This is great for targeted advertising towards their user base, but it's not clear that you can make many conclusions about the general market based on this data. After all, the user of Flickr are a self-selecting group. I don't think you'll find as many Canon 1Ds-Mk II cameras on there, as they're a ~$6800 body and the owners of those cameras are more likely to use them professionaly and have dedicated sites for their photos, rather than relying on Flickr to host.
They paid standard academic price for the original system. The take back and upgrade for $600K was a gift. I'm not saying you couldn't build something of similar performance for a simlar cost right now, but you have to agree that Apple has made a significant reduction in cost for VT on the upgrade. Standard academic pricing doesn't include a return and upgrade of used hardware months down the road for full value.
I'm not saying its not impressive or that someone couldn't doa similar thing with Apple ever again. All I'm saying is that there was consideration given by Apple to VT.
You have to take the costs with a grain of salt. They built the original machine for $5.2M. They then upgraded all the nodes from PowerMac G5s to Xserve G5s for $600K. Even if you assume that the $5.2 was a fair price for their original system, the upgrade price was an absolute gift from Apple. The cost per node to upgrade was about $550. Since they moved from non-ECC RAM to ECC RAM (4GB/node), the memory upgrade should have cost more than that alone.
Vendors will often give away hardware in order to break into a new market. This is incredible marketing for Apple. Who cares if they eat a few million for the press they've gotten?
Re:It's not just about speed and massively paralle
on
Linux Clustering
·
· Score: 2, Informative
There are certain classes of problems that clusters don't map to well. Applications with a very high cost of inter-processor comminucation or that demand a huge piece of contiguous memory are probably always going to be outside the realm of clusters.
However, problems that are embaressingly parallel can be handled by a cluster very adequately for a fraction of the cost of a traditional supercomputer. I don't know that you can ignore this class of problems and say that clusters aren't "true 'supercomputing'".
Yeah right, so you can't properly secure your own cd's or whatever, so go ahead and put a tax on internet access and cd burner's to make up losses because of your own incompetence.
It has nothing to do with properly securing CDs. It has to do with trying to force a business model that no longer works in the Internet age onto consumers.
but if countries like China find themselves at a disadvantage due to a handful of local spammers I would think they would be more motivated to deal with the problem.
Yeah, cause if there's one thing the Chinese government wants, it's an open experience on the Internet.
Bob Slydell: If you would, would you walk us through a typical day, for you?
Peter Gibbons: Yeah.
Bob Slydell: Great.
Peter Gibbons: Well, I generally come in at least fifteen minutes late, ah, I use the side door--that way Lumbergh can't see me, heh--after that I sorta space out for an hour.
Bob Porter: Da-uh? Space out?
Peter Gibbons: Yeah, I just stare at my desk, but it looks like I'm working. I do that for probably another hour after lunch too, I'd say in a given week I probably only do about fifteen minutes of real, actual, work.
Not to mention that most machines on the November 2002 list would have faster numbers now, if they re-ran with the Goto BLAS that came out after list enetries were due. Top 80 in the world is probably overstating things a bit.
128 nodes in the compute partition, plus an additional node (they don't say, but I would assume it's the front-end node - you would want your storage and management node on the high-speed interconnect, most likely).
So you have 128*2=256 procs connected on the high-speed interconnect, plus 1*2=2 procs for the other node. 256+2 = 258 procs. 1GB RAM/proc = 258GB for the entire system. Has nothing to do with 1000 versus 1024.
Hell, if the filter worked 90% of the time I'd use it at my ISP...
Yeah, me too. But that's only if it's 90% false negatives and a 0% false positives. I don't want my ISP filtering if I lose a legitimate message without any way of knowing...
I had a lot of classes I was very interested in during grad school, where the entire campus had WiFi access. I checked e-mail in all of them. I used AIM/ICQ as well. I found that I didn't have to choose between being interested/attentive/participatory and checking/sending e-mail. I think a lot of people can handle both, and for those who can, WiFi access is a blessing. Not only can you download presentations and lecture notes, but you can further investigate interesting problems in real-time. Also, timely communication (about meetings, deadlines, etc.) is a gift when you hit crunch time on projects or your thesis. That, and now true 24/7 access to pr0n - what could be better!
...that I patented the business model of getting generic patents, then going after large companies with lots of money. My patent covers both bullying them to settle out of court and having a technologically impaired judge award in my favor.
The whole idea of patents, AFAIK, is to grant a temporary monopoly for the patent holder, and thereby giving them a greater chance at a return for their invested R&D. This isn't a flaw of the system at all. Now how certain patents (software, etc.) are approved is another story.
Re:Oh that's very responsible of you, SlashDot
on
Microsoft News Update
·
· Score: 2, Funny
But to link directly to the crash-windows-in-one-easy-step binary? That's just plain irresponsible.
But, to be fair, they linked to a Windows executable. What self-respecting/. reader would stoop to running that code?
I'd argue that there is a design flaw here: a failed decryption should only return one bit of information, namely "decryption failed", and not provide a potential adversary with algorithm output.
How would you ever be able to detect a failed decryption? The ability to detect garbage requires a restriction on the valid cleartext input into the algorithm.
Actually going ahead with the demo c^Hwould give Microsoft a leg up in appeal of any unfavorable ruling.
When I first heard that CKK allowed the demo, I was woried that it could be seen as favoring the states over M$. I'm actually glad they dropped the demo, because it's one less thing for M$ to point to and cry foul.
Slashdot Mirror
This is great for targeted advertising towards their user base, but it's not clear that you can make many conclusions about the general market based on this data. After all, the user of Flickr are a self-selecting group. I don't think you'll find as many Canon 1Ds-Mk II cameras on there, as they're a ~$6800 body and the owners of those cameras are more likely to use them professionaly and have dedicated sites for their photos, rather than relying on Flickr to host.
Doesn't that assume that /. readers RTFA?
We all know that's not true.
They paid standard academic price for the original system. The take back and upgrade for $600K was a gift. I'm not saying you couldn't build something of similar performance for a simlar cost right now, but you have to agree that Apple has made a significant reduction in cost for VT on the upgrade. Standard academic pricing doesn't include a return and upgrade of used hardware months down the road for full value.
I'm not saying its not impressive or that someone couldn't doa similar thing with Apple ever again. All I'm saying is that there was consideration given by Apple to VT.
You have to take the costs with a grain of salt. They built the original machine for $5.2M. They then upgraded all the nodes from PowerMac G5s to Xserve G5s for $600K. Even if you assume that the $5.2 was a fair price for their original system, the upgrade price was an absolute gift from Apple. The cost per node to upgrade was about $550. Since they moved from non-ECC RAM to ECC RAM (4GB/node), the memory upgrade should have cost more than that alone.
Vendors will often give away hardware in order to break into a new market. This is incredible marketing for Apple. Who cares if they eat a few million for the press they've gotten?
There are certain classes of problems that clusters don't map to well. Applications with a very high cost of inter-processor comminucation or that demand a huge piece of contiguous memory are probably always going to be outside the realm of clusters.
However, problems that are embaressingly parallel can be handled by a cluster very adequately for a fraction of the cost of a traditional supercomputer. I don't know that you can ignore this class of problems and say that clusters aren't "true 'supercomputing'".
A good compliment to my current laptop battery which, at 3 years of age, discharges in under 30 seconds.
but what can you overclock it to?
Why is it that nowadays, any new cool thing is invented either for military or advertising use?
You forgot the porn industry... How much inovation on the web has come from them?
Think of it this way...
If you pick the correct door the first time (1/3 chance), and he shows you another door, and you switch, you lose.
If you don't pick the right door the first time (2/3 chance), he shows you another loser, and you switch, you win.
So, with the strategy to switch after he shows you another loser, you will win 2/3 of the time.
Yeah right, so you can't properly secure your own cd's or whatever, so go ahead and put a tax on internet access and cd burner's to make up losses because of your own incompetence.
It has nothing to do with properly securing CDs. It has to do with trying to force a business model that no longer works in the Internet age onto consumers.
but if countries like China find themselves at a disadvantage due to a handful of local spammers I would think they would be more motivated to deal with the problem.
Yeah, cause if there's one thing the Chinese government wants, it's an open experience on the Internet.
Bob Slydell: If you would, would you walk us through a typical day, for you?
Peter Gibbons: Yeah.
Bob Slydell: Great.
Peter Gibbons: Well, I generally come in at least fifteen minutes late, ah, I use the side door--that way Lumbergh can't see me, heh--after that I sorta space out for an hour.
Bob Porter: Da-uh? Space out?
Peter Gibbons: Yeah, I just stare at my desk, but it looks like I'm working. I do that for probably another hour after lunch too, I'd say in a given week I probably only do about fifteen minutes of real, actual, work.
Not to mention that most machines on the November 2002 list would have faster numbers now, if they re-ran with the Goto BLAS that came out after list enetries were due. Top 80 in the world is probably overstating things a bit.
Goto site is here
Article about the effects is here
128 nodes in the compute partition, plus an additional node (they don't say, but I would assume it's the front-end node - you would want your storage and management node on the high-speed interconnect, most likely).
So you have 128*2=256 procs connected on the high-speed interconnect, plus 1*2=2 procs for the other node. 256+2 = 258 procs. 1GB RAM/proc = 258GB for the entire system. Has nothing to do with 1000 versus 1024.
Hell, if the filter worked 90% of the time I'd use it at my ISP...
Yeah, me too. But that's only if it's 90% false negatives and a 0% false positives. I don't want my ISP filtering if I lose a legitimate message without any way of knowing...
I had a lot of classes I was very interested in during grad school, where the entire campus had WiFi access. I checked e-mail in all of them. I used AIM/ICQ as well. I found that I didn't have to choose between being interested/attentive/participatory and checking/sending e-mail. I think a lot of people can handle both, and for those who can, WiFi access is a blessing. Not only can you download presentations and lecture notes, but you can further investigate interesting problems in real-time. Also, timely communication (about meetings, deadlines, etc.) is a gift when you hit crunch time on projects or your thesis. That, and now true 24/7 access to pr0n - what could be better!
...that I patented the business model of getting generic patents, then going after large companies with lots of money. My patent covers both bullying them to settle out of court and having a technologically impaired judge award in my favor.
I'm going to sue the pants off of Woolston!
Patents as anti-competitive instrument
The whole idea of patents, AFAIK, is to grant a temporary monopoly for the patent holder, and thereby giving them a greater chance at a return for their invested R&D. This isn't a flaw of the system at all. Now how certain patents (software, etc.) are approved is another story.
But to link directly to the crash-windows-in-one-easy-step binary? That's just plain irresponsible.
/. reader would stoop to running that code?
But, to be fair, they linked to a Windows executable. What self-respecting
I'd argue that there is a design flaw here: a failed decryption should only return one bit of information, namely "decryption failed", and not provide a potential adversary with algorithm output.
How would you ever be able to detect a failed decryption? The ability to detect garbage requires a restriction on the valid cleartext input into the algorithm.
Actually going ahead with the demo c^Hwould give Microsoft a leg up in appeal of any unfavorable ruling.
When I first heard that CKK allowed the demo, I was woried that it could be seen as favoring the states over M$. I'm actually glad they dropped the demo, because it's one less thing for M$ to point to and cry foul.