The idea that extinctions are periodic, and even that this is due to motion relative to the galactic plane, has been around at least since the late 80s - I remember this coming up from back then. You're just looking at the latest incarnation of an old idea.
OK, I've only read the summary (I have a lecture to give half an hour from now to prepare for) but I can see some objections:
* My boss (David Penny, Massey University) argues that the mammals and birds were already outcompeting the dinosaurs at the end of the cretaceous, so the asteroid was at best a coup-de-grace for them. * The "periodic extinctions" idea has been around for decades, including the possible link to oscillations through the galactic plane. * Mass extinctions are sudden. The increase in extragalactic cosmic rays exposure would be slow, over millions of years. * The extragalactic cosmic ray exposure changes should be highly regular. The extinctions are irregular.
Good luck in court arguing that you shouldn't be liable for more than $5. If that were the case, nobody would ever pay for anything - either they don't get caught (and get it free) or they do get caught, and have to pay what they would have paid anyway.
If you steal the twinkies in the knowledge that there is only a 1% chance you'll be caught, then being charged $500 if you do get caught seems fair to me (although I don't think the law allows this argument. IANAL.)
So (interpolating a little) current high-end cards are bottlenecked by PCI bandwidth at about x12 or less. This means that the next generation high-end can be expected use upcoming PCI-e 2.0 (twice the speed per channel) otherwise they'll bottleneck on x16.
I'm having troubles determining how the cards in the old review (6800GT etc, which do fine on x4) compare to the current generation, but I'm guessing that the current mid-range (e.g. 8500, 7600) are comparable, and could therefore run at x4 without a problem.
My motherboard has great big old PCI slots, and tiny little 1xPCI-e slots which are just as capable. PCI-e has taken over for graphics cards, but I've never even seen a 1xPCI-e expansion card. (The motherboard manufacturers don't believe they'll be used either - they put them next to the 16x slot where double-width graphics cards will make them inaccessable.)
When will old PCI die? Perhaps very small format motherboards and laptops will eventually drive demand for 1xPCI-e cards?
For that matter - is there any reason for low-end PCI-e graphics cards to be 16x, rather than 8x or even 4x? (They'd still fit in a 16x slot.) I suppose there is no demand - any PCI-e motherboard has a 16x slot, and there isn't anything you'd want to put in it except a GPU. About the only use I can think of is if you wanted one computer to run many low-performance displays - e.g. 8 monitors off four GPUs, each using a 4x slot.
At a company meeting we had an "inspirational speaker" who was a Olympic rowing gold medalist. He described what they ate during training - it was about 5-6 *huge* meals per day, including one at about 2am. It wouldn't surprise me at all if it was 33,000 kCal per day or more. This was required to supply the energy they expended during training.
I don't really know what happened to multi-node jobs, as I never submitted one (except during a training workshop, when all other work had been cleared from the cluster) and rarely if ever noticed them on the queue. It was the case that the queues were always full of single node jobs. I haven't been a heavy user - there were about 3 occasions where I had number crunching which would have taken 2-4 weeks on my desktop computer which I could break up and submit to Helix and get it all back in a few days.
I think in general it hasn't been too much of a problem - everyone wants to run a bunch of single node jobs anyhow.
In general, however, I see big problems if you can't migrate jobs. Say somebody wants 32 CPUs. You can stop running new jobs until 32 CPUs are free (which could mean 31 CPUs sit idle for a week waiting). You could suspend the jobs already running on 32 CPUs and resume them after the 32 CPU job is complete (which could cause a 1 hour job 5 minutes short of completion to be delayed for a week by the big job.) You could reserve half the CPUs to only be available for 32 CPU jobs (which could lead to half the cluster being idle while there is a huge backlock of single CPU jobs.) You could start the 32 CPU job running concurrantly with the pre-existing jobs on 32 nodes. Then if the single CPU jobs on 31 of the nodes finish promptly but one has a long-running resource-hogging job, you end up with 31 CPUs idleing most of the time waiting for the one shared CPU to catch up.
I think that covers the possibilities. If you can't predict in advance how long jobs will take, you can't schedule jobs with different numbers of nodes efficiently.
I've used a 64 node Beowulf cluster on occasion. The queue was generally full of single-node jobs and multi-node jobs hardly ever ran. There really is no good way of scheduling in this environment without the ability to suspend a job on one node and later resume it on another. So far as I know, our scheduling software was not capable of this. (Fortunately, I was just submitting single node jobs.)
Starting with the information in the summary, I spent a few minutes web searching. "bproc" appears not to be capable of this: it just means that your primary node can "see" the processes running on other nodes, so you can use 'ps', 'kill' etc. on them. However OpenSSI has "process migration". Is this the ability to move processes from one node to another?
Here's the essence of what he has to say: "I don't think it was possible to keep the book under wraps." "There are simply too many people who must be trusted in order for the security to hold." "My guess is that the publishers will lose zero sales"
You're assuming they get to bill 40 hours per week, and have no costs. They have to spend some of their time on non-billable running-the-business work, possibly pay secretaries and legal researchers, rent, malpractise insurance, attending conferences...
IANAL, so I don't know what the overheads are like, but they'll be a non-negligable fraction of that $400,000.
There is an excellent novel "Passage" by Connie Willis involving scientists researching near death experiences (NDEs). (This is fiction about science, rather than Science Fiction, although she writes that too.) They're also doing battle with a crackpot researcher who vigorously prompts patients into "remembering" angels etc etc.
I think that 500W figure came off the top of someone's head. I suspect the described system would actually idle at about 200W and peak at around 500W.
PSU (power supply unit) capacity is being way oversold. A desktop PC just isn't going to break 300W peak unless it is a hard-core gaming machine. Even a decent gaming machine (fast CPU and a single nearly-top-of-the-line GPU) won't break 300W. See here for examples of what 300W will run. (The thread started ~4 years ago, so you might want to skip to the end.)
Here's an example:
Intel Pentium D950, overclocked / overvolted 10% (Presler, 3.74GHz) ATI Radeon X1950XTX-512 PCIe graphics Asus P5LD2-VM motherboard 4 x 1024MB Corsair DDR2-6400 RAM Hitachi Deskstar 7K80 80GB hard drive WD Raptor WD1500ADFD 150GB 10krpm HDD 1 x 120mm fan
Total AC power draw is an absolute max of 298W; estimated DC power draw is 256W. _
Myself:
No TV, never owned one. I've been pretty much TV free for almost 20 years. (I'm occasionally tempted by a flatmate's TV.)
No cell phone. I had one for a couple of years (an old hand-me-down) which was only used or carried occasionally, but a few months ago they shut down the network it ran on.
No digital music player.
I've had a car for ~5 years now, but I've had many years car-free. Going back would be hard. When car-free, I lived in a bigger city* with better public transport and worse traffic
Dishwasher and microwave: I've got them - I'm lazy. I don't have a lawnmower - because I pay someone to do the lawn.
Computer would be hard to live without. I also use it as a substitute TV, for watching DVDs (mostly from the library.)
Lots and lots of books. Sometimes I think they own the house more than I do.
* Was Auckland (pop ~1 million), now Palmerston North (~75,000.)
While "most worthless story ever" is a huge overstatement, the criticism is justified. I knew that Intel were pushing flash cache on the motherboard, but I didn't know (or forgot) that they called it "turbo memory." It would only have taken a few words to change the summary from opaque to transparent for me.
This is a common problem in/. submissions. We get something like:
"The latest version (0.6) of Furball is finally available. It now has support for HCF, and has extended support for TLA, plus a whole bunch of UI improvements, including LERs and SEDs and a BRS. Lead developer Dead Beef says 'This is a huge milestone for us on the way to 1.0.'"
This is fine for the 3% of us who know what Furball is. The rest of us are left guessing that it is probably software.
There is a lot of fuss about whether cell phones, wi-fi etc. can damage bodies and minds by their radio waves. Although there is a lot of fuss, it is not justified by much (any?) significant scientific data.
Now it is shown that "intermediate frequency electrical fields" (whatever that means) can damage cancerous brain cells. Does this mean that a physiological effect (beneficial in this case) has been demonstrated, so that an adverse effect becomes more plausible?
I have no idea of the frequencies and amplitudes involved in the two cases (tumour treating fields vs. cell phones).* I'm guessing that the situations are so different that this result says nothing about the physiological effects of cell-phone exposure, but as the linked article contains no useful information about this, and the paper is unavailable, it is just a guess.
* I've looked for the paper on the PNAS website, but I can't find it - perhaps it is accepted but not yet published.
That sentence is verbatim from the article. It doesn't make any more sense in context. I came to read the posts in the hopes that someone else could interpret it for me.
Interestingly, the first factor is quite small, and trivially easy to find. The following Mathematica code finds it in less then 3.5 seconds on my 4 year old computer:
You may be able to carry just a small quantity of water, and recycle the engine exhaust as a water supply. Now your hydrogen is going in a closed loop, and you're acquiring oxygen from the air. This way you only carry that oxygen at the end, when the fuel is 'spent'. This is a slight improvement on your scenario.
Do you have any conception of how much oxygen is in the atmosphere? And how much is constaintly being created and destroyed by natural processes.
Refining the aluminium oxide into aluminium releases the same amount of oxygen, so you're just 'reclaiming' oxygen released earlier in the fuel cycle.
Hm, actually, I think the refining might release oxygen in the form of carbon dioxide (consuming a graphite electrode) - I'm not sure. However, turning oxygen into carbon dioxide for the purpose of energy production isn't exactly new.
This is rather the point - the aluminium is there precisely because it 'contains' a large amount of chemical energy per kg. The system works by liberating that energy.
Whether (some natural energy source -> electricity -> aluminium -> converted to H2 in car -> internal combustion engine -> kinetic energy) is better than lots of other methods for turning natural energy sources into kinetic energy of your car is quite another matter.
You've missed the point. The hydrogen-producing reaction occurs in the car - the car carries the Al/Ge alloy as fuel.
Refueling does become more complex (fuel is a solid, and you need to dump waste as well as take on fresh fuel) and similarly the spent fuel needs to be trucked back to a refinary. It sounds marginal to me, but not as bad as you make out.
Slashdot Mirror
I accept the first criticism.
The idea that extinctions are periodic, and even that this is due to motion relative to the galactic plane, has been around at least since the late 80s - I remember this coming up from back then. You're just looking at the latest incarnation of an old idea.
OK, I've only read the summary (I have a lecture to give half an hour from now to prepare for) but I can see some objections:
* My boss (David Penny, Massey University) argues that the mammals and birds were already outcompeting the dinosaurs at the end of the cretaceous, so the asteroid was at best a coup-de-grace for them.
* The "periodic extinctions" idea has been around for decades, including the possible link to oscillations through the galactic plane.
* Mass extinctions are sudden. The increase in extragalactic cosmic rays exposure would be slow, over millions of years.
* The extragalactic cosmic ray exposure changes should be highly regular. The extinctions are irregular.
Good luck in court arguing that you shouldn't be liable for more than $5. If that were the case, nobody would ever pay for anything - either they don't get caught (and get it free) or they do get caught, and have to pay what they would have paid anyway.
If you steal the twinkies in the knowledge that there is only a 1% chance you'll be caught, then being charged $500 if you do get caught seems fair to me (although I don't think the law allows this argument. IANAL.)
So (interpolating a little) current high-end cards are bottlenecked by PCI bandwidth at about x12 or less. This means that the next generation high-end can be expected use upcoming PCI-e 2.0 (twice the speed per channel) otherwise they'll bottleneck on x16.
I'm having troubles determining how the cards in the old review (6800GT etc, which do fine on x4) compare to the current generation, but I'm guessing that the current mid-range (e.g. 8500, 7600) are comparable, and could therefore run at x4 without a problem.
My motherboard has great big old PCI slots, and tiny little 1xPCI-e slots which are just as capable. PCI-e has taken over for graphics cards, but I've never even seen a 1xPCI-e expansion card. (The motherboard manufacturers don't believe they'll be used either - they put them next to the 16x slot where double-width graphics cards will make them inaccessable.)
When will old PCI die? Perhaps very small format motherboards and laptops will eventually drive demand for 1xPCI-e cards?
For that matter - is there any reason for low-end PCI-e graphics cards to be 16x, rather than 8x or even 4x? (They'd still fit in a 16x slot.) I suppose there is no demand - any PCI-e motherboard has a 16x slot, and there isn't anything you'd want to put in it except a GPU. About the only use I can think of is if you wanted one computer to run many low-performance displays - e.g. 8 monitors off four GPUs, each using a 4x slot.
At a company meeting we had an "inspirational speaker" who was a Olympic rowing gold medalist. He described what they ate during training - it was about 5-6 *huge* meals per day, including one at about 2am. It wouldn't surprise me at all if it was 33,000 kCal per day or more. This was required to supply the energy they expended during training.
I don't really know what happened to multi-node jobs, as I never submitted one (except during a training workshop, when all other work had been cleared from the cluster) and rarely if ever noticed them on the queue. It was the case that the queues were always full of single node jobs. I haven't been a heavy user - there were about 3 occasions where I had number crunching which would have taken 2-4 weeks on my desktop computer which I could break up and submit to Helix and get it all back in a few days.
I think in general it hasn't been too much of a problem - everyone wants to run a bunch of single node jobs anyhow.
In general, however, I see big problems if you can't migrate jobs. Say somebody wants 32 CPUs. You can stop running new jobs until 32 CPUs are free (which could mean 31 CPUs sit idle for a week waiting). You could suspend the jobs already running on 32 CPUs and resume them after the 32 CPU job is complete (which could cause a 1 hour job 5 minutes short of completion to be delayed for a week by the big job.) You could reserve half the CPUs to only be available for 32 CPU jobs (which could lead to half the cluster being idle while there is a huge backlock of single CPU jobs.) You could start the 32 CPU job running concurrantly with the pre-existing jobs on 32 nodes. Then if the single CPU jobs on 31 of the nodes finish promptly but one has a long-running resource-hogging job, you end up with 31 CPUs idleing most of the time waiting for the one shared CPU to catch up.
I think that covers the possibilities. If you can't predict in advance how long jobs will take, you can't schedule jobs with different numbers of nodes efficiently.
I've used a 64 node Beowulf cluster on occasion. The queue was generally full of single-node jobs and multi-node jobs hardly ever ran. There really is no good way of scheduling in this environment without the ability to suspend a job on one node and later resume it on another. So far as I know, our scheduling software was not capable of this. (Fortunately, I was just submitting single node jobs.)
Starting with the information in the summary, I spent a few minutes web searching. "bproc" appears not to be capable of this: it just means that your primary node can "see" the processes running on other nodes, so you can use 'ps', 'kill' etc. on them. However OpenSSI has "process migration". Is this the ability to move processes from one node to another?
Here.
Here's the essence of what he has to say:
"I don't think it was possible to keep the book under wraps."
"There are simply too many people who must be trusted in order for the security to hold."
"My guess is that the publishers will lose zero sales"
At $200/hr, that's at least $400,000 a year...
You're assuming they get to bill 40 hours per week, and have no costs. They have to spend some of their time on non-billable running-the-business work, possibly pay secretaries and legal researchers, rent, malpractise insurance, attending conferences...
IANAL, so I don't know what the overheads are like, but they'll be a non-negligable fraction of that $400,000.
There is an excellent novel "Passage" by Connie Willis involving scientists researching near death experiences (NDEs). (This is fiction about science, rather than Science Fiction, although she writes that too.) They're also doing battle with a crackpot researcher who vigorously prompts patients into "remembering" angels etc etc.
I think that 500W figure came off the top of someone's head. I suspect the described system would actually idle at about 200W and peak at around 500W.
PSU (power supply unit) capacity is being way oversold. A desktop PC just isn't going to break 300W peak unless it is a hard-core gaming machine. Even a decent gaming machine (fast CPU and a single nearly-top-of-the-line GPU) won't break 300W. See here for examples of what 300W will run. (The thread started ~4 years ago, so you might want to skip to the end.)
Here's an example:
Intel Pentium D950, overclocked / overvolted 10% (Presler, 3.74GHz)
ATI Radeon X1950XTX-512 PCIe graphics
Asus P5LD2-VM motherboard
4 x 1024MB Corsair DDR2-6400 RAM
Hitachi Deskstar 7K80 80GB hard drive
WD Raptor WD1500ADFD 150GB 10krpm HDD
1 x 120mm fan
Total AC power draw is an absolute max of 298W; estimated DC power draw is 256W.
_
Myself:
No TV, never owned one. I've been pretty much TV free for almost 20 years. (I'm occasionally tempted by a flatmate's TV.)
No cell phone. I had one for a couple of years (an old hand-me-down) which was only used or carried occasionally, but a few months ago they shut down the network it ran on.
No digital music player.
I've had a car for ~5 years now, but I've had many years car-free. Going back would be hard. When car-free, I lived in a bigger city* with better public transport and worse traffic
Dishwasher and microwave: I've got them - I'm lazy. I don't have a lawnmower - because I pay someone to do the lawn.
Computer would be hard to live without. I also use it as a substitute TV, for watching DVDs (mostly from the library.)
Lots and lots of books. Sometimes I think they own the house more than I do.
* Was Auckland (pop ~1 million), now Palmerston North (~75,000.)
While "most worthless story ever" is a huge overstatement, the criticism is justified. I knew that Intel were pushing flash cache on the motherboard, but I didn't know (or forgot) that they called it "turbo memory." It would only have taken a few words to change the summary from opaque to transparent for me.
/. submissions. We get something like:
This is a common problem in
"The latest version (0.6) of Furball is finally available. It now has support for HCF, and has extended support for TLA, plus a whole bunch of UI improvements, including LERs and SEDs and a BRS. Lead developer Dead Beef says 'This is a huge milestone for us on the way to 1.0.'"
This is fine for the 3% of us who know what Furball is. The rest of us are left guessing that it is probably software.
There is a lot of fuss about whether cell phones, wi-fi etc. can damage bodies and minds by their radio waves. Although there is a lot of fuss, it is not justified by much (any?) significant scientific data.
Now it is shown that "intermediate frequency electrical fields" (whatever that means) can damage cancerous brain cells. Does this mean that a physiological effect (beneficial in this case) has been demonstrated, so that an adverse effect becomes more plausible?
I have no idea of the frequencies and amplitudes involved in the two cases (tumour treating fields vs. cell phones).* I'm guessing that the situations are so different that this result says nothing about the physiological effects of cell-phone exposure, but as the linked article contains no useful information about this, and the paper is unavailable, it is just a guess.
* I've looked for the paper on the PNAS website, but I can't find it - perhaps it is accepted but not yet published.
Value of including craplets > cost of Windows.
Therefore, there is business case for MS to give away craplet-infested Windows install disks.
I don't know whether to be amused, intrigued, or run screaming in horror.
That sentence is verbatim from the article. It doesn't make any more sense in context. I came to read the posts in the hopes that someone else could interpret it for me.
You're absorbing oxygen from the air.
Try to be correct in future when issuing a put-down. Or be polite, then you're just wrong, rather than an idiot.
Interestingly, the first factor is quite small, and trivially easy to find. The following Mathematica code finds it in less then 3.5 seconds on my 4 year old computer:
With[{x = 2^1039 - 1}, Prime[Select[Range[1, 360000], (Mod[x, Prime[#]] == 0) &]]]
Finding the next factor like this (by trial division) should take a mere 10^70 or so times longer.
I've put these into Mathematica and confirmed they are correct.
This is incorrect.
1 5017753575689937658432079465\1 4451080388584562460719428810\9 2146903328852155899782370013\3 4387684495620912657652829388\
The two factors are prime, and multiply to the given number, but that number is not 2^1039-1.
2^1039-1=
5890680864316836766447387249177476247119386964598
5559932591384900650140340063891615625817543763223
7610698331745992221533871131893632012106238622173
7184806201826907368669534112523820726591354912103
7
You may be able to carry just a small quantity of water, and recycle the engine exhaust as a water supply. Now your hydrogen is going in a closed loop, and you're acquiring oxygen from the air. This way you only carry that oxygen at the end, when the fuel is 'spent'. This is a slight improvement on your scenario.
Are you serious?
Do you have any conception of how much oxygen is in the atmosphere? And how much is constaintly being created and destroyed by natural processes.
Refining the aluminium oxide into aluminium releases the same amount of oxygen, so you're just 'reclaiming' oxygen released earlier in the fuel cycle.
Hm, actually, I think the refining might release oxygen in the form of carbon dioxide (consuming a graphite electrode) - I'm not sure. However, turning oxygen into carbon dioxide for the purpose of energy production isn't exactly new.
This is rather the point - the aluminium is there precisely because it 'contains' a large amount of chemical energy per kg. The system works by liberating that energy.
Whether (some natural energy source -> electricity -> aluminium -> converted to H2 in car -> internal combustion engine -> kinetic energy) is better than lots of other methods for turning natural energy sources into kinetic energy of your car is quite another matter.
You've missed the point. The hydrogen-producing reaction occurs in the car - the car carries the Al/Ge alloy as fuel.
Refueling does become more complex (fuel is a solid, and you need to dump waste as well as take on fresh fuel) and similarly the spent fuel needs to be trucked back to a refinary. It sounds marginal to me, but not as bad as you make out.