"But now that Microsoft has a PC emulator, they could let the x86 programs run on non-x86 Windows versions. Is it enough to change the decision? Probably not, but I'd certainly hope that someone at Microsoft is working on it, just in case Intel and AMD would become hostile to Microsoft."
Apart from reinventing the wheel to make everyone feel obsolete and force an upgrade all round, this is what.NET is for.
When Intel was hyping the crap out of IA-64, I think I lot of people at MS believed that it would be the new Alpha; they had to run on that platform if it was going to be that much faster. And then AMD decided they'd make AMD64, and Microsoft realized that it wouldn't be much faster but the memory thing was going to become an issue and regular desktop users would be switching. Add to that the hundreds of millions of legacy boxen and Windows might quickly become as fragmented as UNIX..NET, I believe, was intended to force everyone to ship software in a processor neutral way so that Windows could have a credible port to these other platforms.
If it works out, that might make a PowerPC port an option. However, as (a great many) others have mentioned, PowerPC is not enough faster to justify a port. Now if we were talking about POWER, that might be a different story. But that would not involve any Apple boxen.
"Other things being equal, performance suffers too. More registers means fewer trips to cache, which at the margin means fewer trips to memory, etc -- all of which are vastly slower than what processors typically operate at these days. Intel cranks up the clock, which increases power consumption and heat load, etc."
And yet... performance looks very similar for G5s, Athlon 64s, and Pentium 4s. I would say that other factors have a greater impact on performance.
While Intel's power consumption is rediculous, that's because the Intel marketing people are asshats, and even then, the Pentium M is quite efficient. AMD's power consumption is pretty reasonable in comparison, and the G5 is only marginally better.
And, at the end of the day, things are not equal. They all have localized advantages, but there's no clear cut performance king at the moment.
Besides, if we're going to whip out an compare architechtural superiority, we'll all lose to IA-64. 128 registers, better parallelism than anybody. Look how far it got them.
"There's a reason that many, many embedded computers (everything from cars to set-top boxes) use PPC rather than x86, and it's not just the ease of assembly programming."
Plenty of embedded processors use x86s or relatives of them. 8051's are pretty popular.
"Sure, Joe User running a GUI based program that spends most of its life waiting for the user to do something won't notice much difference -- except in fan noise and heat load, or battery life on a laptop."
What advantage Apple has is due to the amount of attention they devote to ergonomics. Indeed, they resorted to liquid cooling precisely because their heat load is similar to everyone else but their ergonomic standards are higher. They made it a higher priority than anyone else, and they got better results than anyone else. Wow.
Also... Intel Centrino based laptops get a better battery life than any of the Apple laptops because Intel made power savings a higher priority than Apple did.
"Certainly, for you (and most/. readers) a flexible box is a much better choice. Because you are running a true multi-user *nix based system, your box is intrinsically much more secure than any flavor of windows. Most non-geeks however run windows and are afraid to even look inside their computer. For them an iMac is a better choice since it is less likely to give troubles they wouldn't be able to deal with."
Yup.:) You'll note that nowhere in all my rants on this article did I actually deffend Windows.:) When the user just wants a black (in the conceptual sense) box that works properly, an iMac is a perfectly good choice.
What was bugging me in this thread was the Mac zealots blithely overstating the usefulness of the iMac in situations it's not designed for while simultaneously glossing over the expense of a PowerMac. The reality distortion field at work. Yeah, I know, welcome to every Apple thread on Slashdot. But as a Mac user that's not a zealot, it bugs me, and it was worse in this thread.
That's my point. The differences aren't that massive.
The overhead of x86 inefficiency only ties up a small fraction of the transistors, and both of them translate the user visible opcodes into internal micro-ops. Internally, they're quite different, but none of the user visible stuff is that different, and G3s, G4s, and G5s are also different by about the same amount.
As far as capabilities go, some things might be a tad more efficient on G5s, but they both are about the same as far as capabilities go. RISC was different when it was hard to fit an entire processor on one die. Now, the difference is comparatively small.
If you want a difference, look at Itanium, It doesn't even have opcodes, it's got bundles of 3 instructions. Or cell, which is highly parallel in a different way.
"The reasons why I would NOT get a mac are that they are slower and cost MUCH more than equivalent PCs, but more importantly, can't run Windows."
Jesus Christ. If you're going to troll, at least get your facts straight. There's plenty of negative stuff to make fun of, and you went and picked the stuff that's not true.
"Macs only have a chance vs. PCs because they have very efficient architecture."
There's not that big of a difference anymore.
"Macs only have a chance vs. PCs because they have very efficient architecture. Apple doesn't have nearly enough money to compete with Intel or AMD"
Apple doesn't make the CPUs, Motorola and IBM do. Motorolla is pathetic, but IBM can and does keep up with them. Indeed, they collaborate with AMD.
"Second, you create an assembler to assembler compiler. This has been done imperfectly in the past. But IF, and it is a very big if, you do manage to do so, you CAN get up to a theoretical 80%..."
JIT technology has come a long way. Java does it and it competes with C++, as does C#. There was something last year about a JIT that could emulate HPPA binaries on HPPA hardware and it actually got faster due to optimizations... VirtualPC does a pretty respectable job (without the register gap to worry about), and Transmeta does it for the whole computer.
However, as I've said several times in this article, I am skeptical. Not that it can be done, but that it was done by them and that it's as good as they say.
"and keep a really tight loop for the emulator core."
With the speeds they claim, I would suspect some sort of JIT is involved. That could somewhat map the extra registers onto the available ones, or take advantage of x86's pointer operand abilities (32 registers * 32-bits = 128 bytes. That should fit in one cache line.).
Your argument holds water and I personally agree with it, but when trying to persuade others I prefer the stronger argument even if it misses out on key points like that.
The assertion that there's demand that isn't being satisfied is very, very easy to back up given the numbers of p2p stuff; hundreds of millions of users, growth rates above 10-15%/day for months (slowed only by the rapid saturation of the market) at a time in 2000, etc.
Deffending it is hard. Saying "it's inevitable, deal with it" is easy.
These arguments get made a lot, but the fact of the matter is that they actually do own the copyrights on the materials, and legally should be allowed to dictate what happens to them. Even though no direct loss occurs, their copyright is still being violated.
Unfortunately, the members of the **AA abused the market, and now the invisible hand is tearing them a new one. This is nothing more and nothing less than the market attempting to bring the price down to what people are willing to pay.
Anyone that knows economics knows that if the price is that far above what people are willing to pay, the situation is very unstable and it doesn't take much to bring the whole thing crashing down. Napster was the catalyst, but if it hadn't come along someone else would have done it, probably within a few months of the same time.
The market is built into human nature, The **AA's attempts to win a battle against human nature are evidence of the extent to which they are disconnected from reality. It's like trying to stop the drug trade or prostitution. You can spend an arbitrary amount of effort trying, but all you can do is drive up the price, and in the case of all 3, not by very much.
I would agree had my 3rd party memory upgrade not been damaged one of the times I had the logic board repaired. Since it wasn't Apple memory, I was out of pocket for the replacement.
My point is precisely that PowerMacs are very expensive when the only thing in the entire package that you need is expandability. Sure, they're worth it, if what you want is a dual processor workstation. If that is not the case, you're paying a very large premium for a great deal of overkill.
" 3k is not the starting price. It starts at $2,000, and you can pimp it out from there."
Some upgrades (like more than 256 mb memory) are nearly mandatory.
"Remember: you also get _much_ higher quality components. How much noise does this system make?"
Cheap Athlon boxes make a lot of noise. Most other systems, even homebuilds, aren't a problem.
"Also compare power consumption: G5s use about 40 watts per processor, Opterons can be about 130."
What?
All Opterons stay below 95 watts so people will put them in 1U cases. Even the dual core processors they're releasing next year will stay within that limit. You might hit 130 if you overclocked it, but not otherwise.
"The old G3 iMacs can't deal with large drives, but the G4s apparently can."
Perhaps you missed my point...
There's always something new. A computer that can't be upgraded will more quickly fall behind than one that can be upgraded.
The G4 iMacs have 48-bit addressing but they don't have SATA. The G5 iMacs have SATA but they don't have gigabit ethernet. When you can't upgrade you're stuck forever, no matter how nice the feature set initially is. As I said, my Pentium 2 has SATA and gigabit ethernet. I hadn't even heard those words when it was built, but it works fine with them.
Again, this is not a problem for anyone except people that want to upgrade, and those people should not buy iMacs. And, given that PowerMacs are quite a bit more expensive than other computers that can be upgraded, PowerMacs are not always a viable alternative.
"
I keep hearing arguments like this and I think it ignores one key factor: People generally don't upgrade their PCs. Only a small (geek) fraction do more than upgrade their RAM, HD or videocards, which on a Mac is exactly the same cost/difficulty as a PC (except for video cards which are more expensive). Most people don't have or want 3 NICs."
A typically reasonable sounding argument from an Apple supporter. However you're ignoring what I actually said, so you sound like you're right. I'm going to quote myself from the post that you replied to, because this is the part of my position that you ignore: "I'm not saying no one should ever buy an iMac, but I am saying an upgradable computer is worth a lot more to people that actually upgrade them."
Wow. It turns out you're really just accusing me of ignoring a point that I actually make explicitly, and what I said was consistent with your statement. Therefore, I shall consider what I said to stand uncontradicted.
"
Once people have a computer they don't really think about upgrading the CPU alone, until it is way too slow. Then they will go for what is current, but the current processor will generally require a new motherboard, which will also require new ram (and possibly more eg: PCI-X video card). When you put all that together then most people forget it and buy a whole new system anyway. This cycle happens every 2-5 years."
True. And if the now obsolete computer is upgradable, it can do something else instead of remaining a desktop computer that's "way too slow".
"
So for the geek market, sure iMacs suck for upgradability, but geeks should be looking at a PowerMac anyway. For for the other 95% of the market, they do everything that regular people need. As for pricing, they are a medium to high end system that is priced accordingly. One of the parents said it right, they do compare pricewise to Dell systems with comparable specs. Tack OS X on and I think you end up with the best desktop system on the market of people who don't need the flexability because the won't use it, like 95% of the population."
While it's true that PowerMacs are comparable to similarly priced Dell systems, you're completely missing the point that any such system from any vendor is likely bigtime overkill for the needs of the geek in question. Dell is happy to sell you a much cheaper system that is comparable to an iMac, but is also upgradable. If such a system is cheaper and meets the upgradability requirements, it may be preferable to both an iMac and a PowerMac.
Secure protocols take years of careful design and years of peer review. Every standard out there was created behind closed doors with a tight deadline. None of them will last, and when they fail they will be replaced on an emergency basis.
Slashdot Mirror
"But now that Microsoft has a PC emulator, they could let the x86 programs run on non-x86 Windows versions. Is it enough to change the decision? Probably not, but I'd certainly hope that someone at Microsoft is working on it, just in case Intel and AMD would become hostile to Microsoft."
.NET is for.
.NET, I believe, was intended to force everyone to ship software in a processor neutral way so that Windows could have a credible port to these other platforms.
Apart from reinventing the wheel to make everyone feel obsolete and force an upgrade all round, this is what
When Intel was hyping the crap out of IA-64, I think I lot of people at MS believed that it would be the new Alpha; they had to run on that platform if it was going to be that much faster. And then AMD decided they'd make AMD64, and Microsoft realized that it wouldn't be much faster but the memory thing was going to become an issue and regular desktop users would be switching. Add to that the hundreds of millions of legacy boxen and Windows might quickly become as fragmented as UNIX.
If it works out, that might make a PowerPC port an option. However, as (a great many) others have mentioned, PowerPC is not enough faster to justify a port. Now if we were talking about POWER, that might be a different story. But that would not involve any Apple boxen.
"Other things being equal, performance suffers too. More registers means fewer trips to cache, which at the margin means fewer trips to memory, etc -- all of which are vastly slower than what processors typically operate at these days. Intel cranks up the clock, which increases power consumption and heat load, etc."
And yet... performance looks very similar for G5s, Athlon 64s, and Pentium 4s. I would say that other factors have a greater impact on performance.
While Intel's power consumption is rediculous, that's because the Intel marketing people are asshats, and even then, the Pentium M is quite efficient. AMD's power consumption is pretty reasonable in comparison, and the G5 is only marginally better.
And, at the end of the day, things are not equal. They all have localized advantages, but there's no clear cut performance king at the moment.
Besides, if we're going to whip out an compare architechtural superiority, we'll all lose to IA-64. 128 registers, better parallelism than anybody. Look how far it got them.
"There's a reason that many, many embedded computers (everything from cars to set-top boxes) use PPC rather than x86, and it's not just the ease of assembly programming."
Plenty of embedded processors use x86s or relatives of them. 8051's are pretty popular.
"Sure, Joe User running a GUI based program that spends most of its life waiting for the user to do something won't notice much difference -- except in fan noise and heat load, or battery life on a laptop."
What advantage Apple has is due to the amount of attention they devote to ergonomics. Indeed, they resorted to liquid cooling precisely because their heat load is similar to everyone else but their ergonomic standards are higher. They made it a higher priority than anyone else, and they got better results than anyone else. Wow.
Also... Intel Centrino based laptops get a better battery life than any of the Apple laptops because Intel made power savings a higher priority than Apple did.
"The x86 architecture um, how to put this delicately, leaves something to be desired."
As if anyone can tell when they're not coding in assembly.
They've got all the same capabilites, roughly equal performance in most areas, etc.
Because this has no free speech implications at all.
"Certainly, for you (and most /. readers) a flexible box is a much better choice. Because you are running a true multi-user *nix based system, your box is intrinsically much more secure than any flavor of windows. Most non-geeks however run windows and are afraid to even look inside their computer. For them an iMac is a better choice since it is less likely to give troubles they wouldn't be able to deal with."
:) You'll note that nowhere in all my rants on this article did I actually deffend Windows. :) When the user just wants a black (in the conceptual sense) box that works properly, an iMac is a perfectly good choice.
Yup.
What was bugging me in this thread was the Mac zealots blithely overstating the usefulness of the iMac in situations it's not designed for while simultaneously glossing over the expense of a PowerMac. The reality distortion field at work. Yeah, I know, welcome to every Apple thread on Slashdot. But as a Mac user that's not a zealot, it bugs me, and it was worse in this thread.
That's my point. The differences aren't that massive.
The overhead of x86 inefficiency only ties up a small fraction of the transistors, and both of them translate the user visible opcodes into internal micro-ops. Internally, they're quite different, but none of the user visible stuff is that different, and G3s, G4s, and G5s are also different by about the same amount.
As far as capabilities go, some things might be a tad more efficient on G5s, but they both are about the same as far as capabilities go. RISC was different when it was hard to fit an entire processor on one die. Now, the difference is comparatively small.
If you want a difference, look at Itanium, It doesn't even have opcodes, it's got bundles of 3 instructions. Or cell, which is highly parallel in a different way.
Those are different.
"The reasons why I would NOT get a mac are that they are slower and cost MUCH more than equivalent PCs, but more importantly, can't run Windows."
Jesus Christ. If you're going to troll, at least get your facts straight. There's plenty of negative stuff to make fun of, and you went and picked the stuff that's not true.
"Macs only have a chance vs. PCs because they have very efficient architecture."
There's not that big of a difference anymore.
"Macs only have a chance vs. PCs because they have very efficient architecture. Apple doesn't have nearly enough money to compete with Intel or AMD"
Apple doesn't make the CPUs, Motorola and IBM do. Motorolla is pathetic, but IBM can and does keep up with them. Indeed, they collaborate with AMD.
"Second, you create an assembler to assembler compiler. This has been done imperfectly in the past. But IF, and it is a very big if, you do manage to do so, you CAN get up to a theoretical 80%..."
JIT technology has come a long way. Java does it and it competes with C++, as does C#. There was something last year about a JIT that could emulate HPPA binaries on HPPA hardware and it actually got faster due to optimizations... VirtualPC does a pretty respectable job (without the register gap to worry about), and Transmeta does it for the whole computer.
However, as I've said several times in this article, I am skeptical. Not that it can be done, but that it was done by them and that it's as good as they say.
80% of native speed, eg a Pentium 4 3.0 ghz would be as fast as a Pentium 4 2.4 ghz.
If they're telling the truth. However, I share your skepticism.
"and keep a really tight loop for the emulator core."
With the speeds they claim, I would suspect some sort of JIT is involved. That could somewhat map the extra registers onto the available ones, or take advantage of x86's pointer operand abilities (32 registers * 32-bits = 128 bytes. That should fit in one cache line.).
"Let me get this correct. I'm paying more for more computer? Yeah, that sounds right to me. I guess your objective is to pay less for less computer."
Yes. How is this a difficult concept? I don't want to pay extra for stuff that doesn't benefit me.
You're correct, but changing the law won't eliminate the demand any more than the lawsuits will.
Your argument holds water and I personally agree with it, but when trying to persuade others I prefer the stronger argument even if it misses out on key points like that.
The assertion that there's demand that isn't being satisfied is very, very easy to back up given the numbers of p2p stuff; hundreds of millions of users, growth rates above 10-15%/day for months (slowed only by the rapid saturation of the market) at a time in 2000, etc.
Deffending it is hard. Saying "it's inevitable, deal with it" is easy.
These arguments get made a lot, but the fact of the matter is that they actually do own the copyrights on the materials, and legally should be allowed to dictate what happens to them. Even though no direct loss occurs, their copyright is still being violated.
Unfortunately, the members of the **AA abused the market, and now the invisible hand is tearing them a new one. This is nothing more and nothing less than the market attempting to bring the price down to what people are willing to pay.
Anyone that knows economics knows that if the price is that far above what people are willing to pay, the situation is very unstable and it doesn't take much to bring the whole thing crashing down. Napster was the catalyst, but if it hadn't come along someone else would have done it, probably within a few months of the same time.
The market is built into human nature, The **AA's attempts to win a battle against human nature are evidence of the extent to which they are disconnected from reality. It's like trying to stop the drug trade or prostitution. You can spend an arbitrary amount of effort trying, but all you can do is drive up the price, and in the case of all 3, not by very much.
I do as well. Now. :)
Java 1.5 supports templates/generics/whatever.
I would agree had my 3rd party memory upgrade not been damaged one of the times I had the logic board repaired. Since it wasn't Apple memory, I was out of pocket for the replacement.
Best to have everything covered under Applecare.
My point is precisely that PowerMacs are very expensive when the only thing in the entire package that you need is expandability. Sure, they're worth it, if what you want is a dual processor workstation. If that is not the case, you're paying a very large premium for a great deal of overkill.
" 3k is not the starting price. It starts at $2,000, and you can pimp it out from there."
Some upgrades (like more than 256 mb memory) are nearly mandatory.
"Remember: you also get _much_ higher quality components. How much noise does this system make?"
Cheap Athlon boxes make a lot of noise. Most other systems, even homebuilds, aren't a problem.
"Also compare power consumption: G5s use about 40 watts per processor, Opterons can be about 130."
What?
All Opterons stay below 95 watts so people will put them in 1U cases. Even the dual core processors they're releasing next year will stay within that limit. You might hit 130 if you overclocked it, but not otherwise.
Also, G5s use more than 40 watts.
"The old G3 iMacs can't deal with large drives, but the G4s apparently can."
Perhaps you missed my point...
There's always something new. A computer that can't be upgraded will more quickly fall behind than one that can be upgraded.
The G4 iMacs have 48-bit addressing but they don't have SATA. The G5 iMacs have SATA but they don't have gigabit ethernet. When you can't upgrade you're stuck forever, no matter how nice the feature set initially is. As I said, my Pentium 2 has SATA and gigabit ethernet. I hadn't even heard those words when it was built, but it works fine with them.
Again, this is not a problem for anyone except people that want to upgrade, and those people should not buy iMacs. And, given that PowerMacs are quite a bit more expensive than other computers that can be upgraded, PowerMacs are not always a viable alternative.
Does it strike anyone else that that sounds similar to stopping spam, except less easy?
megabytes per second :)
Wow. It turns out you're really just accusing me of ignoring a point that I actually make explicitly, and what I said was consistent with your statement. Therefore, I shall consider what I said to stand uncontradicted.
True. And if the now obsolete computer is upgradable, it can do something else instead of remaining a desktop computer that's "way too slow".
While it's true that PowerMacs are comparable to similarly priced Dell systems, you're completely missing the point that any such system from any vendor is likely bigtime overkill for the needs of the geek in question. Dell is happy to sell you a much cheaper system that is comparable to an iMac, but is also upgradable. If such a system is cheaper and meets the upgradability requirements, it may be preferable to both an iMac and a PowerMac.
Four words: single point of failure
Secure protocols take years of careful design and years of peer review. Every standard out there was created behind closed doors with a tight deadline. None of them will last, and when they fail they will be replaced on an emergency basis.
Not to give away any information about my password, it's possible for my password generator to generate that. So, maybe.