ATMs are in a competitive market, and the company responsible for bad effects is the same as the one in charge of making sure they don't happen. If your ATM screws up, you blame the bank -- and they fix the problem at their expense. If they don't, you'll probably switch banks. They're quite motivated to have the machine get it right.
No such motivation exists in the case of voting machines, and there are plenty of motivations to the contrary.
Yes, that would be a start. I'm certain there's more to it, though -- and I'm certain I'm not qualified to design a ballot. But, there are people who are, and we should require that ballots be designed by such people -- and we should expect them to get it right, along with the people who wrote the code for the machine.
No, you don't. State legislators aren't qualified to understand UI design any more than they are to understand the code behind the UI. There's a reason many skilled professions have certifying bodies, and that certain jobs are only allowed to be done by someone qualified. This is a job that requires the work of someone qualified. Presenting yourself as qualified to design a ballot when you're not is as big a problem as claiming you can design a bridge without a professional engineer certification.
UI design is an important consideration. Suppose you wanted to make a machine biased toward one candidate, without having anything obviously incriminating in the code. You could do something as simple as arrange the options so that parallax effects like you suggest make it easy to press the wrong portion of the screen. If the effects make people press high on average, and you put the candidate you wanted to favor at the top of the list, then pressing high on your candidate registers no check box, and people just press again. But, sometimes they'll press on the other candidate, get the one you wanted, and give up before figuring it out.
Ballot design needs to be fair, for all the same reasons the code needs to be correct. Badly designed ballots are probably just that -- bad design by someone who didn't know better. But, with something as important as an election, it's not ok to have badly designed ballots, and it's not ok to let people who don't know better design them. Design sufficiently bad that it shows meaningful bias should be treated as criminal election fraud, whether it was intentional or not -- there's simply no reason not to have that level of accountability.
Well, obviously this isn't the answer for the average newbie, but I'm not sure why you didn't do it -- ask her what the name of the file is (check the download link) and then open an xterm and run find.
Actually, some sorts of shielding make things worse. Moderate amounts of shielding just end up providing targets for the really high energy particles, which releases a big cloud of moderate energy particles on impact. The secondary radiation is both more abundant and more likely to interact with the stuff on the inside, and so causes a bigger problem. For space applications, there are intermediate amounts of shielding that will actually *increase* the total dose. (This is the case for cosmic rays, not solar flares; the latter can be fairly effectively shielded against, but is frequently less of a concern.) If you're not willing to put *large* amounts of mass around the thing to be shielded, it's often impossible to improve things all that much.
Hardening often consists of simple changes that are nonetheless expensive because they involve changes to the whole production line -- things like rating all the transistors for a noticeably higher voltage, to reduce the likelihood of a radiation-induced latchup event. As chip voltages get lower, this gets harder. Other changes include things like using isotopically pure boron in your dopants -- boron comes in two common isotopes, 10B and 11B. 11B is relatively immune to cosmic radiation, but 10B will fision when hit -- releasing secondary ionizing particles that cause a much greater problem than the cosmic ray by itself would. So rad-hard chips end up made with (expensive) depleted boron.
Combine these, and you see why it's difficult to find a decent selection of rad-hard chips, and also why an up-to-date radiation hardened CPU can cost over $100k each -- and also why you nonetheless need them, and can't really substitute anything short of a few tons of shielding.
Well, now we have the new Prohibition, in the form of the War on Some Drugs. Except this time around they decided they could do it without an amendment.
No need for convicts. There are plenty of *highly* qualified people who want to go badly enough they'd volunteer even for a one way trip -- and even if they didn't expect to survive all that long. These are smart, sane people who really do mean it. There may not be lots and lots of them, but there are certainly dozens and maybe hundreds to choose from.
Those come with warnings about the interactions with alcohol.
So now you are in favour of government regulation instead of the free market?
But it doesn't help anyone to argue that marijuana is 100% safe any more than it does to overinflate its dangers.
I've never heard anyone claim that it is 100% safe. The air we breathe isn't 100% safe. It is *safer* than alcohol as you said, and we don't put people in jail for brewing their own beer.
A market where buyers don't have good information isn't a free market; there are plenty of agents aside from the government that can get in the way of a free market. I'm generally in favor of labeling laws and product purity laws, but against bans on sales and such. I see nothing incongruous about believing that the government should require sellers of products represent those products accurately. Also, my support for both legalization and free markets is as much pragmatic as idealogical -- in both cases I think they tend to result in better worlds than the alternatives. To the extent that limited and targetted regulation improves the market, I'm in favor of it.
A claim that marijuana has caused zero deaths may or may not be the same as one that it's 100% safe; I really don't care. Either one is at best disingenuous, given that there is no shortage of marijuana-related deaths, even if none can be ascribed to THC overdose. I'm just trying to point out that handing out accurate information is both more helpful and more likely to make your point than an unthinking argument that it's either harmless or the tool of Satan.
The anti-emetic properties of marijuana are well documented; that's one of the main reasons it works well for chemotherapy patients. It also has analgesic properties, which are similarly important for cancer patients. Wikipedia and Google will both offer you plenty of links on the subject. Vomiting from marijuana probably results from large amounts consumed rapidly -- and consuming large amounts of nicotine rapidly can certainly cause vomiting, especially in people who aren't used to it. Smoking several cigarettes rapidly would probably do the same thing.
Those come with warnings about the interactions with alcohol. Responsibly advocates of marijuana legalization, if they're busy comparing marijuana risks to those of other drugs, should do the same. Especially since claiming the "zero deaths" statistic is highly misleading at best, given the incidence of fatal interactions. Combined OD deaths for other recreational drugs aren't normally left out of the statistics completely.
Oh, and if we're being rigorous about our statistics, do the alcohol deaths include drunk driving deaths? If so, fatal accidents while stoned should go in the marijuana column, and while those are rarer they're certainly not unheard of. (Speaking of which... are those deaths being double counted?)
You don't need to convince me that alcohol is worse than marijuana; I'm in full agreement on that. But it doesn't help anyone to argue that marijuana is 100% safe any more than it does to overinflate its dangers. Accurate information and responsible arguments are far more helpful to all concerned.
While there are no deaths directly attributable to marijuana alone, marijuana + alcohol is a noticeably more dangerous combination than just alcohol. As a potent antiemetic, marijuana will prevent you from vomiting up the alcohol still in your stomach unabsorbed when your body otherwise would, making it *far* easier to die from alcohol poisoning. Yes, the alcohol does the actual killing, but the marijuana is far from an innocent bystander.
That said, I fully support legalization of marijuana -- I just think it does people a disservice to claim it's completely safe when there is a common and potentially fatal drug interaction to be aware of.
And for the record, death tolls for the various psychedelics are also quite low. There is significant variation among them, though, with some being quite safe (eg LSD) and others less so (MDMA and other amphetamine / methamphetamine based psychedelics are still potent stimulants, with all their attendant risks).
Thirded. Between easy searching, excellent data on most of their parts, and a *wide* range of items, there's really not much more you could ask for. Prices aren't the cheapest you'll find, but they're usually competitive. You pay a mild premium for the huge inventory, fast shipping times, and truly excellent customer service -- but it's worth every penny, especially for small quantity orders.
For things like gears and sprockets and shafts, they won't have every conceivable size -- but the sizes they lack will be weird specialty ones. If you find yourself specifying gear sizes they don't stock, you should be giving serious thought to whether they know something you don't, and redesigning the part in question.
For the OP who doesn't yet know what to get... buy a few parts that look like they might be plausible, and try them out. Once you have a few parts in hand the question will become much easier. Just don't be afraid of ordering the wrong thing the first time out, and keep your initial orders small -- you'll be making more, after all.
Check your math. The glide ratio impacts the efficiency the whole way, uniformly. (Technically the L/D ratio, but they're nearly the same thing; I'll treat them as such here.) Assume your glide ratio is a conservative 10:1; the Gimli Glider demonstrated 12:1 with a 767-200. Climbing to 10 km at a slope of 1/5 uses 3x the cruising fuel for the first 50 km (3x the fuel per km travelled, but mostly done at lower speed; it's not a 3x change in throttle setting). The glide down then uses no fuel for the last 100 km, or half the fuel for the last 200 km. Either way, you get the energy back.
Energy losses in aircraft come from drag, and not much else. There's a second-order effect resulting from reduced L/D during the high AoA ascent, and another due to reduced AoA at low speed / high air pressure. As a result, you get the majority of the potential energy back, but obviously you don't recover the energy lost to drag -- but that portion is properly accounted as the cost of covering distance, not the cost of the climb and descent.
Regardless, it would be a civil violation, not a criminal one. The owner would have to pursue civil measures to get them to stop wearing it; the police can't enforce trademark usage without a court order to that effect, since no crime is being committed until the person using the trademark violates a court order. Of course, they may have committed a tort and be liable, but that still doesn't mean the police can take their stuff until a court specifically says so.
I'd probably rather have this thing working on me than the normal procedure -- bypass is definitely something I'd rather avoid. But, if this thing manages to make a large mistake and tear something, that might be a *big* problem. Note that software has certainly cause medical problems before -- see the Therac-25 for example. It's rare, but software mistakes can and do happen. Caution is obviously warranted, but this probably represents a vast improvement on the current state of the art.
You won't get national coverage, but you don't need it either. This is easily worthy of a few minute segment on the local news -- which ought to be sufficient to shame somebody into taking you more seriously. The phone company rep really won't want to go on camera saying "sorry, they're abusing the phone system, and we won't help."
Your math is fine, but your conclusion is wrong. We don't care how many kilowatt hours get used, but rather how much coal gets burned (or, equivalently, how much computing gets done). Using your numbers, reducing power consumption by 10% to 45kWh means that the data center consumes 90t of coal instead of 100t -- exactly the same as if we reduce the amount of coal required to produce 50 kWh by 10% without changing the amount of power consumed.
In your credit card example, either 10% reduction applied by itself gives you the same amount of savings -- but if applied together, they're multiplicative. Either one alone would result in you paying 90% of the sticker price, but together they multiply and you pay 81% of the sticker price. The same is true in the power example. Either improvement alone reduces coal consumption from 100t to 90t; taken together, coal consumption is down to 81t.
Alternately, in the power consumption case, you should conclude that by reducing power use by 10% we can now run 1.11 data centers for our 100t of coal (at a power level of 50kWh). By reducing coal consumption, we can run 1.11 data centers for our 100t of coal (at a power level of 55.55 kWh).
(Obviously we're ignoring secondary effects like the fact that reduced power consumption means lower grid infrastructure costs while improved generation efficiency has no impact on the grid itself.)
Of course, the best case is for both improvements to happen. And there's no dichotomy; whether or not Google improves their energy efficiency has no impact on whether or not the power company can or will improve theirs.
False dilemma, anyone? Is there any reason at all that we can't do both? Does the fact that electric generation could be better make this any less of a good thing?
For the most part, data center operators can't do much to improve the efficiency of electricity generation. Reducing how much they use, however, they can control -- and a 10% reduction in coal burned per kWhr produced has the same impact as a 10% reduction in kWhrs used. And, reducing power used is probably far easier at this stage, given the relative amounts of effort that have gone into the two pursuits.
And yes, I'm aware that Google has installed a large collection of solar panels. Good for them; I hope they add more. But even so, it makes sense to reduce the electricity demand of their data centers. They don't make all their own electricity yet, and even if they did, those solar panels have a nonzero environmental and monetary cost (better than coal is still nonzero).
Air conditioning costs have a much larger impact on PUE than UPS inefficiency (which ought to be very small -- do you have a reference that says otherwise?).
I agree with your general sentiment, though -- there are more interesting things to measure. Part of the problem is that different balances of CPU, disk, and network gear will produce different numbers for flops/watt, even with the same efficiency in each case.
There are plenty of places other than PUE to attack efficiency. For example, Google uses power supplies with a single 12V output and much higher efficiency (see their white paper (pdf)).
PUE does have several advantages, though, so I wouldn't call it rubbish. Changes to the servers themselves won't have much impact on it, so it makes sense as a metric of the efficiency of the data center building itself. Also, it's simple enough that there are few places to play games with the numbers -- if one data center has a better PUE than another, then it's better in that aspect (unless someone is just plain lying, of course). Sure, it doesn't measure everything, but having exactly what is and isn't measured in your number be unambiguous is quite valuable.
Why? What makes you think they aren't using those CPU cycles in ways that are useful to them and their customers?
I'd be curious to hear stats on how effectively they use those cycles, but I bet they do worry about it. Most of this is motivated by cost, and unused cycles are an expensive for of inefficiency (and so likely one of the first attacked).
A number of modern games can make use of 2+ cores, but 8 isn't going to happen with any efficiency. Note also that this is a cluster in a single box -- those 8 nodes are each different computers on a very fast local network. That means a different OS image per node, and each process on its own node. For lots of supercomputing applications, this is the norm -- each node does its share of the work and they talk over the network. But no games support this; they all expect to run on a single computer.
Also, for gaming performance, I imagine you'd want dual graphics cards -- which this box doesn't support. (It does include "visualization node" options, which have a single Quadro FX card each.)
Still, for something like a desktop render farm, this might make sense -- except I imagine the customers for such would be more interested in options with better price/performance.
Slashdot Mirror
ATMs are in a competitive market, and the company responsible for bad effects is the same as the one in charge of making sure they don't happen. If your ATM screws up, you blame the bank -- and they fix the problem at their expense. If they don't, you'll probably switch banks. They're quite motivated to have the machine get it right.
No such motivation exists in the case of voting machines, and there are plenty of motivations to the contrary.
Yes, that would be a start. I'm certain there's more to it, though -- and I'm certain I'm not qualified to design a ballot. But, there are people who are, and we should require that ballots be designed by such people -- and we should expect them to get it right, along with the people who wrote the code for the machine.
No, you don't. State legislators aren't qualified to understand UI design any more than they are to understand the code behind the UI. There's a reason many skilled professions have certifying bodies, and that certain jobs are only allowed to be done by someone qualified. This is a job that requires the work of someone qualified. Presenting yourself as qualified to design a ballot when you're not is as big a problem as claiming you can design a bridge without a professional engineer certification.
UI design is an important consideration. Suppose you wanted to make a machine biased toward one candidate, without having anything obviously incriminating in the code. You could do something as simple as arrange the options so that parallax effects like you suggest make it easy to press the wrong portion of the screen. If the effects make people press high on average, and you put the candidate you wanted to favor at the top of the list, then pressing high on your candidate registers no check box, and people just press again. But, sometimes they'll press on the other candidate, get the one you wanted, and give up before figuring it out.
Ballot design needs to be fair, for all the same reasons the code needs to be correct. Badly designed ballots are probably just that -- bad design by someone who didn't know better. But, with something as important as an election, it's not ok to have badly designed ballots, and it's not ok to let people who don't know better design them. Design sufficiently bad that it shows meaningful bias should be treated as criminal election fraud, whether it was intentional or not -- there's simply no reason not to have that level of accountability.
Well, obviously this isn't the answer for the average newbie, but I'm not sure why you didn't do it -- ask her what the name of the file is (check the download link) and then open an xterm and run find.
Actually, some sorts of shielding make things worse. Moderate amounts of shielding just end up providing targets for the really high energy particles, which releases a big cloud of moderate energy particles on impact. The secondary radiation is both more abundant and more likely to interact with the stuff on the inside, and so causes a bigger problem. For space applications, there are intermediate amounts of shielding that will actually *increase* the total dose. (This is the case for cosmic rays, not solar flares; the latter can be fairly effectively shielded against, but is frequently less of a concern.) If you're not willing to put *large* amounts of mass around the thing to be shielded, it's often impossible to improve things all that much.
Hardening often consists of simple changes that are nonetheless expensive because they involve changes to the whole production line -- things like rating all the transistors for a noticeably higher voltage, to reduce the likelihood of a radiation-induced latchup event. As chip voltages get lower, this gets harder. Other changes include things like using isotopically pure boron in your dopants -- boron comes in two common isotopes, 10B and 11B. 11B is relatively immune to cosmic radiation, but 10B will fision when hit -- releasing secondary ionizing particles that cause a much greater problem than the cosmic ray by itself would. So rad-hard chips end up made with (expensive) depleted boron.
Combine these, and you see why it's difficult to find a decent selection of rad-hard chips, and also why an up-to-date radiation hardened CPU can cost over $100k each -- and also why you nonetheless need them, and can't really substitute anything short of a few tons of shielding.
Well, now we have the new Prohibition, in the form of the War on Some Drugs. Except this time around they decided they could do it without an amendment.
No need for convicts. There are plenty of *highly* qualified people who want to go badly enough they'd volunteer even for a one way trip -- and even if they didn't expect to survive all that long. These are smart, sane people who really do mean it. There may not be lots and lots of them, but there are certainly dozens and maybe hundreds to choose from.
So now you are in favour of government regulation instead of the free market?
I've never heard anyone claim that it is 100% safe. The air we breathe isn't 100% safe. It is *safer* than alcohol as you said, and we don't put people in jail for brewing their own beer.
A market where buyers don't have good information isn't a free market; there are plenty of agents aside from the government that can get in the way of a free market. I'm generally in favor of labeling laws and product purity laws, but against bans on sales and such. I see nothing incongruous about believing that the government should require sellers of products represent those products accurately. Also, my support for both legalization and free markets is as much pragmatic as idealogical -- in both cases I think they tend to result in better worlds than the alternatives. To the extent that limited and targetted regulation improves the market, I'm in favor of it.
A claim that marijuana has caused zero deaths may or may not be the same as one that it's 100% safe; I really don't care. Either one is at best disingenuous, given that there is no shortage of marijuana-related deaths, even if none can be ascribed to THC overdose. I'm just trying to point out that handing out accurate information is both more helpful and more likely to make your point than an unthinking argument that it's either harmless or the tool of Satan.
The anti-emetic properties of marijuana are well documented; that's one of the main reasons it works well for chemotherapy patients. It also has analgesic properties, which are similarly important for cancer patients. Wikipedia and Google will both offer you plenty of links on the subject. Vomiting from marijuana probably results from large amounts consumed rapidly -- and consuming large amounts of nicotine rapidly can certainly cause vomiting, especially in people who aren't used to it. Smoking several cigarettes rapidly would probably do the same thing.
Those come with warnings about the interactions with alcohol. Responsibly advocates of marijuana legalization, if they're busy comparing marijuana risks to those of other drugs, should do the same. Especially since claiming the "zero deaths" statistic is highly misleading at best, given the incidence of fatal interactions. Combined OD deaths for other recreational drugs aren't normally left out of the statistics completely.
Oh, and if we're being rigorous about our statistics, do the alcohol deaths include drunk driving deaths? If so, fatal accidents while stoned should go in the marijuana column, and while those are rarer they're certainly not unheard of. (Speaking of which... are those deaths being double counted?)
You don't need to convince me that alcohol is worse than marijuana; I'm in full agreement on that. But it doesn't help anyone to argue that marijuana is 100% safe any more than it does to overinflate its dangers. Accurate information and responsible arguments are far more helpful to all concerned.
While there are no deaths directly attributable to marijuana alone, marijuana + alcohol is a noticeably more dangerous combination than just alcohol. As a potent antiemetic, marijuana will prevent you from vomiting up the alcohol still in your stomach unabsorbed when your body otherwise would, making it *far* easier to die from alcohol poisoning. Yes, the alcohol does the actual killing, but the marijuana is far from an innocent bystander.
That said, I fully support legalization of marijuana -- I just think it does people a disservice to claim it's completely safe when there is a common and potentially fatal drug interaction to be aware of.
And for the record, death tolls for the various psychedelics are also quite low. There is significant variation among them, though, with some being quite safe (eg LSD) and others less so (MDMA and other amphetamine / methamphetamine based psychedelics are still potent stimulants, with all their attendant risks).
You have a music box with 1HP rated mechanical components? Now that sounds like an interesting contraption. Pics, please!
Thirded. Between easy searching, excellent data on most of their parts, and a *wide* range of items, there's really not much more you could ask for. Prices aren't the cheapest you'll find, but they're usually competitive. You pay a mild premium for the huge inventory, fast shipping times, and truly excellent customer service -- but it's worth every penny, especially for small quantity orders.
For things like gears and sprockets and shafts, they won't have every conceivable size -- but the sizes they lack will be weird specialty ones. If you find yourself specifying gear sizes they don't stock, you should be giving serious thought to whether they know something you don't, and redesigning the part in question.
For the OP who doesn't yet know what to get... buy a few parts that look like they might be plausible, and try them out. Once you have a few parts in hand the question will become much easier. Just don't be afraid of ordering the wrong thing the first time out, and keep your initial orders small -- you'll be making more, after all.
Check your math. The glide ratio impacts the efficiency the whole way, uniformly. (Technically the L/D ratio, but they're nearly the same thing; I'll treat them as such here.) Assume your glide ratio is a conservative 10:1; the Gimli Glider demonstrated 12:1 with a 767-200. Climbing to 10 km at a slope of 1/5 uses 3x the cruising fuel for the first 50 km (3x the fuel per km travelled, but mostly done at lower speed; it's not a 3x change in throttle setting). The glide down then uses no fuel for the last 100 km, or half the fuel for the last 200 km. Either way, you get the energy back.
Energy losses in aircraft come from drag, and not much else. There's a second-order effect resulting from reduced L/D during the high AoA ascent, and another due to reduced AoA at low speed / high air pressure. As a result, you get the majority of the potential energy back, but obviously you don't recover the energy lost to drag -- but that portion is properly accounted as the cost of covering distance, not the cost of the climb and descent.
Regardless, it would be a civil violation, not a criminal one. The owner would have to pursue civil measures to get them to stop wearing it; the police can't enforce trademark usage without a court order to that effect, since no crime is being committed until the person using the trademark violates a court order. Of course, they may have committed a tort and be liable, but that still doesn't mean the police can take their stuff until a court specifically says so.
I'd probably rather have this thing working on me than the normal procedure -- bypass is definitely something I'd rather avoid. But, if this thing manages to make a large mistake and tear something, that might be a *big* problem. Note that software has certainly cause medical problems before -- see the Therac-25 for example. It's rare, but software mistakes can and do happen. Caution is obviously warranted, but this probably represents a vast improvement on the current state of the art.
This wouldn't be as annoying if getting both a "funny" and a "redundant" mod didn't cost a point of karma.
You won't get national coverage, but you don't need it either. This is easily worthy of a few minute segment on the local news -- which ought to be sufficient to shame somebody into taking you more seriously. The phone company rep really won't want to go on camera saying "sorry, they're abusing the phone system, and we won't help."
Your math is fine, but your conclusion is wrong. We don't care how many kilowatt hours get used, but rather how much coal gets burned (or, equivalently, how much computing gets done). Using your numbers, reducing power consumption by 10% to 45kWh means that the data center consumes 90t of coal instead of 100t -- exactly the same as if we reduce the amount of coal required to produce 50 kWh by 10% without changing the amount of power consumed.
In your credit card example, either 10% reduction applied by itself gives you the same amount of savings -- but if applied together, they're multiplicative. Either one alone would result in you paying 90% of the sticker price, but together they multiply and you pay 81% of the sticker price. The same is true in the power example. Either improvement alone reduces coal consumption from 100t to 90t; taken together, coal consumption is down to 81t.
Alternately, in the power consumption case, you should conclude that by reducing power use by 10% we can now run 1.11 data centers for our 100t of coal (at a power level of 50kWh). By reducing coal consumption, we can run 1.11 data centers for our 100t of coal (at a power level of 55.55 kWh).
(Obviously we're ignoring secondary effects like the fact that reduced power consumption means lower grid infrastructure costs while improved generation efficiency has no impact on the grid itself.)
Of course, the best case is for both improvements to happen. And there's no dichotomy; whether or not Google improves their energy efficiency has no impact on whether or not the power company can or will improve theirs.
False dilemma, anyone? Is there any reason at all that we can't do both? Does the fact that electric generation could be better make this any less of a good thing?
For the most part, data center operators can't do much to improve the efficiency of electricity generation. Reducing how much they use, however, they can control -- and a 10% reduction in coal burned per kWhr produced has the same impact as a 10% reduction in kWhrs used. And, reducing power used is probably far easier at this stage, given the relative amounts of effort that have gone into the two pursuits.
And yes, I'm aware that Google has installed a large collection of solar panels. Good for them; I hope they add more. But even so, it makes sense to reduce the electricity demand of their data centers. They don't make all their own electricity yet, and even if they did, those solar panels have a nonzero environmental and monetary cost (better than coal is still nonzero).
Air conditioning costs have a much larger impact on PUE than UPS inefficiency (which ought to be very small -- do you have a reference that says otherwise?).
I agree with your general sentiment, though -- there are more interesting things to measure. Part of the problem is that different balances of CPU, disk, and network gear will produce different numbers for flops/watt, even with the same efficiency in each case.
There are plenty of places other than PUE to attack efficiency. For example, Google uses power supplies with a single 12V output and much higher efficiency (see their white paper (pdf)).
PUE does have several advantages, though, so I wouldn't call it rubbish. Changes to the servers themselves won't have much impact on it, so it makes sense as a metric of the efficiency of the data center building itself. Also, it's simple enough that there are few places to play games with the numbers -- if one data center has a better PUE than another, then it's better in that aspect (unless someone is just plain lying, of course). Sure, it doesn't measure everything, but having exactly what is and isn't measured in your number be unambiguous is quite valuable.
Why? What makes you think they aren't using those CPU cycles in ways that are useful to them and their customers?
I'd be curious to hear stats on how effectively they use those cycles, but I bet they do worry about it. Most of this is motivated by cost, and unused cycles are an expensive for of inefficiency (and so likely one of the first attacked).
A number of modern games can make use of 2+ cores, but 8 isn't going to happen with any efficiency. Note also that this is a cluster in a single box -- those 8 nodes are each different computers on a very fast local network. That means a different OS image per node, and each process on its own node. For lots of supercomputing applications, this is the norm -- each node does its share of the work and they talk over the network. But no games support this; they all expect to run on a single computer.
Also, for gaming performance, I imagine you'd want dual graphics cards -- which this box doesn't support. (It does include "visualization node" options, which have a single Quadro FX card each.)
Still, for something like a desktop render farm, this might make sense -- except I imagine the customers for such would be more interested in options with better price/performance.
Also, I see no way to configure more than 32GB of memory per node (so 256GB max in the box).