Open isn't perfect. I had a problem where you could crash the machine by doing a ping flood on a Intel PRO/1000MT gigabit card. That was 3.5-stable. It's closer to NetBSD than FreeBSD though.
Of course, on Linux 2.6 it's a good day where there *isn't* some horrible flaw.
"I use my cell phone, iPod, PS2, pda, and PVR more often than I use my desktop machine. That's enough convincing for me. I'm not in the business of convincing others."
Okay,
a) I think you missed the point. I don't think anyone believes x86 will be cutting edge for the next hundred years, just that it will be supported for the next hundred years. It's basically a matter of support for legacy software. That could easily mean emulators, or any number of other things. IBM still has support for binaries from the 50s, and they barely had general purpose computers back then, so this is not hard to believe.
b) Just because you use your mobile devices more than your desktop doesn't mean you don't use your desktop, or that no one else does. Or that there aren't servers that use x86.
c) Whatever dreams we had for ditching x86 support soon were forgotten when AMD came up with AMD64. That reset the clock for the obsolescence of x86 to at least a decade in the future, and that's assuming a processor comes out tomorrow that's so much better there's no choice but to make the change as fast as possible.
d) Cell is not that processor. The controlling processor (according to the article) is a regular PowerPC (or POWER) processor, and anything that deals with general purpose code is stuck there. That means for everything that doesn't involve crunching through vast arrays of numbers, it won't help a lot.
e) The networkability and extensibility the article discussed means it will be easy to attach one of these to an x86 processor, or stick one in a card and use it as a coprocessor for numerically intensive tasks.
x86 is a terrible hack, but we aren't going to get rid of it soon. Now that it's 64-bit, it'll be with us longer than anyone reading this thread will live.
Mass storage technologies are not restricted by the address space of the CPU. If they were, you wouldn't be able to have multi-terabyte RAID arrays in 32-bit PCs.
"Can you imagine if your bittorrent client could use one or two APU's and leave the others for other tasks."
Uhg... RTFA
The APUs are not general purpose CPUs. They can crunch through big arrays of numbers but NOT run branchy general purpose code. There's a more-or-less standard PowerPC chip to control them and it has to run all the general-purpose code.
"That's ridiculous. x86 is dead. The overheating and power consumption confirms it."... right, because 90 nm Athlon64s and Centrinos use SO much power.
(hint: they don't)
"CISC hardware is horrible in mobile devices because of battery life and power consumption."
68K's are just fine for embedded use.
"All next generation consoles will use CISC hardware. Hence, economies of scale to get the price down."
You mean RISC, right?
"x86 is dead and mobile devices wrote the eulogy"
A more articulate post will be required to convinced me, I've afraid.
I disagree with zealotry in any form, and that includes ignoring Linux as an alternative when it's actually pretty decent at a lot of things.
However, the 2.6 kernel has been a mess from day one. I'm used to new versions being a bit buggy, but the kernel developers keep adding new bugs.
First, there was that memory leak with burners. Now more recently they've been moving to libata, and that prevents SATA and PATA drives from being used at the same time on my system because libata does not yet support PATA drives, and all the ATA controllers are on the same chipset. I've had to install an old Promise ATA controller to have a CD drive on my system.
They say it's up to the distributions to stabalize their own kernels, but the distributions clearly don'thave the resources to do that.
Cringely isn't talking about an HDTV PVR. He's talking about an online movie store, all the user would have to do is decode, which Mac minis should be able to handle.
I think it's safe to assume Apple will talk the studios into allowing DRM that people will be able to live with, given the success of the iTunes music store.
"PPC has a better architecture, and the P4 has such an insanely long pipeline that if the branch predictor make a mistake then you take a big performance hit reloading all 30 stages."
Most processors have something wrong with them.
P4s have the long pipeline, G4s have the impotent bus and low clock speed. Both generally perform well with media tasks, but the G4 is at a significant disadvantage. Media codecs generally do not have much in the way of branches. They're what the P4 was designed for and it really shines there.
However, I am comfortable assuming that the 1.25 ghz Mac mini is fast enough for whatever Apple will choose, if they do indeed start selling movies online. I think it's also safe to assume they can get more out of a G4 than VLC and mplayer, as they can specialize for Mac a lot more and have the resources to hand tune the codec. If they're going to be selling movies for the damn thing, that's worth the investment.
Also, they can probably do stuff like decode and cache difficult parts of the file ahead of time, since they control the whole software stack.
"Actually, IIRC, it is because video encoding software can use intel SSE2 instructions to gain performance, not because of the higher clock speed."
It's partly the hyperthreading (does well when the cache isn't being heavily relied upon), partly the clock speed (codecs don't branch a lot so the P4 is favored here), and yes, partly SSE2 (although the G4 has altivec so it balances out).
P4s aren't good at many things, but they are fantastic at video encoding. Athlon64s can't touch them, G5s can't touch them.
Of course, decoding these movies is not a problem for any of these processors, so most of us don't need to worry. The only one that's close is the Mac mini and if Apple's going to be selling movies for them, I think it's safe to assume they'll be supporting them with the software.
X11 is the protocol. XFree86 is the implementation that nobody likes anymore due to the change in license. X.org is the implementation that everyone is moving to. There are a number of other implementations.
Apple committed to XFree86 long before they changed the license. They may move to X.org as that implementation gets better features.
"Updating a freebsd system (3rd party packages) is much easier with the ports system, and it is FRee. You dont' have to pay a subscription to use up2date, or have a local satellite server."
As a BSD and Linux user, I think apt-get and ports are pretty evenly matched. apt-get is more automated, ports is more deterministic.
Some of the Linuxes are 31337 and unreliable (Gentoo), some are more stable than BSD (By stable I mean no major changes for long periods of time.) (Debian-stable).
It's a mistake to paint all the Linuxes with one brush. Don't confuse Gentoo fanbois with Debian users.
Personally, I've never been able to get OpenBSD or FreeBSD working satisfactorally as a workstation OS, but Debian-testing has been great. Conversely, Debian-testing is probably the best workstation OS I've ever used.
If you're talking about time dilation... it wasn't going fast enough for that to be significant. The only thing to remember is that it's so far away that the signals take 67 minutes to get back here.
"Tucked away in a discussion about Apple's manufacturing partners are references to an iBook G5 and a PowerBook G5, which will ship in Q2 2005."
Even if there were a PowerBook G5 on the way, and that's a big 'if' for H1... there's no way it would happen for the iBooks in that timeframe. None whatsoever. Everybody knows how Apple feels about their market segmentation.
The new logo is so neutral and devoid of meaning, so "politically correct" that it says nothing at all about the project. It's not even distinctive enough to be recognized out of context.
If I see a blowfish, I know it's OpenBSD. If I see Tux, I know it's Linux. But if I saw that damned flag without the word "NetBSD" nearby, there's no way I'd associate it with the project.
This is why mascots should not be chosen by committee.
Apple has an airtight case, and there's absoloutely no way he can cover a judgement against him or even the costs of a trial. People do funny things when there's no way out.
A local root exploit gives anyone root on a Linux box until it's patched. The interface stays the same.
On Windows, the messaging interface allows any program to gain admin privs at any time. It can't be disabled or changed without breaking compatability.
I've had the same thought as the person you responded to... it seems unlikely that clock speeds will increase a whole lot (maybe one or two doublings left), but I think it's safe to assume that we'll be seeing more and more cores on each die, and that increasing the number of memory buesses isn't practical (too many wires), so we'll be moving to higher bandwidth, higher latency memory busses that the CPUs cores will have to share.
I was thinking that we could still have physical memory as we know it today, but it won't be directly usable by the processors. They'll treat it as swap space that's faster than a hard drive would be.
The issue with this is overhead in handling page faults. Currently it's not even close to worth it. Significant optimizations and hardware support would be necessary.
"I base my research on this the ATA chip list posted on linux-kernel in 2002. According to that, Intel chipset since PIIX3 "supports" LBA48. In fact, this support lies mostly on the drive, the ATA interface can even be considered very dumb, most intelligence is in the drive logic. The earliest chipsets could have been too smart being dumb..."
huh. Well, you learn something every day. I thought that 48-bit addressing didn't come in until much later.
"Incidentally, PIIX3 is what "this" machine with a P166 is having, in which I plan to use a Maxtor 250G, so a P120 is not that far off. It probably won't have UDMA, granted, but working it will be. It currently is in an ALI (the horror?) K6-2/400 based machine with 2.6.10"
I have a P100 that I have doing DMA with a Promise ATA-100 controller. The DMA is worth it with computers that slow, if you can spare the PCI slots.
"I realize this is too down in the threads for many to read it, but maybe you will:)"
Slashdot Mirror
"Gentoo Linux"
eeew
Open isn't perfect. I had a problem where you could crash the machine by doing a ping flood on a Intel PRO/1000MT gigabit card. That was 3.5-stable. It's closer to NetBSD than FreeBSD though.
Of course, on Linux 2.6 it's a good day where there *isn't* some horrible flaw.
"I use my cell phone, iPod, PS2, pda, and PVR more often than I use my desktop machine. That's enough convincing for me. I'm not in the business of convincing others."
Okay,
a) I think you missed the point. I don't think anyone believes x86 will be cutting edge for the next hundred years, just that it will be supported for the next hundred years. It's basically a matter of support for legacy software. That could easily mean emulators, or any number of other things. IBM still has support for binaries from the 50s, and they barely had general purpose computers back then, so this is not hard to believe.
b) Just because you use your mobile devices more than your desktop doesn't mean you don't use your desktop, or that no one else does. Or that there aren't servers that use x86.
c) Whatever dreams we had for ditching x86 support soon were forgotten when AMD came up with AMD64. That reset the clock for the obsolescence of x86 to at least a decade in the future, and that's assuming a processor comes out tomorrow that's so much better there's no choice but to make the change as fast as possible.
d) Cell is not that processor. The controlling processor (according to the article) is a regular PowerPC (or POWER) processor, and anything that deals with general purpose code is stuck there. That means for everything that doesn't involve crunching through vast arrays of numbers, it won't help a lot.
e) The networkability and extensibility the article discussed means it will be easy to attach one of these to an x86 processor, or stick one in a card and use it as a coprocessor for numerically intensive tasks.
x86 is a terrible hack, but we aren't going to get rid of it soon. Now that it's 64-bit, it'll be with us longer than anyone reading this thread will live.
Mass storage technologies are not restricted by the address space of the CPU. If they were, you wouldn't be able to have multi-terabyte RAID arrays in 32-bit PCs.
"Can you imagine if your bittorrent client could use one or two APU's and leave the others for other tasks."
Uhg... RTFA
The APUs are not general purpose CPUs. They can crunch through big arrays of numbers but NOT run branchy general purpose code. There's a more-or-less standard PowerPC chip to control them and it has to run all the general-purpose code.
"That's ridiculous. x86 is dead. The overheating and power consumption confirms it." ... right, because 90 nm Athlon64s and Centrinos use SO much power.
(hint: they don't)
"CISC hardware is horrible in mobile devices because of battery life and power consumption."
68K's are just fine for embedded use.
"All next generation consoles will use CISC hardware. Hence, economies of scale to get the price down."
You mean RISC, right?
"x86 is dead and mobile devices wrote the eulogy"
A more articulate post will be required to convinced me, I've afraid.
"Sorry, I saw your sig and couldn't resist. : D"
:)
he he
No appologies necessary. It was funny.
I disagree with zealotry in any form, and that includes ignoring Linux as an alternative when it's actually pretty decent at a lot of things.
However, the 2.6 kernel has been a mess from day one. I'm used to new versions being a bit buggy, but the kernel developers keep adding new bugs.
First, there was that memory leak with burners. Now more recently they've been moving to libata, and that prevents SATA and PATA drives from being used at the same time on my system because libata does not yet support PATA drives, and all the ATA controllers are on the same chipset. I've had to install an old Promise ATA controller to have a CD drive on my system.
They say it's up to the distributions to stabalize their own kernels, but the distributions clearly don'thave the resources to do that.
Time to fork 2.7 guys!
Cringely isn't talking about an HDTV PVR. He's talking about an online movie store, all the user would have to do is decode, which Mac minis should be able to handle.
I think it's safe to assume Apple will talk the studios into allowing DRM that people will be able to live with, given the success of the iTunes music store.
"PPC has a better architecture, and the P4 has such an insanely long pipeline that if the branch predictor make a mistake then you take a big performance hit reloading all 30 stages."
Most processors have something wrong with them.
P4s have the long pipeline, G4s have the impotent bus and low clock speed. Both generally perform well with media tasks, but the G4 is at a significant disadvantage. Media codecs generally do not have much in the way of branches. They're what the P4 was designed for and it really shines there.
However, I am comfortable assuming that the 1.25 ghz Mac mini is fast enough for whatever Apple will choose, if they do indeed start selling movies online. I think it's also safe to assume they can get more out of a G4 than VLC and mplayer, as they can specialize for Mac a lot more and have the resources to hand tune the codec. If they're going to be selling movies for the damn thing, that's worth the investment.
Also, they can probably do stuff like decode and cache difficult parts of the file ahead of time, since they control the whole software stack.
"Actually, IIRC, it is because video encoding software can use intel SSE2 instructions to gain performance, not because of the higher clock speed."
It's partly the hyperthreading (does well when the cache isn't being heavily relied upon), partly the clock speed (codecs don't branch a lot so the P4 is favored here), and yes, partly SSE2 (although the G4 has altivec so it balances out).
P4s aren't good at many things, but they are fantastic at video encoding. Athlon64s can't touch them, G5s can't touch them.
Of course, decoding these movies is not a problem for any of these processors, so most of us don't need to worry. The only one that's close is the Mac mini and if Apple's going to be selling movies for them, I think it's safe to assume they'll be supporting them with the software.
X11 is the protocol. XFree86 is the implementation that nobody likes anymore due to the change in license. X.org is the implementation that everyone is moving to. There are a number of other implementations.
Apple committed to XFree86 long before they changed the license. They may move to X.org as that implementation gets better features.
They have that at Calgary International. Or at least they did in jan 2001...
"Updating a freebsd system (3rd party packages) is much easier with the ports system, and it is FRee. You dont' have to pay a subscription to use up2date, or have a local satellite server."
As a BSD and Linux user, I think apt-get and ports are pretty evenly matched. apt-get is more automated, ports is more deterministic.
Some of the Linuxes are 31337 and unreliable (Gentoo), some are more stable than BSD (By stable I mean no major changes for long periods of time.) (Debian-stable).
It's a mistake to paint all the Linuxes with one brush. Don't confuse Gentoo fanbois with Debian users.
Personally, I've never been able to get OpenBSD or FreeBSD working satisfactorally as a workstation OS, but Debian-testing has been great. Conversely, Debian-testing is probably the best workstation OS I've ever used.
If you're talking about time dilation... it wasn't going fast enough for that to be significant. The only thing to remember is that it's so far away that the signals take 67 minutes to get back here.
"Tucked away in a discussion about Apple's manufacturing partners are references to an iBook G5 and a PowerBook G5, which will ship in Q2 2005."
Even if there were a PowerBook G5 on the way, and that's a big 'if' for H1... there's no way it would happen for the iBooks in that timeframe. None whatsoever. Everybody knows how Apple feels about their market segmentation.
I was just thinking that. The only brought 3.3 into -testing like a week ago.
The new logo is so neutral and devoid of meaning, so "politically correct" that it says nothing at all about the project. It's not even distinctive enough to be recognized out of context.
If I see a blowfish, I know it's OpenBSD. If I see Tux, I know it's Linux. But if I saw that damned flag without the word "NetBSD" nearby, there's no way I'd associate it with the project.
This is why mascots should not be chosen by committee.
The PNG format is lossless, and very widely deployed (pretty much all browsers and image programs, etc).
Isn't DMA broken on the Intel x86-64 chips, slowing things down?
Until companies pay off Microsoft to allow their spyware to be installed?
Apple has an airtight case, and there's absoloutely no way he can cover a judgement against him or even the costs of a trial. People do funny things when there's no way out.
The difference is simple.
A local root exploit gives anyone root on a Linux box until it's patched. The interface stays the same.
On Windows, the messaging interface allows any program to gain admin privs at any time. It can't be disabled or changed without breaking compatability.
That's the difference.
Because it's broken as implemented instead of broken as designed (eg messaging interface that can give anything admin privs and can't be changed).
I've had the same thought as the person you responded to... it seems unlikely that clock speeds will increase a whole lot (maybe one or two doublings left), but I think it's safe to assume that we'll be seeing more and more cores on each die, and that increasing the number of memory buesses isn't practical (too many wires), so we'll be moving to higher bandwidth, higher latency memory busses that the CPUs cores will have to share.
I was thinking that we could still have physical memory as we know it today, but it won't be directly usable by the processors. They'll treat it as swap space that's faster than a hard drive would be.
The issue with this is overhead in handling page faults. Currently it's not even close to worth it. Significant optimizations and hardware support would be necessary.
"I base my research on this the ATA chip list posted on linux-kernel in 2002. According to that, Intel chipset since PIIX3 "supports" LBA48. In fact, this support lies mostly on the drive, the ATA interface can even be considered very dumb, most intelligence is in the drive logic. The earliest chipsets could have been too smart being dumb..."
:)"
:)
huh. Well, you learn something every day. I thought that 48-bit addressing didn't come in until much later.
"Incidentally, PIIX3 is what "this" machine with a P166 is having, in which I plan to use a Maxtor 250G, so a P120 is not that far off. It probably won't have UDMA, granted, but working it will be. It currently is in an ALI (the horror?) K6-2/400 based machine with 2.6.10"
I have a P100 that I have doing DMA with a Promise ATA-100 controller. The DMA is worth it with computers that slow, if you can spare the PCI slots.
"I realize this is too down in the threads for many to read it, but maybe you will
Yup.