Yeah, but you'll have a hard time getting a Transmeta cluster to do 3+ TFLOPS sustained, 12 TFLOP peaks... you'd need so many more of them, it wouldn't save you very much. The actual processors are fairly good on the FLOP/Watt basis (much better than x86), but you get 8000 of them, and all of the boards, switching fabric and everything else (including cooling), and you are going to draw a little power here and there...
And here I was worried that I just installed A/C for my house...
Of course, depending on the number of procs, you get different results with the different -j options make -j2 zImage make -j4 zImage make -j zImage etc...
An AC posted below that the Hitachi/Fujitsu machines are vector processors, like the good ol' Crays. Very good at some things, not so great at others...
Favorite fun fact: The microprocessors inside ASCI White contain 2,000 miles of copper wiring, enough to stretch from Washington, D.C. to Phoenix, Arizona.
This compter isn't on the list yet... ASCI Red is... ASCI White is about three times faster than ASCII Blue - with a few less procs than ASCI Red (top of the list). I can't wait to see the linpack results, though...
I'm not sure what types of uProcs the Fujitsu and Hitachi systems use, but I remember reading that the number of nodes is fairly comprable between the IBM machines and the other maufacturers...
See this post for a more reasonable thought. It's probably not Nike (not that I care for them, but really, they are as much the victim here as that other ISP).
Maybe they should have just listed the floor square footage or cubic feet (or centiliters) for volume. There ya go, a tough to grasp representation. I'll admit the elephant thing is a little dumb (as is the calculator analogy), but the basketball court analogy seems to be right on target with giving a semi-useful description...
Of course, they could have just said, "It's way bigger than a regular PC" but that wouldn't have helped;-)
"If anyone screwed up, said Casler, it was Network Solutions, which apparently allowed the hijacker to change Nike's registry information on the basis of a spoofed email from the Nike billing contact -- a person that did not have password authority to make changes to Nike's domain status."
Yeah, everyone knows that they are a bunch of swindling, boorish jerks. We've heard it before, we'll hear it again...
On a more realistic note, I don't think that Nike can/should be held repsonsible, if in fact, NSI made a change due to an email from an unauthorized account (the billing contact). More details need to be seen on this one - still not good, whatever happened...
Since you've posted this same comment on both threads (no, I don't know why they posted it twice, either)...
The point is more the fact that greater leaps are being made in computing and raw power. I'm not going to touch any of the political aspects of any of this, but the object itself is deserving of praise, though the reasoning behind its present use may be questionable. Some would claim otherwise, but you can't blame the hardware for what it does. It is still just a computer. A damned fast one at that.
By the way, the Constitution doesn't guarantee a whole lot of anything, certainly not health care, food, or material wealth. Of course, you probably haven't read it, judging by that statement, but you don't live here, either.
Also, the money wouldn't have been budgeted in the first place for welfare, etc., it would have just been cut and redirected to another branch of the government - this project doesn't directly contribute the way you seem to indicate. People, not governments, are responsible for taking care of other people.
*That's* where I've seen jar-jar before... those psychedelic Camel bilboard ads... lots of colors, but just a shadow in the middle... I never noticed the resemblence before.
any real hardware router (Cisco, etc) will beat the pants off of a software implementation (Linux/BSD/W2k box) for total throughput. For small networks (like my home net), the linux box does a great job of routing/NAT/firewalling/e-mail/webserverving, but if you are going to have heavy traffic, your dollars are best spent elsewhere. If it is just a few (or even a few dozen boxes), you'll probably be fine with a PC implementation.
A hardware router still runs firmware, but the internal bus structure and routing is far more expedient than a PC trying the same tasks. Since real routing happens a couple layers up on the OSI stack, the frames have to be translated and passed through the driver, and in some cases all the way to user space to be filtered. Granted, routing can be in kernel space, but this is still significantly slower (especially considering PCI and uProc bus latency) than a streamlined piece of specific hardware.
netcard1 -> PCI -> mem.... uProc... mem -> PCI -> netcard2 is a long, slow path.
If this is for a buisness, get some real hardware. If this is for your home network, you could have one of the other boxes do it.
Of course, if you are only routing a 28.8 dial-up, you could probably use bicycle messengers, too... the latency isn't *that* much of a problem. For a real pipe, get a real piece of silicon.
Is that from that claymation show (don't know the name) of the people with the extremely wide mouths and oversized hands? I watched it once or twice... strange... now they've made a chicken/turkey movie out of it or something?
Crazy, and not that good.... but the only british comedy I ever liked was MP anyway... most of the british sitcoms are even worse than "Three's Company"
Last night I just updated my Mandrake 6.1 to 7.1... the CD install program *will* let you selectively update packages. Takes a little time (and the GUI has a nasty little feature that makes it hard to tell which packages are selected in tree view), but it all worked out real well... the only thing I had to do afterward (because I wanted to) was to get the newest kernel and reapply the VPN-MASQ patch (if you don't know, you don't need it).
I've been rather happy with Mandrake for a long while now, though I have used FreeBSD on and off for a couple of years now - a great system, too... that's why I have multiple boxen 8^)
Actually, IBM could have done better with OS/2 by providing better SDK type support to developers and making it more atractive as a platform. The Win32 support was/is pretty solid, but marketing an OS as 'it can run this other OS's stuff just fine' just didn't work out. Better multi-tasking, memory protection, etc... Just late and poorly marketed...
Um... there would be a license in each source file... at least, I'd hope they'd be that organized. The top header should have something like (and pardon the spacing):
/* (C) COPYRIGHT FOO CORP. 2000 */ /* */ /* The source code for this program is not published or otherwise */ /* divested of its trade secrets, irrespective of what has been */ /* deposited with the U.S. Copyright Office. */ /* */ /* End Copyright ************************************************** ****/
Slashdot Mirror
Yeah, but you'll have a hard time getting a Transmeta cluster to do 3+ TFLOPS sustained, 12 TFLOP peaks... you'd need so many more of them, it wouldn't save you very much. The actual processors are fairly good on the FLOP/Watt basis (much better than x86), but you get 8000 of them, and all of the boards, switching fabric and everything else (including cooling), and you are going to draw a little power here and there...
And here I was worried that I just installed A/C for my house...
Of course, depending on the number of procs, you get different results with the different -j options
make -j2 zImage
make -j4 zImage
make -j zImage
etc...
IBM's ASCI
Draws 1 Point 2 Megawatts
The West Coast Goes Dim
Yup, and I just posted more comments on that one, too... D'Oh!
Guess it's just gone from the front page...
Nike Rerouted
ISP is Hopping Mad
NSI to Blame
where is this box that you are stepping out of? Is it the same box that people think outside of, or is it a different color?
An AC posted below that the Hitachi/Fujitsu machines are vector processors, like the good ol' Crays. Very good at some things, not so great at others...
Favorite fun fact:
The microprocessors inside ASCI White contain 2,000 miles of copper wiring, enough to stretch from Washington, D.C. to Phoenix, Arizona.
Maybe with the 28 semis that it took to get it from IBM to Sandia... Does Santa still run that freight line?
This compter isn't on the list yet... ASCI Red is... ASCI White is about three times faster than ASCII Blue - with a few less procs than ASCI Red (top of the list). I can't wait to see the linpack results, though...
I'm not sure what types of uProcs the Fujitsu and Hitachi systems use, but I remember reading that the number of nodes is fairly comprable between the IBM machines and the other maufacturers...
See this post for a more reasonable thought. It's probably not Nike (not that I care for them, but really, they are as much the victim here as that other ISP).
Maybe they should have just listed the floor square footage or cubic feet (or centiliters) for volume. There ya go, a tough to grasp representation. I'll admit the elephant thing is a little dumb (as is the calculator analogy), but the basketball court analogy seems to be right on target with giving a semi-useful description...
;-)
Of course, they could have just said, "It's way bigger than a regular PC" but that wouldn't have helped
sue the ass off of Network Solutions!
"If anyone screwed up, said Casler, it was Network Solutions, which apparently allowed the hijacker to change Nike's registry information on the basis of a spoofed email from the Nike billing contact -- a person that did not have password authority to make changes to Nike's domain status."
Yeah, everyone knows that they are a bunch of swindling, boorish jerks. We've heard it before, we'll hear it again...
On a more realistic note, I don't think that Nike can/should be held repsonsible, if in fact, NSI made a change due to an email from an unauthorized account (the billing contact). More details need to be seen on this one - still not good, whatever happened...
Since you've posted this same comment on both threads (no, I don't know why they posted it twice, either)...
The point is more the fact that greater leaps are being made in computing and raw power. I'm not going to touch any of the political aspects of any of this, but the object itself is deserving of praise, though the reasoning behind its present use may be questionable. Some would claim otherwise, but you can't blame the hardware for what it does. It is still just a computer. A damned fast one at that.
By the way, the Constitution doesn't guarantee a whole lot of anything, certainly not health care, food, or material wealth. Of course, you probably haven't read it, judging by that statement, but you don't live here, either.
Also, the money wouldn't have been budgeted in the first place for welfare, etc., it would have just been cut and redirected to another branch of the government - this project doesn't directly contribute the way you seem to indicate. People, not governments, are responsible for taking care of other people.
Too bad the Pentium (original) and the PPro are generally referred to as the P5 and P6... what next then?
*That's* where I've seen jar-jar before... those psychedelic Camel bilboard ads... lots of colors, but just a shadow in the middle... I never noticed the resemblence before.
Save us all...
My understanding was that most of Neo's next battles were back in meatspace, rather than in the Matrix, so he's not all powerful there...
Have Natalie drown him in Hot Grits ;-)
Sorry.... couldn't resist - maybe just an 'accidental' swing of Samuel L. Jackson's 'saber... yeah, that'd work.
any real hardware router (Cisco, etc) will beat the pants off of a software implementation (Linux/BSD/W2k box) for total throughput. For small networks (like my home net), the linux box does a great job of routing/NAT/firewalling/e-mail/webserverving, but if you are going to have heavy traffic, your dollars are best spent elsewhere. If it is just a few (or even a few dozen boxes), you'll probably be fine with a PC implementation.
... mem -> PCI -> netcard2 is a long, slow path.
A hardware router still runs firmware, but the internal bus structure and routing is far more expedient than a PC trying the same tasks. Since real routing happens a couple layers up on the OSI stack, the frames have to be translated and passed through the driver, and in some cases all the way to user space to be filtered. Granted, routing can be in kernel space, but this is still significantly slower (especially considering PCI and uProc bus latency) than a streamlined piece of specific hardware.
netcard1 -> PCI -> mem.... uProc
If this is for a buisness, get some real hardware. If this is for your home network, you could have one of the other boxes do it.
Of course, if you are only routing a 28.8 dial-up, you could probably use bicycle messengers, too... the latency isn't *that* much of a problem. For a real pipe, get a real piece of silicon.
... the safer than Mir line put a smile on *my* face... that's funny.
There will always be good and bad examples in any group... and good and bad examples for an individual.
Woz is a shining example, though. More beloved by the masses and seemingly less abrasive than the 'other' Steve (Jobs) or BillG.
Is that from that claymation show (don't know the name) of the people with the extremely wide mouths and oversized hands? I watched it once or twice... strange... now they've made a chicken/turkey movie out of it or something?
Crazy, and not that good.... but the only british comedy I ever liked was MP anyway... most of the british sitcoms are even worse than "Three's Company"
Nah, 'bicksp' is a perfectly pronouncable word... 'cept fer mebbe that last leetle bit... o vell...
Last night I just updated my Mandrake 6.1 to 7.1... the CD install program *will* let you selectively update packages. Takes a little time (and the GUI has a nasty little feature that makes it hard to tell which packages are selected in tree view), but it all worked out real well... the only thing I had to do afterward (because I wanted to) was to get the newest kernel and reapply the VPN-MASQ patch (if you don't know, you don't need it).
I've been rather happy with Mandrake for a long while now, though I have used FreeBSD on and off for a couple of years now - a great system, too...
that's why I have multiple boxen 8^)
Actually, IBM could have done better with OS/2 by providing better SDK type support to developers and making it more atractive as a platform. The Win32 support was/is pretty solid, but marketing an OS as 'it can run this other OS's stuff just fine' just didn't work out. Better multi-tasking, memory protection, etc... Just late and poorly marketed...
Um... there would be a license in each source file... at least, I'd hope they'd be that organized. The top header should have something like (and pardon the spacing):
* ********/
* ****/
/* Copyright *************************************************
/* */
/* File stuff.h */
/* */
/* Confidential */
/* (C) COPYRIGHT FOO CORP. 2000 */
/* */
/* The source code for this program is not published or otherwise */
/* divested of its trade secrets, irrespective of what has been */
/* deposited with the U.S. Copyright Office. */
/* */
/* End Copyright *************************************************