In addition to this, until we start seeing widespread use of PCIe, the downstream AGP bus is still a serious bottleneck as well.
I thought this was an AGP driver issue, not a hardware limitation? Who knows, maybe the same guy will write the PCIe driver and history will repeat itself...
Re:Actually, Gentoo releases don't really matter..
on
Gentoo 2004.2 Released
·
· Score: 1
Another alternative is anything that uses apt-get...e.g. debian, fedora, yellow dog. You can make the update process as automatic as you like, without the time it takes to compile from source.
We use apt-get on about 10 FC2 boxes to update everything but the kernel nightly, with emails sent every other night with its status.
I understand that Google has the ability to add nodes quickly, but my understanding is that the never remove nodes. They are left dead on the floor (or wherever...).
Heck, you could even provide a default motd that would give some cursory (no pun intended) help on using the command line; maybe even a URL to some helpful "first time user" pages.
It's not difficult to make the existing tools user friendly.
Let's assume that there was a default icon for the terminal on the panel. Let's assume an unexperienced user clicked on that button. What would their response be?
I can't imagine it would be shock and horror. It's not like the terminal launched with an accompanying blood curdling yell.
Most likely they would click the window closed, and go click another icon til they found one that did something they wanted.
It's not that I am not smart enough to figure out how to get one (but thanks for the implied insult, I needed that), it's that most distros don't aim to please their current user base. They are trying too hard to find that "mythical desktop user".
It seems to me that I could install a functional Linux distibution on the same hardware that the Be geniuses said they were "locked" out of...I guess they just couldn't embrace open source to look at those GPL'd drivers.
Be just seemed to whine rather than get on with the business of doing business. Great ideas, crappy leadership.
If you can write OpenMP/MPI code that runs on the cluster, then yes, the cluster of P4's would probably get you better performance. But the way to code is written, it's horribly non-parallelizable...and of course, rewriting it takes money (or time, and they are equivalent).
If you can't and just need speed from serial code, then the Itanium might be the better choice.
We have an 8 way SGI Altix 350 down the hall. I didn't know I would be in such refined company.
And they can't run 32 bit code natively, it must be done in emulation.
Intel's original purpose was to put these chips in servers and high end workstations, places where 32-bit compatibility was not its main purpose.
For CFD on the codes that I run (full Navier-Stokes equations with LES), an Itanium2 at 1.5 Ghz is twice as fast as a 3.0Ghz P4. (According to SGI these are $10k/chip, the lowly 1.4 Ghz are $3k/chip)
1) This is not the unix way of doing things. Small individual apps that can be combined in powerful ways.
2) What is this "internet" thing you talk about? To me it's a moving target. It does the "big three" (browse, mail, chat). What about streaming MP3s? How about P2P? How about unknown protocol-X? I like mozilla as much as the next person (typing in it now...), but the goal is overstated.
Mozilla should break into separate apps to handle separate tasks.
Then the desktops should provide a standard way of providing inter-app communication (is that what message bus is attempting?), so that clicking on a link in my e-mail client of choice it sends it to my browser of choice...
The real problem is that the Windows OS is largely inseparable from its GUI and, as it currently stands, is way too bloated to run individual HPC nodes efficiently and effectively.
I do CFD for a living. When I started my new position a couple of years back, I convinced my boss to move to Linux because (Linux + ifc) was 50% faster than (Win2k + visual fortran) for the single processor codes we were running.
I can see Microsoft trying to pare that difference down, but it will still be prohibitive when coupled with licensing costs.
The argument against GM foods goes along the same lines as the argument that global warming is bad: while they both maybe on shaky scientific basis, are you sure that you want to risk the consequences?
(People are starving generally because of political problems and distribution issues, not actual crop yield.)
(Consider this: if you thought that one act could possibly (albeit unlikely, but nevertheless finite probablility) ignite the atmosphere of the earth, thereby killing everyone, would you do it? That was the choice faced by the Manhattan Project, and yet they still touched off the first atomic bomb. Yeesh...)
Can you conclude that "Gentoo is faster than Mandrake?" No. This is a limited test. It is likely that Mandrake is faster for some things. Also, we tested load-time performance only.
Slashdot Mirror
In addition to this, until we start seeing widespread use of PCIe, the downstream AGP bus is still a serious bottleneck as well.
I thought this was an AGP driver issue, not a hardware limitation? Who knows, maybe the same guy will write the PCIe driver and history will repeat itself...
Another alternative is anything that uses apt-get...e.g. debian, fedora, yellow dog. You can make the update process as automatic as you like, without the time it takes to compile from source.
We use apt-get on about 10 FC2 boxes to update everything but the kernel nightly, with emails sent every other night with its status.
If disposable income correlates with income, and income correlates with intelligence, then you may be correct.
<troll>
I see you are a PC user, so I will translate: smart people make more money and buy macs...
</troll>
...ah yes, just build a better mousetrap.
I think that you just advocated DRM for hardware.
If I own it, and want mess with it I should be able to (just not under warranty).
I understand that Google has the ability to add nodes quickly, but my understanding is that the never remove nodes. They are left dead on the floor (or wherever...).
I-rate with E-rate?
Thanks, I'll be here all week.
I wish my ISP would mirror porn locally too...
;)
Ever heard of usenet?
Heck, you could even provide a default motd that would give some cursory (no pun intended) help on using the command line; maybe even a URL to some helpful "first time user" pages.
It's not difficult to make the existing tools user friendly.
Let's assume that there was a default icon for the terminal on the panel. Let's assume an unexperienced user clicked on that button. What would their response be?
I can't imagine it would be shock and horror. It's not like the terminal launched with an accompanying blood curdling yell.
Most likely they would click the window closed, and go click another icon til they found one that did something they wanted.
It's not that I am not smart enough to figure out how to get one (but thanks for the implied insult, I needed that), it's that most distros don't aim to please their current user base. They are trying too hard to find that "mythical desktop user".
Which being such an intuitive key stroke to hit, only makes the parent's point.
The shell still is an integral part of the desktop on linux. Why there can't be a button on the panel/kicker by default is beyond me.
Hmmm...a fix for the fix. Interesting.
How nested can that go?
Uhh, yeah...
It seems to me that I could install a functional Linux distibution on the same hardware that the Be geniuses said they were "locked" out of...I guess they just couldn't embrace open source to look at those GPL'd drivers.
Be just seemed to whine rather than get on with the business of doing business. Great ideas, crappy leadership.
Define cost effective.
If you can write OpenMP/MPI code that runs on the cluster, then yes, the cluster of P4's would probably get you better performance. But the way to code is written, it's horribly non-parallelizable...and of course, rewriting it takes money (or time, and they are equivalent).
If you can't and just need speed from serial code, then the Itanium might be the better choice.
...probably about 3 people have them.
We have an 8 way SGI Altix 350 down the hall. I didn't know I would be in such refined company.
And they can't run 32 bit code natively, it must be done in emulation.
Intel's original purpose was to put these chips in servers and high end workstations, places where 32-bit compatibility was not its main purpose.
For CFD on the codes that I run (full Navier-Stokes equations with LES), an Itanium2 at 1.5 Ghz is twice as fast as a 3.0Ghz P4. (According to SGI these are $10k/chip, the lowly 1.4 Ghz are $3k/chip)
Natural gas is better, but also somewhat finite.
;-)
Not after a meal of burritos...
Sorry for the pithy comment
Mozilla is a platform for internet.
1) This is not the unix way of doing things. Small individual apps that can be combined in powerful ways.
2) What is this "internet" thing you talk about? To me it's a moving target. It does the "big three" (browse, mail, chat). What about streaming MP3s? How about P2P? How about unknown protocol-X? I like mozilla as much as the next person (typing in it now...), but the goal is overstated.
Mozilla should break into separate apps to handle separate tasks.
Then the desktops should provide a standard way of providing inter-app communication (is that what message bus is attempting?), so that clicking on a link in my e-mail client of choice it sends it to my browser of choice...
After seeing the requirements for a single processor version of Longhorn, I can only imagine what the HPC version would require.
The real problem is that the Windows OS is largely inseparable from its GUI and, as it currently stands, is way too bloated to run individual HPC nodes efficiently and effectively.
I do CFD for a living. When I started my new position a couple of years back, I convinced my boss to move to Linux because (Linux + ifc) was 50% faster than (Win2k + visual fortran) for the single processor codes we were running.
I can see Microsoft trying to pare that difference down, but it will still be prohibitive when coupled with licensing costs.
hell, we've been doing it (albeit slower) for thousands of years!
We are splicing in genes from different species. That is one hell of a long way from selective breeding.
The argument against GM foods goes along the same lines as the argument that global warming is bad: while they both maybe on shaky scientific basis, are you sure that you want to risk the consequences?
(People are starving generally because of political problems and distribution issues, not actual crop yield.)
(Consider this: if you thought that one act could possibly (albeit unlikely, but nevertheless finite probablility) ignite the atmosphere of the earth, thereby killing everyone, would you do it? That was the choice faced by the Manhattan Project, and yet they still touched off the first atomic bomb. Yeesh...)
My textbooks still stay that Americans value freedom and free speech more than Canadians...
You sir, are correct. We Americans are prepared to pay a higher price for freedom than we are for Canandians. They smell like bacon.
There are a couple of other good repositories that have FC2 packages out now .
Try this to fix the unneeded space in 'hints' above...
Knoppix does it by using cramfs...
From the page that you linked to:
Can you conclude that "Gentoo is faster than Mandrake?"
No. This is a limited test. It is likely that Mandrake is faster for some things. Also, we tested load-time performance only.