"If you handn't noticed that all of those things that break are binary only drivers. vmware, ati and nvidia all depend on a binary kernel module."
If that were true, I wouldn't have to have a Promise IDE card to be able to use my CD drive. It used to work... but now using the SATA and PATA ports at the same time is impossible with Linux.
Re:One thing I'm a bit confused about...
on
Kernel 2.6.12 Released
·
· Score: 2, Informative
Because development was going quickly and they didn't want to lose momentum. We're getting new features much sooner than we otherwise would have.
The downside is that 2.6 kernels are now a regression-fest that makes Windows look positively stable. They claim distros are able to stabalize their own kernels, which is a theory I have yet to see put into practice. The idea now is to find a kernel version that doesn't have any show-stopper regressions for your hardware.
"That's kind of neat, but I am having some difficulty in deciding what it's useful for. Perhaps teaching Unix to new users? It is, after all, a "fake" environment that looks like the real thing."
It would be very useful for web applications. One of the limitations in pushing functionality out to the client are the limitations in Javascript. A fully general purpose environment makes more things possible.
"More than likely, the lost profits from not reaching the <10 percent of the market willing to pay only for devices with public documentation are less than the lost profits from incurring the expense of cleaning up internal documents."
That's why we have to bitch and moan as loud as possible to make the public relations value exceed the cost of cleaning up the docs.;)
"except the mac rarely crashes, doesn't get viruses, never needs formatting (or rarely). Add all the time you spend dealing with those things to the cost of a PC and the difference shirnks."
I'm not going to dispute that OS X beats Windows as a desktop OS, but the hardware is a lot more expensive. In some cases, it's within reason. In other cases, like the single-CPU PowerMac, it's outrageously overpriced.
And buying a dual-CPU PowerMac doesn't help even if those aren't overpriced. If I didn't need a second CPU in the first place, overpricing everything but the dual-CPU computers is still a rip off.
"Really, the only reason they went with BSD is because they can keep their version closed (right?)."
The Darwin kernel is open.
"But once again, let me point out why: Because Apple wants to sell pretty, uncrackable, all-in-one, gold standard computers. They've been doing that since the 1970's and they will continue to do so. As we see with OSX and the ability to easily install Linux on a "Macintosh" that the software is just another layer."
I dunno. Most Linux users tend to like to customize the hardware on their computers. The only Macs that can accept substantial upgrades are PowerMacs, and these cost much more money than equivilant PCs. I don't see Apple changing their pricing, so people primarily interested in running Linux can save a lot of money doing it on other hardware. People peripherally interested in Linux will be more likely to use Macs.
" 20 years from now??? I so hope we are not still running nothing but Unix and Unix wannabes? Sure it "may" have a posix layer but I hope that some progress has been made by then. Maybe something like Plan 9."
The UNIX way of doing things might not be perfect, but it's good enough and more importantly very entrenched. There will be alternatives in 20 years just like there are alternatives now, but UNIX will still be with us.
Yes there's a huge speed advantage for Cell with some things, but for other things there's a big disadvantage. They might be capable of branching, but that doesn't mean they're good at it. You don't give someone 128 registers unless you want them to unroll their loops.
You have to rewrite code for Cell because it's a different environment (limited directly accessible memory, different libraries, etc). The fact that they're not PowerPC is a secondary issue, that's what compilers are for. Getting people to rewrite code for a different environment is like pulling teeth because it's really annoying.
Programmers are more expensive than hardware unless the advantage is huge. Games? Sure. Big CAD workstations? Why not? Scientific simulations? Sweet. Pixar's render farms? Absoloutely. Databases, dynamtic content, web browsers, fancy searches (eg spotlight), etc? Not. A. Chance.
x86 isn't sitting still. I don't know about Intel, but AMD's roadmap talks about co-processors on the die. They'd be x86 compatible, but optimized for different cases like vector-heavy code. If these are a fraction as fast as a Cell, but they work transparently with your existing code, you'll find them winning in the market for all but the most demanding applications.
Also, Cell isn't mutually exclusive with x86. There's no reason a Cell can't sit on a PCIe card much like we have dedicated 3D hardware, and I've no doubt this will happen for some things. The issue is that Cell is the part that you can sometimes do without, while a general purpose processor is almost always mandatory.
"Multi-threaded code is very difficult to write correctly and debug. It's hardly 'easy'."
Let's not get carried away.:)
When all you want is a sane alternative to non-blocking I/O, or an application that remains responsive when it's busy, no problem. Multi-threaded code is easy to write. Multi-threading CPU-bound code such that it's actually faster is sometimes impossible and usually hard.
"On the PC? Unscrew case, remove the HDD Cage as the memory bank just HAPPNES to be half way under it, make sure to unplug all cables because they are running all across the main board, then put in memory and reverse the whole thing. Takes me roughly 4x as long to upgrade the RAM in a standard PC than in the Mac."
Price of memory: > $60 Time to earn price of memory: > 1 hour (for me anyway, if you make more STFU) Time to upgrade PowerMac: 60 seconds. Value of time to upgrade PowerMac: < $1 Time to upgrade PC: 240 seconds (assuming you're correct) Value of time to upgrade PC: < $4 Difference: $3, 1/20th the cost of the memory
Total lifetime instances of open case: 50 (guessing high) Total additional cost: $150
Sometimes, specialization and generalization are mutually exclusive. Linux is a very powerful general purpose OS, but there's a lot of specializations that they can't have or don't have because their focus is elsewhere.
Fine with me. A monoculture would be a Bad Thing.
Anyone who ties themself to one OS is automatically at a disadvantage.
"On my home XP Pro box, freshly after a reboot, I currently have 15 distinct processes running, with FireFox as the only obviously user-interactive one."
Just because processes are present doesn't mean they're running. They're almost certainly not running. Most processes sit idle, taking up memory and waiting for something to happen. When it does, they take care of it in 5 milliseconds, and then they sit there again. Your CPU is idle for the vast majority of the time, and when you're using it these other processes don't get in your way very much.
Linux does a pretty reasonable job. I use XP at work and when it's doing something CPU-bound (generating a key pair with putty sticks out in my mind) the machine becomes unresponsive, but doing the same thing on my Linux machine doesn't have any perceptible effect. 2.4 kernels kinda sucked at that, but 2.6 classifies threads based on whether they use up all their CPU time. If they sleep voluntarily or wait for I/O, they are given higher priority.
Even if the CPU usage is at 100%, benchmarks have shown that interactive processess generally respond in under a millisecond. It's really impressive how a system can be under heavy load but you wouldn't even be able to tell if you couldn't see the network lights blinking like mad, hear the hard drive, and see the CPU temperature going up.
"What I would worry about is more on the chip perfromance side of things - namely the additonal capacitance loading, cross talk and the overall routing density for this approach."
Also the power consumption. I don't have a breakdown of where this technology would be used and where chips spend the most power (apart from knowing cache doesn't take much power), but it might hurt on laptops.
"If the Mathmatica CEO can get called on Wednesday night the week before, asked to bring the source code to Apple, and turn around a native Intel program in two hours of changes, then your developers don't need a year advanced warning. Right?"
That's an already-portable application without many architechture specific optimizations and they only had to get it to "demo quality".
The really big stuff like Photoshop is going to take longer. Even if they gave big vendors more warning, thousands of people would have known about it and it's too big a secret and too easy to leak anonymously to stay secret for long.
If Apple hadn't come clean, the rumors and FUD would have been much worse.
His theories are generally internally consistent but they often have little basis in reality. He also has trouble distinguishing what he would do from something that a company like Apple would do.
"I like the fact that both of my macs running together at full noise are quieter than my PC with one fan. This is because of the CPU. PowerPCs are cooler in both senses of the word."
Maybe you should check with Steve before you continue spouting propaganda. Apple wants cool and quiet computers, so they picked Intel.
Pentium 4s are rediculous, but as far as I can tell Apple will be using Pentium M derived chips for everything except those developer boxes they're shipping now. Pentium M blows Apple away in terms of battery life, they are easily the cooler chip.
"Nobody buys a Mac because it's got a cool processor - they but it because it's got a great interface that makes life easier for them."
Are you kidding? There's a lot of architechture bigots around here. And there's a few (not many, but a few) people that even have rational reasons for prefering PowerPC.
Even though it's completely transparent to most people and 99% of developers aren't programming in assembly, I can't count the number of times I've seen something to the effect of "I can't stand the thought of using such an ugly architechture.".
In the dark days before the G5 they had to convince themselves there was an advantage as they dropped further and further behind. I guess they can't leave that behind. A picture of Steve in front of that giant "performance per watt" slide isn't enough to convince some of them.
I'm thinking that it'll be possible to run MacOS X/x86 binaries on Linux with a Wine-like layer if you have the Apple userspace available. Then we can have iTunes, Office, even Photoshop running natively without having to run a full instance of OS X.
The obvious way to do this would be to buy Apple hardware and dual-boot with Linux, simply mounting the OS X partition and using the libraries from there. It might also be possible to extract the files from an install CD for use on cheaper and faster AMD hardware.
Before you flame me: MacOS X isn't fully compatible with Linux on PowerPC or x86, and not all software you might want can/has been ported. Also note that there are plenty of reasons for wanting to do this even if you yourself wouldn't want to.
Slashdot Mirror
"If you handn't noticed that all of those things that break are binary only drivers. vmware, ati and nvidia all depend on a binary kernel module."
If that were true, I wouldn't have to have a Promise IDE card to be able to use my CD drive. It used to work... but now using the SATA and PATA ports at the same time is impossible with Linux.
Because development was going quickly and they didn't want to lose momentum. We're getting new features much sooner than we otherwise would have.
The downside is that 2.6 kernels are now a regression-fest that makes Windows look positively stable. They claim distros are able to stabalize their own kernels, which is a theory I have yet to see put into practice. The idea now is to find a kernel version that doesn't have any show-stopper regressions for your hardware.
Indeed.
Everyone who hasn't read the Mars trilogy by Robinson, please leave the thread.
"OpenBSD users see Linux users as losers"
We do?
I guess I should stop using Debian on my other computer.
"That's kind of neat, but I am having some difficulty in deciding what it's useful for. Perhaps teaching Unix to new users? It is, after all, a "fake" environment that looks like the real thing."
It would be very useful for web applications. One of the limitations in pushing functionality out to the client are the limitations in Javascript. A fully general purpose environment makes more things possible.
"More than likely, the lost profits from not reaching the <10 percent of the market willing to pay only for devices with public documentation are less than the lost profits from incurring the expense of cleaning up internal documents."
;)
That's why we have to bitch and moan as loud as possible to make the public relations value exceed the cost of cleaning up the docs.
"except the mac rarely crashes, doesn't get viruses, never needs formatting (or rarely). Add all the time you spend dealing with those things to the cost of a PC and the difference shirnks."
I'm not going to dispute that OS X beats Windows as a desktop OS, but the hardware is a lot more expensive. In some cases, it's within reason. In other cases, like the single-CPU PowerMac, it's outrageously overpriced.
And buying a dual-CPU PowerMac doesn't help even if those aren't overpriced. If I didn't need a second CPU in the first place, overpricing everything but the dual-CPU computers is still a rip off.
"Really, the only reason they went with BSD is because they can keep their version closed (right?)."
The Darwin kernel is open.
"But once again, let me point out why: Because Apple wants to sell pretty, uncrackable, all-in-one, gold standard computers. They've been doing that since the 1970's and they will continue to do so. As we see with OSX and the ability to easily install Linux on a "Macintosh" that the software is just another layer."
I dunno. Most Linux users tend to like to customize the hardware on their computers. The only Macs that can accept substantial upgrades are PowerMacs, and these cost much more money than equivilant PCs. I don't see Apple changing their pricing, so people primarily interested in running Linux can save a lot of money doing it on other hardware. People peripherally interested in Linux will be more likely to use Macs.
" 20 years from now??? I so hope we are not still running nothing but Unix and Unix wannabes?
Sure it "may" have a posix layer but I hope that some progress has been made by then. Maybe something like Plan 9."
The UNIX way of doing things might not be perfect, but it's good enough and more importantly very entrenched. There will be alternatives in 20 years just like there are alternatives now, but UNIX will still be with us.
Yes there's a huge speed advantage for Cell with some things, but for other things there's a big disadvantage. They might be capable of branching, but that doesn't mean they're good at it. You don't give someone 128 registers unless you want them to unroll their loops.
You have to rewrite code for Cell because it's a different environment (limited directly accessible memory, different libraries, etc). The fact that they're not PowerPC is a secondary issue, that's what compilers are for. Getting people to rewrite code for a different environment is like pulling teeth because it's really annoying.
Programmers are more expensive than hardware unless the advantage is huge. Games? Sure. Big CAD workstations? Why not? Scientific simulations? Sweet. Pixar's render farms? Absoloutely. Databases, dynamtic content, web browsers, fancy searches (eg spotlight), etc? Not. A. Chance.
x86 isn't sitting still. I don't know about Intel, but AMD's roadmap talks about co-processors on the die. They'd be x86 compatible, but optimized for different cases like vector-heavy code. If these are a fraction as fast as a Cell, but they work transparently with your existing code, you'll find them winning in the market for all but the most demanding applications.
Also, Cell isn't mutually exclusive with x86. There's no reason a Cell can't sit on a PCIe card much like we have dedicated 3D hardware, and I've no doubt this will happen for some things. The issue is that Cell is the part that you can sometimes do without, while a general purpose processor is almost always mandatory.
"Multi-threaded code is very difficult to write correctly and debug. It's hardly 'easy'."
:)
Let's not get carried away.
When all you want is a sane alternative to non-blocking I/O, or an application that remains responsive when it's busy, no problem. Multi-threaded code is easy to write. Multi-threading CPU-bound code such that it's actually faster is sometimes impossible and usually hard.
"On the PC? Unscrew case, remove the HDD Cage as the memory bank just HAPPNES to be half way under it, make sure to unplug all cables because they are running all across the main board, then put in memory and reverse the whole thing. Takes me roughly 4x as long to upgrade the RAM in a standard PC than in the Mac."
Price of memory: > $60
Time to earn price of memory: > 1 hour (for me anyway, if you make more STFU)
Time to upgrade PowerMac: 60 seconds.
Value of time to upgrade PowerMac: < $1
Time to upgrade PC: 240 seconds (assuming you're correct)
Value of time to upgrade PC: < $4
Difference: $3, 1/20th the cost of the memory
Total lifetime instances of open case: 50 (guessing high)
Total additional cost: $150
Price of PowerMac: $1800 (1.8 ghz, 1 gb, 160 gb 2xDDR, 9600 XT)
Price of equivilant PC: $934 (Dell Dimension 4700, 2.8 ghz, 1 gb DDR2, 160 gb, x16 X300 SE)
Adjusted price of equivilant PC: $1084
"But I guess some peoples time is just not worth anything."
I know what my time is worth. The PowerMac looks like a pretty bad deal.
Sometimes, specialization and generalization are mutually exclusive. Linux is a very powerful general purpose OS, but there's a lot of specializations that they can't have or don't have because their focus is elsewhere.
Fine with me. A monoculture would be a Bad Thing.
Anyone who ties themself to one OS is automatically at a disadvantage.
"On my home XP Pro box, freshly after a reboot, I currently have 15 distinct processes running, with FireFox as the only obviously user-interactive one."
Just because processes are present doesn't mean they're running. They're almost certainly not running. Most processes sit idle, taking up memory and waiting for something to happen. When it does, they take care of it in 5 milliseconds, and then they sit there again. Your CPU is idle for the vast majority of the time, and when you're using it these other processes don't get in your way very much.
Dual-cores are great, but not for this base case.
That's exactly what they'll do. It won't be any cheaper but at least it'll be faster than a G4.
Exclusively Intel OEMs get large discounts.
Linux does a pretty reasonable job. I use XP at work and when it's doing something CPU-bound (generating a key pair with putty sticks out in my mind) the machine becomes unresponsive, but doing the same thing on my Linux machine doesn't have any perceptible effect. 2.4 kernels kinda sucked at that, but 2.6 classifies threads based on whether they use up all their CPU time. If they sleep voluntarily or wait for I/O, they are given higher priority.
Even if the CPU usage is at 100%, benchmarks have shown that interactive processess generally respond in under a millisecond. It's really impressive how a system can be under heavy load but you wouldn't even be able to tell if you couldn't see the network lights blinking like mad, hear the hard drive, and see the CPU temperature going up.
"What I would worry about is more on the chip perfromance side of things - namely the additonal capacitance loading, cross talk and the overall routing density for this approach."
Also the power consumption. I don't have a breakdown of where this technology would be used and where chips spend the most power (apart from knowing cache doesn't take much power), but it might hurt on laptops.
"If the Mathmatica CEO can get called on Wednesday night the week before, asked to bring the source code to Apple, and turn around a native Intel program in two hours of changes, then your developers don't need a year advanced warning. Right?"
That's an already-portable application without many architechture specific optimizations and they only had to get it to "demo quality".
The really big stuff like Photoshop is going to take longer. Even if they gave big vendors more warning, thousands of people would have known about it and it's too big a secret and too easy to leak anonymously to stay secret for long.
If Apple hadn't come clean, the rumors and FUD would have been much worse.
His theories are generally internally consistent but they often have little basis in reality. He also has trouble distinguishing what he would do from something that a company like Apple would do.
"I like the fact that both of my macs running together at full noise are quieter than my PC with one fan. This is because of the CPU. PowerPCs are cooler in both senses of the word."
Maybe you should check with Steve before you continue spouting propaganda. Apple wants cool and quiet computers, so they picked Intel.
Pentium 4s are rediculous, but as far as I can tell Apple will be using Pentium M derived chips for everything except those developer boxes they're shipping now. Pentium M blows Apple away in terms of battery life, they are easily the cooler chip.
"Nobody buys a Mac because it's got a cool processor - they but it because it's got a great interface that makes life easier for them."
Are you kidding? There's a lot of architechture bigots around here. And there's a few (not many, but a few) people that even have rational reasons for prefering PowerPC.
Even though it's completely transparent to most people and 99% of developers aren't programming in assembly, I can't count the number of times I've seen something to the effect of "I can't stand the thought of using such an ugly architechture.".
In the dark days before the G5 they had to convince themselves there was an advantage as they dropped further and further behind. I guess they can't leave that behind. A picture of Steve in front of that giant "performance per watt" slide isn't enough to convince some of them.
Yes. The computer had one processor, and that processor was a Pentium 4.
I'm thinking that it'll be possible to run MacOS X/x86 binaries on Linux with a Wine-like layer if you have the Apple userspace available. Then we can have iTunes, Office, even Photoshop running natively without having to run a full instance of OS X.
The obvious way to do this would be to buy Apple hardware and dual-boot with Linux, simply mounting the OS X partition and using the libraries from there. It might also be possible to extract the files from an install CD for use on cheaper and faster AMD hardware.
Before you flame me: MacOS X isn't fully compatible with Linux on PowerPC or x86, and not all software you might want can/has been ported. Also note that there are plenty of reasons for wanting to do this even if you yourself wouldn't want to.
I had a quick look at PearPC. Apparently they hope to get OS 9 working properly soon, and there's another emulator project called Basilisk.
...
They'll both be really slow, but if you lose two factors of 10 you won't be that far from 68k speeds.
I'm going to cry myself to sleep now...