First of all flash has replaced Hard Drives below a certain size. I doubt that you can find a sub 50 GB hard drive these days. If you do they are pretty rare and the price per gigabyte will be really high. Lets look at this for laptop drives (since most SSDs come in laptop form factors).
The cheapest laptop HDD newegg sell is 80GB for $36.99. If you wanted all 80GB that would work out at about $0.46 per gigabyte but lets say you decided you only need 32GB so most of the drive is wasted and the effective cost per gigabyte is about $1.16
The cheapest SSD newegg offer is 32GB for $84.99 giving a cost per gigabyte of about $2.66. Plus reading the reviews it's really not a drive I'd want to buy.
For desktops the difference is marginally greater at these capacities (the HDD will be slightly cheaper and the SSD will need adaptor rails)
There are several reasons to buy SSDs over HDDs but at least if you want a SATA device designed to mount in an ordinary desktop or laptop HDD bay then afaict the lines for capacity vs cost never cross.
If you want to put them in laptops then you either have to make them fit in the standard laptop drive form factor or convince laptop manufacturers to redesign thier product line to fit your device.
Besides a cubiod package seems like quite a nice form factor to design electronics for, just stack a load of boards that are all the same size and shape. What would you suggest instead?
Afaict the real problem at the moment is cost not ability to pack the parts into a given package. You can get a 512GB SSD in the 2.5 inch 9.5mm high format that most laptops take ( http://www.newegg.com/Product/Product.aspx?Item=N82E16820139115 ) which is not that much smaller than the largest HDD you can get in that form factor (750GB) but it will set you back $1,439.00 ( ) which is likely as much as the laptop you are fitting it to.
Linux distros likewise tend to use thier own trackers.
Tracking your own torrents uses minimal bandwidth and afaict is pretty easy to set up. So the main reason to use a public tracker is because you don't want to draw attention to yourself.
Kodachrome Film Status: The last day of processing for all types of Kodachrome film will be December 30th, 2010. The last day Kodak will accept prepaid Kodachrome film in Europe is November 30th, 2010. Film that is not in our lab by noon on December 30th will not be processed.
Dwayne’s Photo IS NOT CLOSING! We will continue to process other types of film and provide all our other normal services in 2011. Only Kodachrome film processing is being discontinued.
I'd guess that even with them being the only kodachrome processor left in the world they realise that with the film discontinued there just won't be enough buisness left to justify keeping the line (kodachrome is quite a complex process afaict).
Had they access to the source code to the shit they where actually running perhaps they would have upgraded the OS long ago. Probablly not, security isn't considered a huge issue because the systems in question probablly aren't exposed to the internet and every change brings a risk of new problems (better the devil you know than the devil you don't). So PCs performing embedded functions are unlikely to get upgraded unless something forces it.
From other comments here it does actually sound like in this case they were investigating an OS upgrade but only because they were being pushed by dying hardware.
remember Linus *created* a distributed version management tool (git) when he couldn't use anymore BitKeeper.. remember linus refused to use version control at all until the author of bitkeeper backed him into a corner fixing by every perceived issue until he ran out of excuses.
A DVCS actually makes it easier for a head honcho to retain complete control of the main tree. With a traditional vcs you pretty much have to let major developers commit directly to the main repositry or manually manage patches from them like you would if you weren't using a vcs at all. With a DVCS you can pull in commits while keeping thier history.
With most digital storage it's very much an all or nothing affair, either a block is preserved completely, it's unreadable or it's corrupted in a way that throws the compression of. And with compressed formats if you lose a block there is a good chance you won't be able to decipher anything after it anyway (you may still be able to get bits of the image before them).
If you really care about your data keep multiple copies and checksums (the built in checksuming of the drives is NOT sufficiant, use something like md5 or sha1 where the chance of damaged data matching is incrediblly small) and check the copies of the data against the checksums periodically.
the nameplate output is peak, not average. The difference is for conventional power plants the nameplate output is what they can give you either on demand (for a peaker plant) or pretty much all the time (for a base load plant).
For renewables other than dam based hydro the nameplate output is what they produce under ideal conditions. Those ideal conditions probablly won't line up with peak demand.
Dam based hydro is an unusual case in that it can have very high peak power and deliver that peak power on demand but it's average power is limited by the water entering the top lake
that doesn't take into account that half of the sales growth this quarter was from first-time buyers. That means 16.5% of their Mac sales came from switchers. Without knowing how they define first time buyer (e.g. does someone buying thier second mac but buying it for delivery to a different address than last time count as a first time buyer) that statistic is kind of meaningless.
The former were the first readily available consumer way of copying music without much in the way of generation loss IIRC most consumer MD decks have SCMS so you could only make single generation copies digitally and for a second gen copy analog was the only option.
Plus even if you defeated SCMS you would still have the sound being encoded and decoded into atrac every generation.
Better than cassete certainly but still generation loss there.
Afaict the original structure was already gone when the pdf was made, you can only try to reverese engineer it from the drawing objects.
You might want to try converting to postscript using ghostscript and then converting to svg using pstoedit. You still won't have the original structure but at least you should have the table shape as a vector drawing rather than a bitmap.
Much like ebay/paypal in that regard, the network effects caused by being the biggest site of their type outweigh for many users the bad behaviour of the site operator.
From there, increasing NNNs indicate generally increasing performance and "features" Generally the first digit indicates a series with distinct features while the later digits give an indication of performance within that series. For example a 920 is (according to tom's hardware's benchmarks) slower than an 870 but the 920 will give you more ram channels (and as a consequence higher max ram) and more fast PCIe lanes.
Things have got a bit messed up with these 6-core chips though as intel have decided to shoehorn them into the i7-9xx series rather than put them in thier own series.
If you mean craft in orbit arround earth the trouble is it's hard to put huge antennas in space. In general antenna gain is related to physical size and antenna gain is important for long distance work because (unlike amplifier gain) it makes the antenna pick up more signal without making it pick up more noise (assuming noise is equal in all directions)
If you mean craft in orbit around the target planet then you still need to downlink the data to a big antennae on earth. You just may be able to be more flexible about when you do it.
PCI is nowhere close to being fast enough for USB 3, USB 2 sure, but not USB 3. True it's not fast enough to max out USB3 but i'd still expect a PCI USB3 card to beat a USB2 connection.
On the other hand, PCIe is a totally different story, and just about every motherboard these days includes at least a couple PCIe slots. Kind of, there is the graphics slot which is PCIe x16 but often that can only reasonablly be used for graphics either because the system doesn't have onboard graphics or because using it disables the onboard graphics. Granted this may be more of an issue on the intel side than the amd side (I don't have personal experiance with recent AMD systems). Of course the disabling of onboard graphics is rarely documented in the manual so unless you know for sure your board doesn't do this the only way to find out is to buy a card and try to install it:(
Afaict a typical LGA1156 board comes with the graphics slot, sometimes an x4 slot (which may be x16 physical allowing it's use for a secondary graphics card) and the remaining two positions are filled with either PCIe x1 slots or PCI slots. The x4 and x1 slots only run at 1.0 speeds (meaning that afaict they can't max out a 1x USB 3 controller).
If you want SLI or some other card that needs more than 1.0 x4 you need a board with the graphics slot either split (which cuts into your graphics cards bandwidth even when you aren't usuing the other card) or taken to a NF200 PCIe bridge (which drives up the cost). Or abandon the LGA1156 platform altogether (which means either going for a last gen processor, going to LGA1366 and paying a heavy premium on the CPU price or going over to AMD who can't match intel in performance per core)
Altogether far more complex and far more desisions to be made upfront than the days when you had an AGP slot that was just for graphics and PCI slots you could use for anything else and all you had to decide was how many of them you wanted.
Note that 1.0 x1 slots are only just over twice as fast as PCI slots (though they do get that bandwidth dedicated and full duplex while PCI is half duplex and often shared though nowadays usually not shared with very much)
From the benchmarks i've seen i series chips tend to be a bit faster than core 2 series chips of the same clock speed but not hugely so (not like the huge difference between P4 and core 2)
So mostly getting more performance is a matter of buying either higher clock speeds or (if your apps support it) more cores.
Newegg has the 6 core 2.8ghz AMD 1055T for $200 [newegg.com], Indeed, it's only double the price of a quad-core amd chip of the same clock speed. With the intel 960/970 the difference is only just over 1.5x.
The real issue is there is a huge gap between the top end AMD 6-core chip and the bottom end intel 6-core chip. So the midrange buyer essentially has a choice between buying the top end AMD 6-core or buying an intel chip with only four cores but better performance of each individual core. IMO for desktop use the latter is more important.
A big problem is it's very difficult to judge obviousness, something that would be obvious to someone who has spent some time working on the same problem won't nessacerlly be obvious to a patent examiner. After the fact it's even harder.
Another big problem is those soloutions which aren't particularlly obvious but where there is only a very small number (sometimes only one) of good soloutions which are likely to be found eventually by multiple parties. Lightbulbs are a good example of this, both edison and swan came to the same conclusion on how to make a usable filament at about the same time independently.
And a final big problem is cases where you have to use a patented method not because it's the only or even the best soloution to the underlying problem but because it's the soloution compatible with what everyone else is doing.
I remember seeing a claim (this was a while back but afaict virgin media haven't extended thier network since before the merger) that virgin media covered about half of UK households.
They say the per home of installation of broadband (via fiber) is between 850 and 1300 Euros. True but if they are going for a gauranteed minimum speed of 2megabit they don't need to run fiber to every home, they only need to get it to a cabinet close enough to the homes that DSL over the existing phone lines can cary the rest.
even in my remote town in the middle of nowhere But how big is this town and how does service degrade as you move away from that town. Even a fairly small town should have enough buisness to support an incoming fiber and a phone exchange+DSLAM.
Knowing UK broadband the actual service here will most likely be rate adaptive ADSL2, that is up to 24Mb but more likely 10 or so on a fairly good line and much less as you get further away from the exchange.
It all comes down to number of DSLAMs. the faster they want the minimum service level to be the shorter they have to make the worst case line from the DSLAM to the customer. That in turn means they need more fiber and more DSLAM cabinets and that means more cost.
What they really mean when they say "at least 2Mb" is getting the DSLAMs (fed by fiber then some form of DSL to the cutomers) close enough to the customers that they can get 2Mb. Most people will likely get more than that.
I do wonder why it's so bloody expensive though and especially I wonder if BT are putting stuff that isn't really part of the universal broaband project in those figures.
Slashdot Mirror
First of all flash has replaced Hard Drives below a certain size. I doubt that you can find a sub 50 GB hard drive these days. If you do they are pretty rare and the price per gigabyte will be really high.
Lets look at this for laptop drives (since most SSDs come in laptop form factors).
The cheapest laptop HDD newegg sell is 80GB for $36.99. If you wanted all 80GB that would work out at about $0.46 per gigabyte but lets say you decided you only need 32GB so most of the drive is wasted and the effective cost per gigabyte is about $1.16
The cheapest SSD newegg offer is 32GB for $84.99 giving a cost per gigabyte of about $2.66. Plus reading the reviews it's really not a drive I'd want to buy.
For desktops the difference is marginally greater at these capacities (the HDD will be slightly cheaper and the SSD will need adaptor rails)
There are several reasons to buy SSDs over HDDs but at least if you want a SATA device designed to mount in an ordinary desktop or laptop HDD bay then afaict the lines for capacity vs cost never cross.
If you want to put them in laptops then you either have to make them fit in the standard laptop drive form factor or convince laptop manufacturers to redesign thier product line to fit your device.
Besides a cubiod package seems like quite a nice form factor to design electronics for, just stack a load of boards that are all the same size and shape. What would you suggest instead?
Afaict the real problem at the moment is cost not ability to pack the parts into a given package. You can get a 512GB SSD in the 2.5 inch 9.5mm high format that most laptops take ( http://www.newegg.com/Product/Product.aspx?Item=N82E16820139115 ) which is not that much smaller than the largest HDD you can get in that form factor (750GB) but it will set you back $1,439.00 ( ) which is likely as much as the laptop you are fitting it to.
Linux distros likewise tend to use thier own trackers.
Tracking your own torrents uses minimal bandwidth and afaict is pretty easy to set up. So the main reason to use a public tracker is because you don't want to draw attention to yourself.
I wonder if another option could be to add a regulator IC of some sort and power it from an alkaline cell.
According to his website he isn't shutting down the buisness just stopping developing kodachrome.
http://www.kansas.com/2010/07/14/1403115/last-kodachrome-roll-processed.html
Kodachrome — The End of an Era
Kodachrome Film Status: The last day of processing for all types of Kodachrome film will be December 30th, 2010. The last day Kodak will accept prepaid Kodachrome film in Europe is November 30th, 2010. Film that is not in our lab by noon on December 30th will not be processed.
Dwayne’s Photo IS NOT CLOSING! We will continue to process other types of film and provide all our other normal services in 2011. Only Kodachrome film processing is being discontinued.
I'd guess that even with them being the only kodachrome processor left in the world they realise that with the film discontinued there just won't be enough buisness left to justify keeping the line (kodachrome is quite a complex process afaict).
Had they access to the source code to the shit they where actually running perhaps they would have upgraded the OS long ago.
Probablly not, security isn't considered a huge issue because the systems in question probablly aren't exposed to the internet and every change brings a risk of new problems (better the devil you know than the devil you don't). So PCs performing embedded functions are unlikely to get upgraded unless something forces it.
From other comments here it does actually sound like in this case they were investigating an OS upgrade but only because they were being pushed by dying hardware.
remember Linus *created* a distributed version management tool (git) when he couldn't use anymore BitKeeper..
remember linus refused to use version control at all until the author of bitkeeper backed him into a corner fixing by every perceived issue until he ran out of excuses.
A DVCS actually makes it easier for a head honcho to retain complete control of the main tree. With a traditional vcs you pretty much have to let major developers commit directly to the main repositry or manually manage patches from them like you would if you weren't using a vcs at all. With a DVCS you can pull in commits while keeping thier history.
With most digital storage it's very much an all or nothing affair, either a block is preserved completely, it's unreadable or it's corrupted in a way that throws the compression of. And with compressed formats if you lose a block there is a good chance you won't be able to decipher anything after it anyway (you may still be able to get bits of the image before them).
If you really care about your data keep multiple copies and checksums (the built in checksuming of the drives is NOT sufficiant, use something like md5 or sha1 where the chance of damaged data matching is incrediblly small) and check the copies of the data against the checksums periodically.
Would it be possible to set up a seperate storage location close to where the equipment is used but with much tighter rules on who has access?
the nameplate output is peak, not average.
The difference is for conventional power plants the nameplate output is what they can give you either on demand (for a peaker plant) or pretty much all the time (for a base load plant).
For renewables other than dam based hydro the nameplate output is what they produce under ideal conditions. Those ideal conditions probablly won't line up with peak demand.
Dam based hydro is an unusual case in that it can have very high peak power and deliver that peak power on demand but it's average power is limited by the water entering the top lake
that doesn't take into account that half of the sales growth this quarter was from first-time buyers. That means 16.5% of their Mac sales came from switchers.
Without knowing how they define first time buyer (e.g. does someone buying thier second mac but buying it for delivery to a different address than last time count as a first time buyer) that statistic is kind of meaningless.
The former were the first readily available consumer way of copying music without much in the way of generation loss
IIRC most consumer MD decks have SCMS so you could only make single generation copies digitally and for a second gen copy analog was the only option.
Plus even if you defeated SCMS you would still have the sound being encoded and decoded into atrac every generation.
Better than cassete certainly but still generation loss there.
Afaict the original structure was already gone when the pdf was made, you can only try to reverese engineer it from the drawing objects.
You might want to try converting to postscript using ghostscript and then converting to svg using pstoedit. You still won't have the original structure but at least you should have the table shape as a vector drawing rather than a bitmap.
Much like ebay/paypal in that regard, the network effects caused by being the biggest site of their type outweigh for many users the bad behaviour of the site operator.
From there, increasing NNNs indicate generally increasing performance and "features"
Generally the first digit indicates a series with distinct features while the later digits give an indication of performance within that series. For example a 920 is (according to tom's hardware's benchmarks) slower than an 870 but the 920 will give you more ram channels (and as a consequence higher max ram) and more fast PCIe lanes.
Things have got a bit messed up with these 6-core chips though as intel have decided to shoehorn them into the i7-9xx series rather than put them in thier own series.
start having craft in orbit
Craft in orbit where?
If you mean craft in orbit arround earth the trouble is it's hard to put huge antennas in space. In general antenna gain is related to physical size and antenna gain is important for long distance work because (unlike amplifier gain) it makes the antenna pick up more signal without making it pick up more noise (assuming noise is equal in all directions)
If you mean craft in orbit around the target planet then you still need to downlink the data to a big antennae on earth. You just may be able to be more flexible about when you do it.
PCI is nowhere close to being fast enough for USB 3, USB 2 sure, but not USB 3.
True it's not fast enough to max out USB3 but i'd still expect a PCI USB3 card to beat a USB2 connection.
On the other hand, PCIe is a totally different story, and just about every motherboard these days includes at least a couple PCIe slots. :(
Kind of, there is the graphics slot which is PCIe x16 but often that can only reasonablly be used for graphics either because the system doesn't have onboard graphics or because using it disables the onboard graphics. Granted this may be more of an issue on the intel side than the amd side (I don't have personal experiance with recent AMD systems). Of course the disabling of onboard graphics is rarely documented in the manual so unless you know for sure your board doesn't do this the only way to find out is to buy a card and try to install it
Afaict a typical LGA1156 board comes with the graphics slot, sometimes an x4 slot (which may be x16 physical allowing it's use for a secondary graphics card) and the remaining two positions are filled with either PCIe x1 slots or PCI slots. The x4 and x1 slots only run at 1.0 speeds (meaning that afaict they can't max out a 1x USB 3 controller).
If you want SLI or some other card that needs more than 1.0 x4 you need a board with the graphics slot either split (which cuts into your graphics cards bandwidth even when you aren't usuing the other card) or taken to a NF200 PCIe bridge (which drives up the cost). Or abandon the LGA1156 platform altogether (which means either going for a last gen processor, going to LGA1366 and paying a heavy premium on the CPU price or going over to AMD who can't match intel in performance per core)
Altogether far more complex and far more desisions to be made upfront than the days when you had an AGP slot that was just for graphics and PCI slots you could use for anything else and all you had to decide was how many of them you wanted.
Note that 1.0 x1 slots are only just over twice as fast as PCI slots (though they do get that bandwidth dedicated and full duplex while PCI is half duplex and often shared though nowadays usually not shared with very much)
From the benchmarks i've seen i series chips tend to be a bit faster than core 2 series chips of the same clock speed but not hugely so (not like the huge difference between P4 and core 2)
So mostly getting more performance is a matter of buying either higher clock speeds or (if your apps support it) more cores.
Newegg has the 6 core 2.8ghz AMD 1055T for $200 [newegg.com],
Indeed, it's only double the price of a quad-core amd chip of the same clock speed. With the intel 960/970 the difference is only just over 1.5x.
The real issue is there is a huge gap between the top end AMD 6-core chip and the bottom end intel 6-core chip. So the midrange buyer essentially has a choice between buying the top end AMD 6-core or buying an intel chip with only four cores but better performance of each individual core. IMO for desktop use the latter is more important.
I thought it was only amd who were doing that at the moment (though intel have done it in the past).
A big problem is it's very difficult to judge obviousness, something that would be obvious to someone who has spent some time working on the same problem won't nessacerlly be obvious to a patent examiner. After the fact it's even harder.
Another big problem is those soloutions which aren't particularlly obvious but where there is only a very small number (sometimes only one) of good soloutions which are likely to be found eventually by multiple parties. Lightbulbs are a good example of this, both edison and swan came to the same conclusion on how to make a usable filament at about the same time independently.
And a final big problem is cases where you have to use a patented method not because it's the only or even the best soloution to the underlying problem but because it's the soloution compatible with what everyone else is doing.
I remember seeing a claim (this was a while back but afaict virgin media haven't extended thier network since before the merger) that virgin media covered about half of UK households.
They say the per home of installation of broadband (via fiber) is between 850 and 1300 Euros.
True but if they are going for a gauranteed minimum speed of 2megabit they don't need to run fiber to every home, they only need to get it to a cabinet close enough to the homes that DSL over the existing phone lines can cary the rest.
even in my remote town in the middle of nowhere
But how big is this town and how does service degrade as you move away from that town. Even a fairly small town should have enough buisness to support an incoming fiber and a phone exchange+DSLAM.
Knowing UK broadband the actual service here will most likely be rate adaptive ADSL2, that is up to 24Mb but more likely 10 or so on a fairly good line and much less as you get further away from the exchange.
It all comes down to number of DSLAMs. the faster they want the minimum service level to be the shorter they have to make the worst case line from the DSLAM to the customer. That in turn means they need more fiber and more DSLAM cabinets and that means more cost.
What they really mean when they say "at least 2Mb" is getting the DSLAMs (fed by fiber then some form of DSL to the cutomers) close enough to the customers that they can get 2Mb. Most people will likely get more than that.
I do wonder why it's so bloody expensive though and especially I wonder if BT are putting stuff that isn't really part of the universal broaband project in those figures.