New power supplies are doing this, because it sells units, but it will cost you in the end in lost efficiency and lower reliability (connectors can break).
It works today, because you haven't seen it much before. IE's box "are you sure you want to install/download this?" used to work before, when it was new. But it becomes part of the process after a while. You click yes automatically. It's just fatigue. You can't remain vigilant all the time.
Outsourcing makes these kind of measurements useless. IBM could contract out all their work, and thus they would have less than 5% the employees they have right now. And they would seem to be 20x more efficient, even though it wouldn't change anything. Return on working capital is a better measure. Google will still destroy IBM there, due to being in markets with a lower cost of entry.