I think the reason that moto/freescale didn't aggressively clock the G4 had to do with the slow memory bus. Eventually there's only a tiny advantage from boosting the clock, when it's so far past the bus.
Right now Apple only feeds the bus at 133MHz (maybe 200, but I'm not sure) and it's not 'double-pumped' or anything. Also, from what I understand, even if you fed the chip more, there are limitations of the MPX bus and the chip design that would still limit the performance. The 8641 aims to fix those problems by widening the paths both inside and outside the chip.
I can say that I've used several VNC implementations and Microsoft's RDC for Mac, and MSRDC rocks the house. I find that it's almost as snappy as 'the real thing' when you disable the XP fancypants appearances and use a W9X theme.
I do Mac and PC support at work, so I have to have a PC handy to run virus scans on NTFS drives (god bless firewire enclosures!). Our ticketing system also runs on PC only. Since I switched from VNC to RDP, I've been enjoying work more, opening and closing tickets is much snappier.
VNC is great, but if you must remote in to a Windows box, RDP is really a much better way to go. I hate the idea too, but it's profoud enough of a difference to ooutweigh my 'use open stuff' policy.
Don't get a KVM, run RDP. You'll have a better time watching your Windows box from a nice iMac screen anyway.
That's exactly the sort of thing that'll hopefully happen. Apple will probably just call it a 'G5' though, and leave the 'mobile' for the techies to know.
I'd like Freescale to get their act together and release these CPUs though, they've been talking them up for a long time now.
And BTW, there are dual-core 8651 CPUs coming 'real soon now' according to Freescale. I'd DEFINITELY want a dual 8651 with RapidIO over a single underclocked 970 in a portable.
No it wouldn't. There's little to no benefits from 64-bit computing on a portable. The G5 was built for machines that can draw a fair amount of wattage. A G5 PowerBook would be hotter, larger, and more power hungry than a machine based on Freescale's 8641 series, a branch off the G4 family.
The only limiting factor of the G4 today is the memory bus, which Freescale has to keep compatible with the ancient 60x bus because of their other clients (like Cisco). The 8641 is a G4 with a totally rebuilt memory controller onboard and RapidIO, an alternative to HyperTransport.
You'd be happier with an 8641-based PowerBook than a 970-based PowerBook. Trust me.
I do think Apple will _call_ the 8641-based laptops 'G5's though, they'll say it has to do with the 'generation of the technology, not a specific type of CPU'.
It almost certainly is, but Comcast is saying that equipment upgrades were the cause. I don't believe it AT ALL, since I run a caching DNS proxy on my LAN which is on Comcast cable.
Last night I was able to browse perfectly fine to all the sites in the server's cache, but I couldn't resolve new sites until I added a non-comcast DNS server to my server's resolv.conf file.
Thank god DNS is as open as it is, I just added a uri.edu DNS server from some old documentation from an old job and things started working again. I needed the 'net last night to test some things, and I'd be damned if I was going to let Comcast get in the way.
On a sidenote, it wouldn't surprise me if the 'equipment upgrade' was actually Comcast replacing their Windows DNS with BSD boxen. I'm hesitant to run nmap against the DNS servers for fear of hacking accusations.
"We let them take our jobs, we pay for their education and healthcare, and they don't even speak our language! I say we round them up and use them as SHUTTLE FUEL!"
Ahh, it's called 'Darwin' and you can download it for free. Instead of Quartz you can use the included framebuffer or standard X11.
Esentially, Darwin is a super-modern BSD with a totally revamped device driver structure, more modern multithreaded startup scripting, a high-performance balanced-tree filesystem with journaling and metadata, included compiler, fantastic hardware support, advanced networking with configd, OpenDirectory for LDAP and AD integration, and self-tuning performance features.
Or the other way around...
on
Hacking Mac OS X
·
· Score: 2, Insightful
Or instead of Apple dropping Quartz, which is a huge part of their appeal (to both users and developers), the Open Source community should start working on and in GNUstep, which is an API-alike of the same technologies that OS X is based on.
Hell, APIs aren't protected I.P., you could make Quartz-compatible APIs for X11 and add them to GNUstep.
It would serve us better to emulate the good things we see out there, not knock them down to our level.
m I going to have to rebuild my entire base system manually because I waited too long between syncs?
No. The message that your configuration isn't valid anymore isn't a death-knell. Change the symlink of/etc/make.profile to a newer target and emerge -u world.
With a system that old, I'd do an emerge -e world too, to recompile the whole system with your new compiler.
The analyst is probably like half the people who see me working on the Mac. They think expose is just something neat you can tap on when you're bored. I personally find it VERY useful to be able to look at any window in under a second (seeing if that 17 meg folder is copoed yet?) or switch to any window in two.
I get a lot of people noting how good my system looks, how nice the fonts render, and how it seems to look good at the high resolutions that make Windows cry, but then they walk away muttering that it's still a toy. There will always be people like that, people who prefer telnet.exe on cmd.exe over ssh in terminal.app.
It's EASY to add a device of an already defined type to an OS, but it's not easy to add support for an entirely new type of bus. Without the Windows source code, you'd have to implement USB as paired kernel and userland serial drivers, it would be ugly and proabably wouldn't work.
Even Linux didn't get USB until 2.4 came around. That might be a long time for most people reading this, but I remember when AGP and USB were shipping on hardware but Linux and Windows couldn't use them (sort of). For a while after AGP came out, the AGP bus enumerated as PCI, and you couldn't use that nifty 'borrow some system RAM for textures' feature (this was back when really good video cards had 4MB onboard). USB only worked for keyboards and mice, and only if you enabled 'HIDBP' in the CMOS, it basically told the BIOS to translate the HID devices to standard PS/2 devices for the OS.
I've been thinking recently about the direction of computing, it seems everything is 'going serial'. SCSI, ATA, FireWire, HyperTransport, USB, these are all serial protocols. It's time an OS focused on having fantastic and robust serial capabilities, and defined the various busses as limits against the entire set of capabilities. Maybe this is getting more towards the microkernel state of mind, but shouldn't all the serial protocols share a command set as far as the kernel is concerned?
Anywho, enough ranting, I'm gonna finish these beers and pray that GCC-3.4.3 compiles cleanly on OS X, wish me luck!
I was under the impression that the amd64 chips would be faster at the builds than the athlon-xp. I'm not sure why, but I remember hearing it somewhere. Maybe more registers to play with, so less 'pressure' on GCC to find working configs?
So can I and I don't use distcc, I've got an athlon xp running at 1600MHz.
If you set all your settings properly, multithread the builds (-j3), enable DMA and readahead on your drives, and keep your CFLAGS sane, the build shouldn't take longer than 15 or so hours on a similar system.
Sure, they might drop NTLMv* authentication, but if you get a ticket from the KDC (usually an Active Directory Controller), you'll have access to what's yours.
This article has to do with authentication only, not the transport, however. They might drop CIFS for something else, which would mean that yes, we'd have re re-reverse engineer whatever file-transport MS uses next round.
I wouldn't worry much abou tit though, it's legal to reverse-engineer this sort of thing. They can't sue you for designing a compatible system, you only need to license stuff based on their implementation using THEIR code.
Re:give us cheap Linux-based PPC machines
on
LinuxPPC64 Contest
·
· Score: 1
BTW, just had the same issue with wireless coming up after login. It totally screwed my plans for wireless AD integration. The workaround is to 'join a specific network' and dial in your SSID in the prefs>network>airport control panel.
This seems to me like it's actually good security. You wouldn't want to be sending out authentication requests over an untrusted network, right?
Nope. GCC-4.0 has a totally new abstracted layer called GIMPLE, from the tree-ssa project. All front ends (C, C++, Java, FORTRAN) output to GIMPLE code, which gets optimized and turned into machine code. It should be MUCH easier to add support for new languages now, all yo have to do is write a front end that turns the input code into GIMPLE.
This is a dramatic oversimplification, but from what I've read on the GCC lists, it appears to be how it works.
You don't want one. Trust me. There's no advantage of slapping a 970-class chip into a portable. Who needs 64-bit computing in a form factor you can't put more than 2GB of RAM into?
What you want is a PowerBook that uses the freescale MPC8641D CPU. It's the heart of a dual-core G4 but it's got the memory controller built onto the CPU, which solves the age-old problem of the G4 being bandwidth-starved.
A bona-fide G5 portable would have to be horribly underclocked, and it would be a square peg in a round hole, a power-hungry and moderately efficient (pipeline-wise) 64-bit CPU where a low-power super-efficient 32-bit CPU can do better.
The only problem now is that the Freescale chip uses RapidIO, which is not HyperTransport, to talk to the system. It would probably behoove freescale to HyperTransport-ize the G4 series and get in on the game.
I expect the next major PowerBook revision to be MPC8641 based, it'll require a new motherboard, but a dual-core 1.8GHz PowerBook will more than meet anyone's needs.
They're good if you run their Windoze drivers, but the open-source drivers don't support the hardware video en/decoding, IIRC.
I had an EPIA 800 and couldn't stand how 'fuzzy' the output was on my nice monitor, VIA's mini-ITX products seem to be quite low-quality, but they do get the job done.
If I was going to 'go mini' I'd have to go with something like this, a Pentium-M on an Intel chipset, their stuff sems a lot more 'solid' than VIA's.
If you need AD itself, and can't use SAMBA/OD, you've got to either run parallel systems or put AD on top. There's no way to slave AD off of OD for now, only the other way around.
It would be cool if someone built an authentication/policy interface to OD for Windows though, or made some sort of AD-compatible transport and attribute mode for OpenLDAP.
Slashdot Mirror
I think the reason that moto/freescale didn't aggressively clock the G4 had to do with the slow memory bus. Eventually there's only a tiny advantage from boosting the clock, when it's so far past the bus.
Right now Apple only feeds the bus at 133MHz (maybe 200, but I'm not sure) and it's not 'double-pumped' or anything. Also, from what I understand, even if you fed the chip more, there are limitations of the MPX bus and the chip design that would still limit the performance. The 8641 aims to fix those problems by widening the paths both inside and outside the chip.
I can say that I've used several VNC implementations and Microsoft's RDC for Mac, and MSRDC rocks the house. I find that it's almost as snappy as 'the real thing' when you disable the XP fancypants appearances and use a W9X theme.
I do Mac and PC support at work, so I have to have a PC handy to run virus scans on NTFS drives (god bless firewire enclosures!). Our ticketing system also runs on PC only. Since I switched from VNC to RDP, I've been enjoying work more, opening and closing tickets is much snappier.
VNC is great, but if you must remote in to a Windows box, RDP is really a much better way to go. I hate the idea too, but it's profoud enough of a difference to ooutweigh my 'use open stuff' policy.
Don't get a KVM, run RDP. You'll have a better time watching your Windows box from a nice iMac screen anyway.
How about "G5 Mobile"?
That's exactly the sort of thing that'll hopefully happen. Apple will probably just call it a 'G5' though, and leave the 'mobile' for the techies to know.
I'd like Freescale to get their act together and release these CPUs though, they've been talking them up for a long time now.
And BTW, there are dual-core 8651 CPUs coming 'real soon now' according to Freescale. I'd DEFINITELY want a dual 8651 with RapidIO over a single underclocked 970 in a portable.
No it wouldn't. There's little to no benefits from 64-bit computing on a portable. The G5 was built for machines that can draw a fair amount of wattage. A G5 PowerBook would be hotter, larger, and more power hungry than a machine based on Freescale's 8641 series, a branch off the G4 family.
The only limiting factor of the G4 today is the memory bus, which Freescale has to keep compatible with the ancient 60x bus because of their other clients (like Cisco). The 8641 is a G4 with a totally rebuilt memory controller onboard and RapidIO, an alternative to HyperTransport.
You'd be happier with an 8641-based PowerBook than a 970-based PowerBook. Trust me.
I do think Apple will _call_ the 8641-based laptops 'G5's though, they'll say it has to do with the 'generation of the technology, not a specific type of CPU'.
It almost certainly is, but Comcast is saying that equipment upgrades were the cause. I don't believe it AT ALL, since I run a caching DNS proxy on my LAN which is on Comcast cable.
Last night I was able to browse perfectly fine to all the sites in the server's cache, but I couldn't resolve new sites until I added a non-comcast DNS server to my server's resolv.conf file.
Thank god DNS is as open as it is, I just added a uri.edu DNS server from some old documentation from an old job and things started working again. I needed the 'net last night to test some things, and I'd be damned if I was going to let Comcast get in the way.
On a sidenote, it wouldn't surprise me if the 'equipment upgrade' was actually Comcast replacing their Windows DNS with BSD boxen. I'm hesitant to run nmap against the DNS servers for fear of hacking accusations.
Next thing you know, it'll be illegal immigrants.
That's a winning party platform right there:
"We let them take our jobs, we pay for their education and healthcare, and they don't even speak our language! I say we round them up and use them as SHUTTLE FUEL!"
Ahh, it's called 'Darwin' and you can download it for free. Instead of Quartz you can use the included framebuffer or standard X11.
Esentially, Darwin is a super-modern BSD with a totally revamped device driver structure, more modern multithreaded startup scripting, a high-performance balanced-tree filesystem with journaling and metadata, included compiler, fantastic hardware support, advanced networking with configd, OpenDirectory for LDAP and AD integration, and self-tuning performance features.
Or instead of Apple dropping Quartz, which is a huge part of their appeal (to both users and developers), the Open Source community should start working on and in GNUstep, which is an API-alike of the same technologies that OS X is based on.
Hell, APIs aren't protected I.P., you could make Quartz-compatible APIs for X11 and add them to GNUstep.
It would serve us better to emulate the good things we see out there, not knock them down to our level.
m I going to have to rebuild my entire base system manually because I waited too long between syncs?
/etc/make.profile to a newer target and emerge -u world.
No. The message that your configuration isn't valid anymore isn't a death-knell. Change the symlink of
With a system that old, I'd do an emerge -e world too, to recompile the whole system with your new compiler.
Well 'Chimera' reminds ME of the lump on my back that was supposed to be my twin brother. Oh the humanity!
The analyst is probably like half the people who see me working on the Mac. They think expose is just something neat you can tap on when you're bored. I personally find it VERY useful to be able to look at any window in under a second (seeing if that 17 meg folder is copoed yet?) or switch to any window in two.
I get a lot of people noting how good my system looks, how nice the fonts render, and how it seems to look good at the high resolutions that make Windows cry, but then they walk away muttering that it's still a toy. There will always be people like that, people who prefer telnet.exe on cmd.exe over ssh in terminal.app.
It's EASY to add a device of an already defined type to an OS, but it's not easy to add support for an entirely new type of bus. Without the Windows source code, you'd have to implement USB as paired kernel and userland serial drivers, it would be ugly and proabably wouldn't work.
Even Linux didn't get USB until 2.4 came around. That might be a long time for most people reading this, but I remember when AGP and USB were shipping on hardware but Linux and Windows couldn't use them (sort of). For a while after AGP came out, the AGP bus enumerated as PCI, and you couldn't use that nifty 'borrow some system RAM for textures' feature (this was back when really good video cards had 4MB onboard). USB only worked for keyboards and mice, and only if you enabled 'HIDBP' in the CMOS, it basically told the BIOS to translate the HID devices to standard PS/2 devices for the OS.
I've been thinking recently about the direction of computing, it seems everything is 'going serial'. SCSI, ATA, FireWire, HyperTransport, USB, these are all serial protocols. It's time an OS focused on having fantastic and robust serial capabilities, and defined the various busses as limits against the entire set of capabilities. Maybe this is getting more towards the microkernel state of mind, but shouldn't all the serial protocols share a command set as far as the kernel is concerned?
Anywho, enough ranting, I'm gonna finish these beers and pray that GCC-3.4.3 compiles cleanly on OS X, wish me luck!
The Holy Empire of Intellectual Rights?
:-)
Seems to work. Though I hadn't thought of it.
I was under the impression that the amd64 chips would be faster at the builds than the athlon-xp. I'm not sure why, but I remember hearing it somewhere. Maybe more registers to play with, so less 'pressure' on GCC to find working configs?
So can I and I don't use distcc, I've got an athlon xp running at 1600MHz.
If you set all your settings properly, multithread the builds (-j3), enable DMA and readahead on your drives, and keep your CFLAGS sane, the build shouldn't take longer than 15 or so hours on a similar system.
Nope. It doesn't work that way.
Sure, they might drop NTLMv* authentication, but if you get a ticket from the KDC (usually an Active Directory Controller), you'll have access to what's yours.
This article has to do with authentication only, not the transport, however. They might drop CIFS for something else, which would mean that yes, we'd have re re-reverse engineer whatever file-transport MS uses next round.
I wouldn't worry much abou tit though, it's legal to reverse-engineer this sort of thing. They can't sue you for designing a compatible system, you only need to license stuff based on their implementation using THEIR code.
BTW, just had the same issue with wireless coming up after login. It totally screwed my plans for wireless AD integration. The workaround is to 'join a specific network' and dial in your SSID in the prefs>network>airport control panel.
This seems to me like it's actually good security. You wouldn't want to be sending out authentication requests over an untrusted network, right?
Nope. GCC-4.0 has a totally new abstracted layer called GIMPLE, from the tree-ssa project. All front ends (C, C++, Java, FORTRAN) output to GIMPLE code, which gets optimized and turned into machine code. It should be MUCH easier to add support for new languages now, all yo have to do is write a front end that turns the input code into GIMPLE.
This is a dramatic oversimplification, but from what I've read on the GCC lists, it appears to be how it works.
You don't want one. Trust me. There's no advantage of slapping a 970-class chip into a portable. Who needs 64-bit computing in a form factor you can't put more than 2GB of RAM into?
What you want is a PowerBook that uses the freescale MPC8641D CPU. It's the heart of a dual-core G4 but it's got the memory controller built onto the CPU, which solves the age-old problem of the G4 being bandwidth-starved.
A bona-fide G5 portable would have to be horribly underclocked, and it would be a square peg in a round hole, a power-hungry and moderately efficient (pipeline-wise) 64-bit CPU where a low-power super-efficient 32-bit CPU can do better.
The only problem now is that the Freescale chip uses RapidIO, which is not HyperTransport, to talk to the system. It would probably behoove freescale to HyperTransport-ize the G4 series and get in on the game.
I expect the next major PowerBook revision to be MPC8641 based, it'll require a new motherboard, but a dual-core 1.8GHz PowerBook will more than meet anyone's needs.
They're good if you run their Windoze drivers, but the open-source drivers don't support the hardware video en/decoding, IIRC.
I had an EPIA 800 and couldn't stand how 'fuzzy' the output was on my nice monitor, VIA's mini-ITX products seem to be quite low-quality, but they do get the job done.
If I was going to 'go mini' I'd have to go with something like this, a Pentium-M on an Intel chipset, their stuff sems a lot more 'solid' than VIA's.
I know, check my username, I've been 'MarcQuadra' since my Dad bought me a 660av when I was 14 so I would get the heck off his computer.
Many fond memories of A-TRAIN, SimCity 2000, and A-10 Attack! on my Quadra float through my head when I'm drifting off to sleep each night.
I take personal offense! Quadra is a perfectly acceptable name!
If you need AD itself, and can't use SAMBA/OD, you've got to either run parallel systems or put AD on top. There's no way to slave AD off of OD for now, only the other way around.
It would be cool if someone built an authentication/policy interface to OD for Windows though, or made some sort of AD-compatible transport and attribute mode for OpenLDAP.
Linux already has the technologies that comprise WinFS...
But does Linux have a very persistent yellow doggie to step me through the process?
I didn't think so. I'll stick with what I know.
OOH! HE DOES TRICKS!
Alright, that's what I was hoping you'd say.
I was really hoping you weren't going to tell me about how your dual-boot box didn't play UT2004 as well.