Was unfortunately left out. I mean, Athlon 64 makes a fine Pentium 4 competitor when running a legacy 32-bit operating system, but it's so much more. Those cool extra registers you get in 64-bit mode make the thing just scream!
And no, the intel EM64T stuff isn't even competing in the same league, 40-45% slower with 40% more GHz is what I've seen in real-life workloads (heavy numbercrunching). For some other types of loads it does just about as well as the a64/opteron, though.
Revised x86_64 support (possibly in the pentium m core and in the same price range as the new 90nm a64's) and Intel has a chance. That and Microsoft delaying 64-bit Windows for a couple more years.
Has been "proven" for a long time. Every time they send something up there they seem to do the same story. Just google for water on mars and the first hit is a NASA press release from 2000.
And that's hardly the first "evidence", there's been plenty of evidence to show that there have been glaciers on Mars at some point. Or so I've been told by someone who studies Mars for a living:-)
Although privatization (especially in Britain) often does not work, some financial incentive for efficiently utilizing the available frequencies would be useful.
In other words, if you're wasting say 5MHz of perfectly good bandwidth, and replacing all your kit with something modern doing the same in 0.5MHz would cost you X, it would be reasonable if you'd end up saving money if you did so (of course replacing all those millions of radios is not an option, just look at where DAB went).
HAM people already do this, sort of (not benefiting financially ofcourse, just in a everyone benefits sort of way). It pays off to use the very limited resources they get as efficiently as possible so they do so.
The problem is that it penalizes countries that started reducing their emissions _BEFORE_ 1990. Now if you were polluting like hell back then, reducing your emissions is easy...
This is actually what happened to Finland. Stupid politicians. It would have been totally fine to just ignore the treaty and just announce that we were unilaterally doing the equivalent thing except using the peak emission year of the 80's as the basis.
Indeed. And a64 really likes 64-bit code, even if it is a bit more bloated. Those extra registers really make things fly (unlike some other 64 bit archs like sparc64 where the biggest benefit is getting a 64-bit virtual address space and everything that doesn't need that typically gets compiled 32-bit because it's faster).
It's just that the proprietary operating system most hardware sites concentrate on doesn't have a non-beta 64-bit version (if you don't count that CPU that Linux has probably a 95% market share per sold cpu on, per box 2003 Server might do a bit better)
Well to be honest Anandtech is actually one of the few hardware sites that does have some kind of Linux benchmarks. Alas they don't "get" the open source way and just do "distro X is 2% faster than distro Y when running Z" style benchmarks when the approach should be to figure out why the differences are there. Sure it requires more effort and competence, but it's be a _lot_ more useful than the benchmarks they do now.
One of the Finnish election observers says he saw a touchscreen device break down (with 200 votes inside), and it was replaced and the officials were left pondering what to do with the broken one. Apparently it was one of those touch-screen no-paper-trail ones. Interesting to see if this is an isolated incident or not.
Drives from just about any vendor can suck. I know people who had 1000 IBM 75GXP drives (hey, their previous models were ok!) and had to swap a drive a day due to errors (not total failures tho, just early signs of it). Although many people seem to recommend "identical drives" for raid clusters for performance reasons or whatever, mixing vendors or at least production batches does the benefit of a single vendor fuckup resulting in data loss due to all of your drives having similar early deaths.
As for controllers, lots of big installations use 3ware cards. A bit more expensive than the basic ide/sata adapters, but you get lots of ports in one pci (preferably something better than 32-bit/33MHz) slot, and each drive gets their own ide bus. Now, even if the things have hardware raid5, you might be better off using plain old software raid with them. Depending on the card model it can easily be twice as fast. Embedded CPU's aren't that fast compared to a p4 or a64:-)
Once you have 32 CPUs or so and have an OS that is able to run pretty well on those (Linux does, depending on the workload of course, people are using Altixes with 256-512 in a single system image and are quite happy with them), being to able to replace cpus without rebooting starts getting pretty important (since occasionally they do fail).
See, if you didn't visit third party sites, just microsofts marketing pages so you'd know which of their products you should be buying next, you'd be totally safe...
Because they only support older models (pre-Radeon 9600 or so). While some work to add open source support to the R300 series is being done, it doesn't work yet. http://volodya-project.sourceforge.net/R300.php seems to be the site of the effort.
From what I've used the binary drivers, they're not _that_ hard to get running, on a friends fc2 laptop it was a matter of copying a few dri header files from the kernel sourcecode (the ATI drivers should be including a copy of these since there's no guarantee the kernel DRM stuff will continue using it in the future)
Anyway, it was a matter of running make after that. Not as easy as the nvidia stuff, but still not that hard either.
I've played it that way on my athlon 2000+geforce 4 ti4200. It's quite playable, most notable problems are that it's _SLOW_ for about 15-20 secs after loading a game. Going to the menus and using the PDA for a while and it goes back to a reasonable 20-25fps or so. Also I end up having to restart the game completely after dying, trying to load a game at that point just gives a black screen.
After a hardware upgrade and a native client it should be great, hardware upgrade for the speed and native client to get rid of those glitches;-)
Two 3ware 12-port adapters, 24 400GB drives, a double-width ATX case and you're almost there.
We had a bunch of these (well, 3*8-port and 80GB PATA it was back then) as el-cheapo RAID-0 disk servers in 2000. Losing a drive was no biggie, the boxes were part of a HSM solution so the master copy of the data was always on tape, these were just temporary caches.
And finished it a few days ago (amazon.co.uk has had it since the 1st)
Without spoiling the plot too much, I'd recommend it for fans of Quicksilver, even though unlike the original reviewer, I didn't think it was as good as the first book.
I really liked the "history" of modern science bits in Quicksilver, which were lacking in "The Confusion". (Economy and exotic locations being the theme for this book, which aren't as cool as science:-) ). Written as well as the first one, in the style one either hates or likes. I don't mind the extended descriptions as I'm a speed-reader when I decide to do so:-)
You can get 28 cpus (either xeon or PPC 970, mixing is no problem) in a 7U rack, which is pretty impressive. And you get some reasonably good remote management stuff (including a vnc based console so no need to drive up to the office when a machine fails to boot).
Your facts about what you get are true, but as for storage, the internal drives are pretty much meant for the basic OS, the drives are 2.5" ones intended the laptops that honestly are pretty slow by modern standards.
But that's why you get those $$$ SAN things (or NFS over GigE, which gets performance compareable to a local drive. Sure it's slower but not THAT much.)
I did a high-speed network protocol (Scheduled Transfer Protocol on GigE) demo a few years back, and to get something more interesting than a text console showing a transfer speed of 100MB/s (crappy 32-bit/33 MHz PCI buses and PC100 memory:( ) I did a quick&dirty video transmission thing.
What I used was just some random ~= 2Mbps video clip, but to get the bandwidth use up, I split the file into blocks of 1 MB, transmitted each block 500 times or so and the receiver chose a random copy of the block for playing the file back.
Avoids the slow disk problem yet lets you send "500 videos at the same time". It took something like 15 minutes to hack this feature into the network benchmark tool I was using:-)
But I did feel a bit stupid watching my school spend $160/box to roll out 16-bit Microsoft TCP/IP for WFWG 3.11 on all of the machines. A week after the order was made, a free 32-bit beta that worked quite a bit better was announced:-(
One of the worst personal mistakes is when I installed Linux 0.98 from the boot/rootdisk combination to my harddrive (previous versions I had to use from floppy due to missing support for my scsi adapter) I had a partition setup for it and everything. When the thing asked where I wanted to install, I answered/dev/sda instead of/dev/sda2:-( (where sda1 was my DOS partition with code I didn't want to lose of which I had no backups:( )
Of course, that episode made my transition to Linux as the primary OS a lot faster since I had to restart those projects from scratch and started using gcc instead of borland while I was at it:-)
Measure the weight of your wallet before and after? Price tends to correlate with taste (and thus size of bubbles)
Fortunately there are plenty of cheap alternatives to the over-priced (but excellent) authentic stuff. I like the Hungarian Torley (dry, of course). Almost as good as the low-end 20-25 euro champagnes for a fraction of the price.
Would be nice to have a really good excuse for buying a bottle of Dom Perignon, though;)
Next version of the Sun Ray server-side software will supposedly be available for Linux. Wonder if they'll continue the "for peanuts" educational site license with that one.
They're not bad boxes at all as long as you have a fast sparc around already. Otherwise not much sense unless someone reverse-engineers the protocol (as been partially done by someone AFAIK) or the Linux port materializes and you can get it for a reasonable price.
The normal price for the server side is like $1500/20 clients, so it really doesn't make much sense unless you think $350 (sun ray)+$75 (software)/box+$500 (decent monitor) is a reasonable price for fast but dumb terminals.
But they're very easy to maintain, just plug it in, insert a smart card and you're up & running. Remove smart card and move it to another box and your session moves there. Great when your office has less seats than workers:-)
Shame I don't have any mod points today, or I would mod the parent up. For those who are not into poetry, it's based on a very nice poem by Tennyson, the Charge of the Light Brigade (which is also the title of an excellent Outer Limits episode where mankind makes its last stand against aliens that plan on conquering Earth against impossible odds).
No. For the Microsoft campus licenses you still need to buy an OEM Windows license for every machine (around $100). This lets you install any version of Windows, Office and Visual Studio on the machine. If you don't have one, your "free" XP is as illegal as it is if you warezed it off the net.
Of course, your university pays something in the range of hundreds of thousands to millions of dollars / year for the contract, but since that's not out of your pocket (apart from tuition or taxes), it's obviously a really great deal every university should buy into!
Compare this to the $2500 site license that article mentions for RHEL, assuming that happens. I hope it will, that would give students a possibility to get a long-term supported OS for free, or if they feel like it suits them better, they can use Fedora or Debian (also for free).
"Unless Microsoft decides to make radical changes to their service model, a Windows-based version of Media Player will still remain a closed system, where users of any other operating systems cannot access content", said an anonymous computer user.
"Additionally, users of Windows Media Player are limited to using operating systems from Microsoft... this is a drawback for computer users, who expect choice in operating systems, choice in devices, and choice in what they want to do tomorrow, even it's burning music they've legally bought to a CD or put it on a portable OGG/MP3 player.
Lastly, if you use Windows along with a DRM-based system, you won't have the ability of using the several different free operating systems. When I'm getting software for free or a modest fee, I want to know that I have choices today and in the future.
IANAL, but naming your Linux distribution Hat is pretty close to the fine line.
Oh well, they'll probably call the final version something lame like "Sun Enterprise Linux Desktop" and the whole thing will be a non-issue, if it ever was to begin with:-)
One issue that might affect some people is the exec-shield anti-stack overflow technology, which most notably doesn't play will with wine and alsalib (latter might be fixed nowadays, alsalib used to use a gcc feature that made it place code on the stack).
You can easily disable it through/proc, or use a program called chstk to enable executable stacks for specific programs. It's not included in this beta, but you can grab it from here
Well, RHAS 2.1 is a server OS, so the GNOME desktop probably does suck. Haven't tried it myself, but it should be equivalent of RH 7.2, where the desktop was usable but far from polished (not that it's perfect even now).
The advanced workstation product they're planning should be interesting, I'd assume it'll be the same "7.2" (for RHAS compatibility) with the 8.x "Bluecurve" desktop for both KDE and GNOME.
For scrolling, try shift-pageup/pagedown (which does an entire page at a time), of course having support for scrolling a line at a time like in konsole might be useful in some situations.
Slashdot Mirror
Was unfortunately left out. I mean, Athlon 64 makes a fine Pentium 4 competitor when running a legacy 32-bit operating system, but it's so much more. Those cool extra registers you get in 64-bit mode make the thing just scream!
And no, the intel EM64T stuff isn't even competing in the same league, 40-45% slower with 40% more GHz is what I've seen in real-life workloads (heavy numbercrunching). For some other types of loads it does just about as well as the a64/opteron, though.
Revised x86_64 support (possibly in the pentium m core and in the same price range as the new 90nm a64's) and Intel has a chance. That and Microsoft delaying 64-bit Windows for a couple more years.
Has been "proven" for a long time. Every time they send something up there they seem to do the same story. Just google for water on mars and the first hit is a NASA press release from 2000.
:-)
And that's hardly the first "evidence", there's been plenty of evidence to show that there have been glaciers on Mars at some point. Or so I've been told by someone who studies Mars for a living
Although privatization (especially in Britain) often does not work, some financial incentive for efficiently utilizing the available frequencies would be useful.
In other words, if you're wasting say 5MHz of perfectly good bandwidth, and replacing all your kit with something modern doing the same in 0.5MHz would cost you X, it would be reasonable if you'd end up saving money if you did so (of course replacing all those millions of radios is not an option, just look at where DAB went).
HAM people already do this, sort of (not benefiting financially ofcourse, just in a everyone benefits sort of way). It pays off to use the very limited resources they get as efficiently as possible so they do so.
The problem is that it penalizes countries that started reducing their emissions _BEFORE_ 1990. Now if you were polluting like hell back then, reducing your emissions is easy...
This is actually what happened to Finland. Stupid politicians. It would have been totally fine to just ignore the treaty and just announce that we were unilaterally doing the equivalent thing except using the peak emission year of the 80's as the basis.
Indeed. And a64 really likes 64-bit code, even if it is a bit more bloated. Those extra registers really make things fly (unlike some other 64 bit archs like sparc64 where the biggest benefit is getting a 64-bit virtual address space and everything that doesn't need that typically gets compiled 32-bit because it's faster).
It's just that the proprietary operating system most hardware sites concentrate on doesn't have a non-beta 64-bit version (if you don't count that CPU that Linux has probably a 95% market share per sold cpu on, per box 2003 Server might do a bit better)
Well to be honest Anandtech is actually one of the few hardware sites that does have some kind of Linux benchmarks. Alas they don't "get" the open source way and just do "distro X is 2% faster than distro Y when running Z" style benchmarks when the approach should be to figure out why the differences are there. Sure it requires more effort and competence, but it's be a _lot_ more useful than the benchmarks they do now.
One of the Finnish election observers says he saw a touchscreen device break down (with 200 votes inside), and it was replaced and the officials were left pondering what to do with the broken one. Apparently it was one of those touch-screen no-paper-trail ones. Interesting to see if this is an isolated incident or not.
Drives from just about any vendor can suck. I know people who had 1000 IBM 75GXP drives (hey, their previous models were ok!) and had to swap a drive a day due to errors (not total failures tho, just early signs of it). Although many people seem to recommend "identical drives" for raid clusters for performance reasons or whatever, mixing vendors or at least production batches does the benefit of a single vendor fuckup resulting in data loss due to all of your drives having similar early deaths.
:-)
As for controllers, lots of big installations use 3ware cards. A bit more expensive than the basic ide/sata adapters, but you get lots of ports in one pci (preferably something better than 32-bit/33MHz) slot, and each drive gets their own ide bus. Now, even if the things have hardware raid5, you might be better off using plain old software raid with them. Depending on the card model it can easily be twice as fast. Embedded CPU's aren't that fast compared to a p4 or a64
Once you have 32 CPUs or so and have an OS that is able to run pretty well on those (Linux does, depending on the workload of course, people are
using Altixes with 256-512 in a single system image and are quite happy with them), being to able to replace cpus without rebooting starts getting pretty important (since occasionally they do fail).
See, if you didn't visit third party sites, just microsofts marketing pages so you'd know which of their products you should be buying next, you'd be totally safe...
Because they only support older models (pre-Radeon 9600 or so). While some work to add open source support to the R300 series is being done, it doesn't work yet. http://volodya-project.sourceforge.net/R300.php seems to be the site of the effort.
From what I've used the binary drivers, they're not _that_ hard to get running, on a friends fc2
laptop it was a matter of copying a few dri header files from the kernel sourcecode (the ATI drivers should be including a copy of these since there's no guarantee the kernel DRM stuff will continue
using it in the future)
Anyway, it was a matter of running make after that. Not as easy as the nvidia stuff, but still not that hard either.
I've played it that way on my athlon 2000+geforce 4 ti4200. It's quite playable, most notable problems are that it's _SLOW_ for about 15-20 secs
;-)
after loading a game. Going to the menus and using the PDA for a while and it goes back to a reasonable 20-25fps or so. Also I end up having to restart the game completely after dying, trying to load a game at that point just gives a black screen.
After a hardware upgrade and a native client it should be great, hardware upgrade for the speed
and native client to get rid of those glitches
Two 3ware 12-port adapters, 24 400GB drives, a double-width ATX case and you're almost there.
We had a bunch of these (well, 3*8-port and 80GB PATA it was back then) as el-cheapo RAID-0 disk servers in 2000. Losing a drive was no biggie, the boxes were part of a HSM solution so the master copy of the data was always on tape, these were just temporary caches.
And finished it a few days ago (amazon.co.uk has had it since the 1st)
:-) ). Written as well as the first one, in the style one either hates or likes. I don't mind the extended descriptions as I'm a speed-reader when I decide to do so :-)
Without spoiling the plot too much, I'd recommend it for fans of Quicksilver, even though unlike the original reviewer, I didn't think it was as good as the first book.
I really liked the "history" of modern science bits in Quicksilver, which were lacking in "The Confusion". (Economy and exotic locations being the theme for this book, which aren't as cool as science
Density and maintainability (but not price!)
You can get 28 cpus (either xeon or PPC 970, mixing is no problem) in a 7U rack, which is pretty impressive. And you get some reasonably good remote management stuff (including a vnc based console so no need to drive up to the office
when a machine fails to boot).
Your facts about what you get are true, but as for storage, the internal drives are pretty much meant for the basic OS, the drives are
2.5" ones intended the laptops that honestly are pretty slow by modern standards.
But that's why you get those $$$ SAN things (or NFS over GigE, which gets performance compareable to a local drive. Sure it's slower but not THAT much.)
I did a high-speed network protocol (Scheduled Transfer Protocol on GigE) demo a few years back, and to get something more interesting than a text console showing a transfer speed of 100MB/s (crappy 32-bit/33 MHz PCI buses and PC100 memory :( ) I did a quick&dirty video transmission thing.
:-)
What I used was just some random ~= 2Mbps video clip, but to get the bandwidth use up, I split the file into blocks of 1 MB, transmitted each block 500 times or so and the receiver chose a random copy of the block for playing the file back.
Avoids the slow disk problem yet lets you send "500 videos at the same time". It took something like 15 minutes to hack this feature into the network benchmark tool I was using
But I did feel a bit stupid watching my school spend $160/box to roll out 16-bit Microsoft TCP/IP for WFWG 3.11 on all of the machines. A week after the order was made, a free 32-bit beta that worked quite a bit better was announced :-(
/dev/sda instead of /dev/sda2 :-( (where sda1 was my DOS partition with code I didn't want to lose of which I had no backups :( )
:-)
One of the worst personal mistakes is when I installed Linux 0.98 from the boot/rootdisk combination to my harddrive (previous versions I had to use from floppy due to missing support for my scsi adapter) I had a partition setup for it and everything. When the thing asked where I wanted to install, I answered
Of course, that episode made my transition to Linux as the primary OS a lot faster since I had to restart those projects from scratch and started using gcc instead of borland while I was at it
Measure the weight of your wallet before and after? Price tends to correlate with taste (and thus size of bubbles)
;)
Fortunately there are plenty of cheap alternatives to the over-priced (but excellent) authentic stuff. I like the Hungarian Torley (dry, of course). Almost as good as the low-end 20-25 euro champagnes for a fraction of the price.
Would be nice to have a really good excuse for buying a bottle of Dom Perignon, though
Next version of the Sun Ray server-side software will supposedly be available for Linux. Wonder if they'll continue the "for peanuts" educational site license with that one.
/box+$500 (decent monitor) is a reasonable price for fast but dumb terminals.
:-)
They're not bad boxes at all as long as you have a fast sparc around already. Otherwise not much sense unless someone reverse-engineers the protocol (as been partially done by someone AFAIK) or the Linux port materializes and you can get it for a reasonable price.
The normal price for the server side is like $1500/20 clients, so it really doesn't make much sense unless you think $350 (sun ray)+$75 (software)
But they're very easy to maintain, just plug it in, insert a smart card and you're up & running.
Remove smart card and move it to another box and your session moves there. Great when your office
has less seats than workers
Shame I don't have any mod points today, or I would
mod the parent up. For those who are not
into poetry, it's based on a very nice poem by Tennyson, the Charge of the Light Brigade (which is also the title of an excellent Outer Limits episode where mankind makes its last stand against aliens that plan on conquering Earth against impossible odds).
No. For the Microsoft campus licenses you still need to buy an OEM Windows license for every machine (around $100). This lets you install
any version of Windows, Office and Visual Studio
on the machine. If you don't have one, your "free" XP is as illegal as it is if you warezed it off the net.
Of course, your university pays something in the range of hundreds of thousands to millions of
dollars / year for the contract, but since that's
not out of your pocket (apart from tuition or taxes), it's obviously a really great deal every university should buy into!
Compare this to the $2500 site license that article mentions for RHEL, assuming that happens. I hope it will, that would give students a possibility to get a long-term supported OS for free, or if they feel like it suits them better, they can use Fedora or Debian (also for free).
"Unless Microsoft decides to make radical changes to their service model, a Windows-based version of Media Player will still remain a closed system, where users of any other operating systems cannot access content", said an anonymous computer user.
"Additionally, users of Windows Media Player are limited to using operating systems from Microsoft... this is a drawback for computer users, who expect choice in operating systems, choice in devices, and choice in what they want to do tomorrow, even it's burning music they've legally bought to a CD or put it on a portable OGG/MP3 player.
Lastly, if you use Windows along with a DRM-based system, you won't have the ability of using the several different free operating systems. When I'm getting software for free or a modest fee, I want to know that I have choices today and in the future.
( Hat that is, I really should learn to use that preview button)
IANAL, but naming your Linux distribution Hat is pretty close to the fine line.
:-)
Oh well, they'll probably call the final version
something lame like "Sun Enterprise Linux Desktop"
and the whole thing will be a non-issue, if it ever was to begin with
Yes, should work just fine both ways
/proc, or use a program called chstk to enable executable stacks for specific programs. It's not included in this beta, but you can grab it from
One issue that might affect some people is the exec-shield anti-stack overflow technology, which
most notably doesn't play will with wine and alsalib (latter might be fixed nowadays, alsalib used to use a gcc feature that made it place code on the stack).
You can easily disable it through
here
Well, RHAS 2.1 is a server OS, so the GNOME desktop probably does suck. Haven't tried it myself, but it should be equivalent of RH 7.2, where the desktop was usable but far from polished (not that it's perfect even now).
The advanced workstation product they're planning should be interesting, I'd assume it'll be the same "7.2" (for RHAS compatibility) with the 8.x "Bluecurve" desktop for both KDE and GNOME.
For scrolling, try shift-pageup/pagedown (which does an entire page at a time), of course having support for scrolling a line at a time like in konsole might be useful in some situations.