I haven't failed to realize it, I simply don't have enough information to make a judgment one way or the other. It's equally possible that those computers are used for more than just interfacing with the Shuttle, and those other uses require a rolling-over date counter. Again, this comes down to the criticality issue: it's far more important that the Shuttle systems remain in working order than that those on the ground do, and the designers may have decided that it simply wasn't worth the risk of trying to implement rollover on the Shuttle systems, whereas the ground systems needed rollover for e.g. data logging. Without knowing more about the systems, you can't just go and say "they screwed up"--that would be like blaming a toaster maker because the toaster shorted out when you put it in the bathtub: it's simply not designed for that.
I would say that requiring a reboot every year on December 31 is a pretty huge error.
I wouldn't. When you're designing something like Shuttle software that has to work absolutely flawlessly 100% of the time, you don't put in any frills. And on something that is only ever in space for 10-15 consecutive days at most, year-end handling is most certainly a frill. (If you are a professional software developer, it ought to be obvious just how many things could break by adding a feature like that. If the original design calls for a monotonically increasing day number, for example, there's very likely to be some code that relies on that, so you have to go through the entire system, checking everything that even touches the day counter to ensure it can handle a reset from 365 to 1--and then check everything that uses those routines, and so on and so on.)
I suspect this is routine to NASA, and the reporter just blew it out of proportion. After all, Windows can handle end-of-year rollover, so if the Shuttle can't then it's broken, right?
The ISS software has a `maintenance mode' that is supposed to be only accessible to members of the "verkiezingswacht", the Nedap election-day helpdesk. You need a password to get the software in this mode. A quick look in the binary revealed this password to be "GEHEIM", the Dutch word for "SECRET".
Hello? Did someone not get the memo about secure passwords? Or better yet, no default passwords at all? Granted, physical access makes the point rather moot, but if this is the kind of security the designer had in mind, it looks like they can give Diebold a run for their money . ..
FWIW, the three drives I bought in a system upgrade about a year ago (two 2.5" Toshiba MK4032GSX's and a 3.5" Seagate ST3400633AS) have been doing just fine--though I did inadvertently discover that those little plastic ribs on the SATA connectors aren't quite as strong as you might expect . . . *snap*
I assume you're comparing TDP with "I/O power", which is not a sane comparison.
It seems I was. Mea culpa.
You'll also see that:
"800MHz" "Min P-State"...(which is the state CnQ uses when idle) on it's own, gives lower power than HLT. The two working together, of course, produce the best results ("IDDC1 Max @ Min P-State").
This, however, doesn't jive with the numbers I'm seeing. Looking at ADA4400DAA6CD (page 91), HLT at Pmin ought to use nearly 50W less than HLT at P0 (12W versus 60W, going by P=IE), but I'm not even seeing a 10W difference. (Granted, the IDD values given are all maximums, so the reality may be that none of it makes much of a difference for Linux.)
A powered-on CPU is never stopped... HLT just tells it there's no work to do for this cycle, and so it uses slightly less power.
AMD's Athlon(TM) 64 Processor Power and Thermal Data Sheet would seem to disagree with you. See the table starting at page 30, where the Halt/Stop Grant power usage is less than 10% of TDP even at Pmin. (It seems that dual-core CPUs don't support Halt/Stop Grant, so that's admittedly not relevant to my case.) The HLT instruction description in the instruction reference also explicitly says that
Entering the HALT state puts the processor in low-power mode. Execution resumes when an unmasked hardware interrupt (INTR), non-maskable interrupt (NMI), system management interrupt (SMI), RESET, or INIT occurs.
which seems pretty clear to me that it's not just "idling for a cycle"--it's stopped completely. (Again, I don't know how relevant this is to dual-core CPUs, but in my case, idle power usage with everything back in the system--including a 3.5" HD I forgot was still in there--is around 68W. Not a trivial change from 75W, certainly, but not the kind of massive drop your comments had suggested to me, especially given that Windows eats 140W on the same machine.)
They're getting a bit behind the times, but looking through the CPUs on this site will give you a better understanding of my point: http://users.erols.com/chare/elec.htm
Interesting, I hadn't been aware of that site. Thanks for the pointer. It does seem that AMD's own documentation may be a little out of date, as it doesn't list the dual-core Socket AM2 processors, only the Socket 939 89/110W versions.
And in any case, the OP's point about VIA chips still holds, even if his numbers are somewhat skewed.
I'm not entirely convinced that the numbers still come out against VIA, but it certainly would be interesting to run a comparison with a CPU that does support a low-power halt state. (It would also be nice to see some mini-ITX motherboards for such; I've grown quite fond of my 26x21x6cm box.)
HLT and CnQ are two entirely seperate issues, both of which reduce your CPU's power usage, but CnQ does dramatically more than HLT ever could.
I'm pretty sure that's wrong--why would a CPU use more power stopped than when processing at any speed?--but just for kicks, I pulled everything but video card, one drive, and one stick of RAM, made sure powernow-k8 was loaded, and set the governor to powersave; and idle power usage still bottomed out at 55W. I did learn a couple of interesting tidbits (did you know that an SBLive PCI card pulls 3W even when you're not using it?), but I think it's pretty clear you're not going to get a dual-core Athlon system into the 20-30W range.
And yes, even in Linux power draw shoots over 100W when I load both cores, and performance drops along with power usage (but still close to 100W) if I set powersave mode.
Well, it looks like you were right--but in the wrong direction: I booted in Windows, unloaded the CnQ driver completely, and saw my idle power usage jump to 140W(!) with massive fan noise. I also experimented with cpufreq on the Linux side, but it didn't make any difference during idle, only when loaded; presumably the HLT in the kernel idle loop drops the CPU into low-power mode.
By my measurements, the video card (a GeForce 6600) takes 15W when idle (I can't test the video on the C3 system because the video chip is built in). I do have one more hard drive on the Athlon, but it's a 2.5" drive like the others in both systems, and shouldn't take more than 5W even when active. I'd accept that the motherboard and associated hardware takes a fair amount of power to run, but since it's necessary in order to make use of the CPU I think it's fair to count that as part of the CPU's power usage. As for CnQ, I haven't been able to measure any difference in idle power usage whether it's on or off.
So, while you're still waiting on your Via C7 to crunch those numbers (at 20w), a Core2 Duo or A64 X2 system can do it in 1/4-1/6 the time (at 35w), and clock down to low-power state (3-5w).
That's funny. I have a Via C3 box and an Athlon64 X2 box right here. One uses 30W when idle; the other, 75W. Care to guess which? (Hint: the systems and the numbers are given in the same order.)
"Planet" has only ever been defined by those nine things orbiting Sol, so of course it's ambigious when applied to anything else--and that's why I argue that astronomers ought to stay away from it entirely. You don't see scientists arguing over the precise definition of "moment" (as in "just a moment", not the physics term), do you? "Planet" ought to be the same way. It always has been and probably always will be ambiguous, and trying to override years and years of common sense with some arcane (to laymen) set of parameters is only asking for trouble.
Besides, the human brain is actually quite good at figuring out what a "similar object" is, as long as you don't need scientific precision. It's what we're built to do, after all.
. . . something you look at and say, "hey, that's a planet."
No, seriously. Given all the historical baggage surrounding the term "planet", people shouldn't be trying to use it as a scientific term in the first place. If you want something that can be used to scientifically denote a certain class of astronomical objects, call them "primary satellites" or something. What's wrong with saying something like this, for example? "A planet is one of the nine satellites of Sol: Mercury, Venus, Earth,..., Pluto; or a similar object orbiting another star that is widely recognized as a planet." That keeps the status quo with respect to our solar system, which doesn't seem to have hurt anything in the 76 years it's been around, and lets public opinion decide on anything else that pops up. Which leaves astronomers free to spend their valuable time actually watching the sky rather than trying to convince people that something that looks like a planet and smells like a planet isn't actually a planet.
NTT DoCoMo, in Japan, has a little hole-punch-like device they use to destroy the internal memory chip when you give your phone back, and best of all they do it right there on the spot: you give them your old phone, and they stick it in the device and go "crunch!" Of course, I haven't actually seen the schematics for any (much less all) of the DoCoMo phones so I could theoretically be being fooled, but given the nearly paranoid attitude among Japanese these days over personal information, I doubt DoCoMo would take that risk.
The European Space Agency estimates it to be around 100 billion Euros which isn't cheap.
I dunno . . . if the US can spend several hundred billion dollars to kill people in the Middle East, then a mere hundred billion or so for potential scientific advances doesn't seem all that bad.
Having just made use of this on a trans-Pacific (13h in the air) flight, I can definitely say it's not overpriced. Upstream is only around 4 kbps (making scp's to my home box a bit frustrating), but downstream was fast and stable enough that I didn't have much problem using either ssh or Mozilla. From the viewpoint of someone who does a Japan/US round trip about once a year, it's unfortunate they're dropping the service, though with the infrastructure necessary to support it I suppose I can't blame them.
Literally. It's a QR code, not a standard linear barcode, but it's the same concept, and these days probably more ads, information posters and the like have them than don't, and virtually all cell phones have cameras that can decode them into URLs. No links handy (who'd need barcodes when you're already on the web?), so you'll have to take my word for it, but they really are everywhere. Even the wrappers on McDonalds burgers have codes that take you to their *cough* nutrition information page.
Did anyone foresee that in the 90s the largest empire humans ever built would evaporate like a soap bubble? (Except Poul Anderson in the 1953 story "The Last Deliverer"). Talk about existing models of how things work falling apart.
You could probably say the same thing about most of the empires and other large societal structures of human history. People don't generally build things with the intent of them falling apart. (Companies do, but that's another issue entirely.)
Imagine an intelligent and curious human from rural Nepal, or Papua New Guinea. Could you explain your job to them?
Sure (language barrier aside). "In my country, we have lots of machines. My job is to figure out how to make new ones that work better." (software R&D) At a fundamental level, computer programs aren't that different from machine blueprints. If they don't know what machines are, I explain them as "tools that work by themselves", and maybe even build a simple hand-cranked something-or-other to show them; yes, that's skipping over the electricity part that makes it really automated, but then how many people even in our society really understand how electricity works, rather than just thinking of it as "the juice coming out of the wall"?
Could you do your job without the embryonic augmentations we have now, such as Google?
Sure. It might take a little longer, and I might actually have to drag myself out to the library (horrors! bright light in the sky!), but on the other hand I wouldn't have to deal with a flood of information tending to force my mind into thinking inside a box. And who knows, maybe I'd even meet a cute girl at the library.
We're partway up that vertical curve now.
Curve, maybe, just like the various revolutions we've had before. Vertical, hell no.
Slashdot Mirror
the comment settings page, where you can assign -6 to Funny comments and never have to have your reading interrupted by them again!
I haven't failed to realize it, I simply don't have enough information to make a judgment one way or the other. It's equally possible that those computers are used for more than just interfacing with the Shuttle, and those other uses require a rolling-over date counter. Again, this comes down to the criticality issue: it's far more important that the Shuttle systems remain in working order than that those on the ground do, and the designers may have decided that it simply wasn't worth the risk of trying to implement rollover on the Shuttle systems, whereas the ground systems needed rollover for e.g. data logging. Without knowing more about the systems, you can't just go and say "they screwed up"--that would be like blaming a toaster maker because the toaster shorted out when you put it in the bathtub: it's simply not designed for that.
I would say that requiring a reboot every year on December 31 is a pretty huge error.
I wouldn't. When you're designing something like Shuttle software that has to work absolutely flawlessly 100% of the time, you don't put in any frills. And on something that is only ever in space for 10-15 consecutive days at most, year-end handling is most certainly a frill. (If you are a professional software developer, it ought to be obvious just how many things could break by adding a feature like that. If the original design calls for a monotonically increasing day number, for example, there's very likely to be some code that relies on that, so you have to go through the entire system, checking everything that even touches the day counter to ensure it can handle a reset from 365 to 1--and then check everything that uses those routines, and so on and so on.)
I suspect this is routine to NASA, and the reporter just blew it out of proportion. After all, Windows can handle end-of-year rollover, so if the Shuttle can't then it's broken, right?
And now I have to worry about not just nuclear excitement to the northwest, but also radioactive snails??
Well, I guess I can at least give a play-by-play when they---holy shit WHAT'S THAT?! NO NO NOT WITH THE TAIL NOOOOOOOO
NO CARRIER
I bet the folks over at the Dihydrogen Monoxide Research Division would be delighted to have you on board.
I was not drunk when I wrote this; that's just some immature coder making fun of me.
Granted, being drunk is about the only valid excuse I could make for only initializing half of a doubly-linked list node . . .
His previous record was set last July.
He commented, "It's good that I was able to do it relaxed." (This is unambiguous.)
Hello? Did someone not get the memo about secure passwords? Or better yet, no default passwords at all? Granted, physical access makes the point rather moot, but if this is the kind of security the designer had in mind, it looks like they can give Diebold a run for their money . . .
Obviously I need sleep. So here's another mirror.
here (primed).
FWIW, the three drives I bought in a system upgrade about a year ago (two 2.5" Toshiba MK4032GSX's and a 3.5" Seagate ST3400633AS) have been doing just fine--though I did inadvertently discover that those little plastic ribs on the SATA connectors aren't quite as strong as you might expect . . . *snap*
I assume you're comparing TDP with "I/O power", which is not a sane comparison.
It seems I was. Mea culpa.
You'll also see that: ...(which is the state CnQ uses when idle) on it's own, gives lower power than HLT. The two working together, of course, produce the best results ("IDDC1 Max @ Min P-State").
"800MHz" "Min P-State"
This, however, doesn't jive with the numbers I'm seeing. Looking at ADA4400DAA6CD (page 91), HLT at Pmin ought to use nearly 50W less than HLT at P0 (12W versus 60W, going by P=IE), but I'm not even seeing a 10W difference. (Granted, the IDD values given are all maximums, so the reality may be that none of it makes much of a difference for Linux.)
A powered-on CPU is never stopped... HLT just tells it there's no work to do for this cycle, and so it uses slightly less power.
AMD's Athlon(TM) 64 Processor Power and Thermal Data Sheet would seem to disagree with you. See the table starting at page 30, where the Halt/Stop Grant power usage is less than 10% of TDP even at Pmin. (It seems that dual-core CPUs don't support Halt/Stop Grant, so that's admittedly not relevant to my case.) The HLT instruction description in the instruction reference also explicitly says that
which seems pretty clear to me that it's not just "idling for a cycle"--it's stopped completely. (Again, I don't know how relevant this is to dual-core CPUs, but in my case, idle power usage with everything back in the system--including a 3.5" HD I forgot was still in there--is around 68W. Not a trivial change from 75W, certainly, but not the kind of massive drop your comments had suggested to me, especially given that Windows eats 140W on the same machine.)
They're getting a bit behind the times, but looking through the CPUs on this site will give you a better understanding of my point: http://users.erols.com/chare/elec.htm
Interesting, I hadn't been aware of that site. Thanks for the pointer. It does seem that AMD's own documentation may be a little out of date, as it doesn't list the dual-core Socket AM2 processors, only the Socket 939 89/110W versions.
And in any case, the OP's point about VIA chips still holds, even if his numbers are somewhat skewed.
I'm not entirely convinced that the numbers still come out against VIA, but it certainly would be interesting to run a comparison with a CPU that does support a low-power halt state. (It would also be nice to see some mini-ITX motherboards for such; I've grown quite fond of my 26x21x6cm box.)
HLT and CnQ are two entirely seperate issues, both of which reduce your CPU's power usage, but CnQ does dramatically more than HLT ever could.
I'm pretty sure that's wrong--why would a CPU use more power stopped than when processing at any speed?--but just for kicks, I pulled everything but video card, one drive, and one stick of RAM, made sure powernow-k8 was loaded, and set the governor to powersave; and idle power usage still bottomed out at 55W. I did learn a couple of interesting tidbits (did you know that an SBLive PCI card pulls 3W even when you're not using it?), but I think it's pretty clear you're not going to get a dual-core Athlon system into the 20-30W range.
And yes, even in Linux power draw shoots over 100W when I load both cores, and performance drops along with power usage (but still close to 100W) if I set powersave mode.
Well, it looks like you were right--but in the wrong direction: I booted in Windows, unloaded the CnQ driver completely, and saw my idle power usage jump to 140W(!) with massive fan noise. I also experimented with cpufreq on the Linux side, but it didn't make any difference during idle, only when loaded; presumably the HLT in the kernel idle loop drops the CPU into low-power mode.
By my measurements, the video card (a GeForce 6600) takes 15W when idle (I can't test the video on the C3 system because the video chip is built in). I do have one more hard drive on the Athlon, but it's a 2.5" drive like the others in both systems, and shouldn't take more than 5W even when active. I'd accept that the motherboard and associated hardware takes a fair amount of power to run, but since it's necessary in order to make use of the CPU I think it's fair to count that as part of the CPU's power usage. As for CnQ, I haven't been able to measure any difference in idle power usage whether it's on or off.
So, while you're still waiting on your Via C7 to crunch those numbers (at 20w), a Core2 Duo or A64 X2 system can do it in 1/4-1/6 the time (at 35w), and clock down to low-power state (3-5w).
That's funny. I have a Via C3 box and an Athlon64 X2 box right here. One uses 30W when idle; the other, 75W. Care to guess which? (Hint: the systems and the numbers are given in the same order.)
"Planet" has only ever been defined by those nine things orbiting Sol, so of course it's ambigious when applied to anything else--and that's why I argue that astronomers ought to stay away from it entirely. You don't see scientists arguing over the precise definition of "moment" (as in "just a moment", not the physics term), do you? "Planet" ought to be the same way. It always has been and probably always will be ambiguous, and trying to override years and years of common sense with some arcane (to laymen) set of parameters is only asking for trouble.
Besides, the human brain is actually quite good at figuring out what a "similar object" is, as long as you don't need scientific precision. It's what we're built to do, after all.
. . . something you look at and say, "hey, that's a planet."
No, seriously. Given all the historical baggage surrounding the term "planet", people shouldn't be trying to use it as a scientific term in the first place. If you want something that can be used to scientifically denote a certain class of astronomical objects, call them "primary satellites" or something. What's wrong with saying something like this, for example? "A planet is one of the nine satellites of Sol: Mercury, Venus, Earth, ..., Pluto; or a similar object orbiting another star that is widely recognized as a planet." That keeps the status quo with respect to our solar system, which doesn't seem to have hurt anything in the 76 years it's been around, and lets public opinion decide on anything else that pops up. Which leaves astronomers free to spend their valuable time actually watching the sky rather than trying to convince people that something that looks like a planet and smells like a planet isn't actually a planet.
NTT DoCoMo, in Japan, has a little hole-punch-like device they use to destroy the internal memory chip when you give your phone back, and best of all they do it right there on the spot: you give them your old phone, and they stick it in the device and go "crunch!" Of course, I haven't actually seen the schematics for any (much less all) of the DoCoMo phones so I could theoretically be being fooled, but given the nearly paranoid attitude among Japanese these days over personal information, I doubt DoCoMo would take that risk.
The European Space Agency estimates it to be around 100 billion Euros which isn't cheap.
I dunno . . . if the US can spend several hundred billion dollars to kill people in the Middle East, then a mere hundred billion or so for potential scientific advances doesn't seem all that bad.
Having just made use of this on a trans-Pacific (13h in the air) flight, I can definitely say it's not overpriced. Upstream is only around 4 kbps (making scp's to my home box a bit frustrating), but downstream was fast and stable enough that I didn't have much problem using either ssh or Mozilla. From the viewpoint of someone who does a Japan/US round trip about once a year, it's unfortunate they're dropping the service, though with the infrastructure necessary to support it I suppose I can't blame them.
Literally. It's a QR code, not a standard linear barcode, but it's the same concept, and these days probably more ads, information posters and the like have them than don't, and virtually all cell phones have cameras that can decode them into URLs. No links handy (who'd need barcodes when you're already on the web?), so you'll have to take my word for it, but they really are everywhere. Even the wrappers on McDonalds burgers have codes that take you to their *cough* nutrition information page.
Did anyone foresee that in the 90s the largest empire humans ever built would evaporate like a soap bubble? (Except Poul Anderson in the 1953 story "The Last Deliverer"). Talk about existing models of how things work falling apart.
You could probably say the same thing about most of the empires and other large societal structures of human history. People don't generally build things with the intent of them falling apart. (Companies do, but that's another issue entirely.)
Imagine an intelligent and curious human from rural Nepal, or Papua New Guinea. Could you explain your job to them?
Sure (language barrier aside). "In my country, we have lots of machines. My job is to figure out how to make new ones that work better." (software R&D) At a fundamental level, computer programs aren't that different from machine blueprints. If they don't know what machines are, I explain them as "tools that work by themselves", and maybe even build a simple hand-cranked something-or-other to show them; yes, that's skipping over the electricity part that makes it really automated, but then how many people even in our society really understand how electricity works, rather than just thinking of it as "the juice coming out of the wall"?
Could you do your job without the embryonic augmentations we have now, such as Google?
Sure. It might take a little longer, and I might actually have to drag myself out to the library (horrors! bright light in the sky!), but on the other hand I wouldn't have to deal with a flood of information tending to force my mind into thinking inside a box. And who knows, maybe I'd even meet a cute girl at the library.
We're partway up that vertical curve now.
Curve, maybe, just like the various revolutions we've had before. Vertical, hell no.
That's why you use a mirror with frickin' lasers on its frame.