Why does every 'douche' keep comparing a Xeon to a consumer grade processor? The reason is simply that apple refuses to sell a normal desktop. The choice of apple desktop machines comes down to
* mini: a small form factor machine built with latop parts and with specs comparable to a low end laptop
* imac: an all-in one with midrange desktop specs, no room for more hard drives or expansion cards, no USB3 or ESATA, and the options for screen size and graphics memeory tied together with the CPU options and hard drive options also tied into those choices to some extent.
* pro: a monster built out of components intended for dual socket server/workstation systems and with a pricetag to match.
All depends what education authorities do. If they decide to follow the IAU (which is likely as afaict there is no other organsation of similar standing pushing a rival definition) then that is what kids will be taught.
http://en.wikipedia.org/wiki/Eris_(dwarf_planet) "Eris, formal designation 136199 Eris, is the most massive known dwarf planet in the Solar System and the ninth most massive body known to orbit the Sun directly. It is estimated to be approximately 2300-2400 km in diameter,[14] and 27% more massive than Pluto or about 0.27% of the Earth's mass."
It is the object that finally lead to pluto's "demotion" from planet to dwarf plant.
One thing I could see is having a "kit" of this sort of thing so you determine what the line type is ( by a combination of observation and knowing local standards) get the appropriate tap out of your kit, maybe set the blade depth and deploy it.
Afaict steam only got going at all because valve forced anyone who bought HL2 and wanted to play it and anyone who wanted to continue playing HL1 online to sign up for steam.
It could but they would have to plan to screw you in advance.
Whereas with online activation they can screw you at any time. Maybe they simply lose interest in thier older stuff. Maybe they are bought by a competitor whose sole aim is to kill them. While I don't think it has happened with games yet it has certainly happened with online music.
True, their hook still looks limited to a fairly small range of wiring types though, it relies on the ground/neutral being bare and a phase conductor twisted in with it covered in fairly limited insulation. Probablly also limited in terms of the overall size of the bundle.
It also looks prone to shorting if it grips the cable the wrong way.
Firstly the advertised speed for DSL is generally the speed on a "perfect" line. Unless you have a FTTP setup or similar scenario where all copper lines are very short speeds are often much lower than advertised.
Secondly while there may not be contention on the local loop there is almost certain to be contention somewhere either in the ADSL backend network or in the network that joins it to the internet.
"A lot" is relative, Given the small ammounts of data involved and the short range it should easilly be possible at under 100ma. USB ports are supposed to be able to provide 500ma.
So either the GP is lying or he has machines with REALLY REALLY shitty USB ports or the wireless receiver is using far more power than nessacery.
Unfortunately I suspect that due to contractual limitations the content available to international viewers will only be a subset of that available to british ones.
Geographic restrictions aren't really a software issue. IIRC amazon mp3 for example has geographic restrictions yet doesn't need any special client software at all.
Geographic restrictions on digital content come primerally from the fact that media is generally licensed on a per-country basis. To follow the terms of their licenses from the content providers the service must enforce geographical restrictions.
Sorry I wasn't clear enough, assume you are not making gigantic sequential accesses, but just say 64K writes that are otherwise "randomly" distributed. But 64K might be large enough to get you the sequential write performance speed. If you stripe that 64K access over 4 SDDs it might turn out that the performance you see for 16K "random" writes to each SDD is less than 1/4th the performance you were getting doing 64K "random" writes to one SSD. Which results in lower over all performance, which was my original point I'd think you could mostly avoid that by using a large stripe size for your striping so that small accesses go to just one drive and only large stuff gets interleaved.
What's I'd like to see is a push toward faster and cheaper, at the expense of capacity. Performance and capacity of SSDs tends to be at least somewhat linked, assuming the flash chips themselves are equal and you have a good controller more chips gives you both more speed and more capacity.
The average out-of-the-box OS should only take on the order of 10GB of space with a few more GB for common applications It probably "should" (i'm not a fan of the way software seems to keep growing in size with very little improvement in functionality) but it doesn't and in the short term there isn't really anything anyone except Microsoft can do about that.
The fact is that most users run windows and most buisness users also run office. Therefore any mainstream laptop/desktop needs at least to be able to accomodate the latest versions of windows, office, a virus checker and have a reasonable ammount of space left over for updates, user data, domain specific applications etc.
so lets add things up (based on official minimum requirements here)
Win7: 20GB (assuming that a machine with a SSD will be high-end in other ways too a 64-bit install is probably desirable) Office 2010: 3.5GB (for "professional plus" but other editions were only slightly lower) Antivirus: 0.5GB (for avg free)
So we are talking about 24GB just for the "basics". Add a bit more for optional components/OEM crapware/user data/domain specific applications and different definitions of gigabyte (Afaict MS use 2^30 while drive manufacturers use 10^9)and you are probably talking about 32GB as the absolute minimum a new machine can reasonably ship with and 48GB-64GB as more realistic.
Linux (at least debian, but I think other distros are similar) is much better in this regard with a standard desktop install including openoffice still being under 5GB but it remains a niche option.
One thing I have noticed is that office paper and non-fiction books are about the whitest paper commonly seen. Fiction books are less white and newspapers can't really be called white at all. Whiteness makes a document look more professional and is obviously an advantage when printing photos etc but i'm not at all convinced it's needed to make text comfortable to read.
I've even known people who need to have things printed on special paper and/or use filter sheets or filtered glasses because office paper was too white for them to see properly.
Would you consider the cost of beefing up internet connections and the equipment behind them to mitigate the impact of an ongoing DDOS to be legitimate damages? There is certainly no way a company that offers internet services can reasonably make themselves DDOS proof, they can only beef up their systems to handle more traffic than the attacker can generate (a quantity which is entirely dependent on the specific attacker)
if you do it right Java/C++ performance is so close as to not matter. The thing is java's language design seems to put design for performance and memory efficiency at odds with design for readability.
Specifically Java encourages you to create lots of tiny objects each with a seperate heap allocation, a pointer to a virtual method table, a pointer pointing at the object and probably other overheads too.
You can avoid this by various tricks but those tricks will all make the code less readable.
C and C++ on the other hand encourage you to create arrays of structures which have virtually zero overhead since they are allocated in a single block of memory and afaict C++ doesn't create anything like a virtual method table unless you specifically use things that need it as part of your object.
Another thing to consider when benchmarking is that GC changes WHEN the price for your actions is paid. With traditional memory allocation/deallocation you pay for cleaning stuff up as part of the process of using it. With GC you pay for it as a seperate GC process that happens at times outside of your control (either when the vm thinks it has time to spare or when it hits it's memory limit, whichever comes first).
Nothing has changed. What has changed is that with core 2 stuff you have the option of a nvidia chipset with integrated graphics that were better than intel integrated graphics while being physically smaller and lower power consumption than a discrete solution with it's own memory.
With current gen intel stuff that option is gone (though admittedly from a users point of view the fact that intel integrated graphics are better than they used to be somewhat makes up for it).
Afaict Nvidia was afraid that intel would take it even further and remove fast PCIe support from some of their chips so they persuaded the FTC to include a clause about that in the antitrust settlement.
it's still a year before they take them out Umm at least according to the migration document they will stop taking orders for xserves at the end of january 2011. That is less than 3 months away.
The "migration strategy" doesn't really tell you anything that a deployment guy with half a brain couldn't already figure out which is that apple has two other server options both of which are vastly inferior for many situations that involve xserves in a datacenter.
The mac pro can offer similar performance to the xserve but can't offer anywhere the density (yes mac pros can be configed with more grunt than xserves but this can't make up for the fact you only get two of them in 12U).
The mini is also lagging the xserve on density at least if you belive apples figure of two in 1U but more importantly it's horribly lacking in storage options. Your only choices are laptop drives (small, slow) or SSDs (fast but even smaller than laptop drives) and if you want SSDs you will have to fit them yourself since apple doesn't sell the mini with them. There is also the increased admin cost of the higher machine count.
And from the migration document it appears neither of them offers server features like lights out management and redundant power supplies.
You get much more oomph from four Mac Minis than you do from an XServe. On what do you base this claim?
CPU wise a mini maxes out at 2.66GHz dual core, an xserve maxes out at 2.93GHz 8 cores (two sockets each with a quad core processor).
Ram wise a mini maxes out at 8GB, an xserve maxes out at 48 GB (note: ram figures are maximums availible from apple, it may be possible to fit more especially in the xserve).
On storage the mini takes up to two drives while the xserve takes up to three but the mini's drives have to be laptop drives (and I think they have to be thin laptop drives) while the xserve takes full desktop drives.
So on CPU the xserve would appear to beat four minis, on ram the minis beat the xserve on paper but running four OS images will mean higher ram overheads. On storage it's a bit of a tossup depending on your needs but having more storage in the same box definiately gives more flexibility. The xserve also has far better options for external storage than the minis.
If you can print a car [slashdot.org] You can't print a car in any meaningful sense, you can print the bodywork but the important parts of the car still have to be made by more traditional methods.
I don't think apple or MS ceasing to exist tomorrow is a realistic fear unless there is a nuclear war or something. Even if MS did dissapear though i'm sure the governement could introduce an interim law that allowed vendors to keep supplying windows hardware until a permanent soloution could be found. With apple you would have to run your OS on hardware where it has never had any significant testing.
A much more immediate concern though is supply chain foul-ups. If the supply chain for a MS OEM fouls up you just go to another one. If the supply chain for apple fouls up you are SOL.
Given the choice between single vendor software on single vendor hardware VS single vendor software on multi-vendor hardware i'd definately prefer the latter.
and most especially the _testing_ costs for multiple regions. Heh you still need to test on multiple language versions of your OS even if all your text is 7-bit ascii. For example you need to figure out where you will be using the local conventions for decimal seperators and where you will be using the dot and make sure you use the right conversion routines in the right place. Failure to do this will lead to software that works fine on english systems but may break on continental european ones.
Another trick MS has pulled is making volume license windows sales upgrade/downgrade only and OEM ones non-transferable. So either you pay full retail upfront (few people do) or every time you replace a computer you pay for windows again even if you don't particulally want a new version (witness the number of machines that were available with vista or win7 downgraded to XP). Worse I belive OEM downgrades have now ceased so if you want XP now afaict you need to buy a machine with win7 and then buy a volume license copy to let you downgrade it.
This gives MS a practically guaranteed revenue stream.
I'd be pretty peeved if someone did this to my property, leaving it there is a huge liability, removing it and making good properly would be quite a hassle.
Slashdot Mirror
Why does every 'douche' keep comparing a Xeon to a consumer grade processor?
The reason is simply that apple refuses to sell a normal desktop. The choice of apple desktop machines comes down to
* mini: a small form factor machine built with latop parts and with specs comparable to a low end laptop
* imac: an all-in one with midrange desktop specs, no room for more hard drives or expansion cards, no USB3 or ESATA, and the options for screen size and graphics memeory tied together with the CPU options and hard drive options also tied into those choices to some extent.
* pro: a monster built out of components intended for dual socket server/workstation systems and with a pricetag to match.
All depends what education authorities do. If they decide to follow the IAU (which is likely as afaict there is no other organsation of similar standing pushing a rival definition) then that is what kids will be taught.
http://en.wikipedia.org/wiki/Eris_(dwarf_planet)
"Eris, formal designation 136199 Eris, is the most massive known dwarf planet in the Solar System and the ninth most massive body known to orbit the Sun directly. It is estimated to be approximately 2300-2400 km in diameter,[14] and 27% more massive than Pluto or about 0.27% of the Earth's mass."
It is the object that finally lead to pluto's "demotion" from planet to dwarf plant.
One thing I could see is having a "kit" of this sort of thing so you determine what the line type is ( by a combination of observation and knowing local standards) get the appropriate tap out of your kit, maybe set the blade depth and deploy it.
Afaict steam only got going at all because valve forced anyone who bought HL2 and wanted to play it and anyone who wanted to continue playing HL1 online to sign up for steam.
It could but they would have to plan to screw you in advance.
Whereas with online activation they can screw you at any time. Maybe they simply lose interest in thier older stuff. Maybe they are bought by a competitor whose sole aim is to kill them. While I don't think it has happened with games yet it has certainly happened with online music.
True, their hook still looks limited to a fairly small range of wiring types though, it relies on the ground/neutral being bare and a phase conductor twisted in with it covered in fairly limited insulation. Probablly also limited in terms of the overall size of the bundle.
It also looks prone to shorting if it grips the cable the wrong way.
so the user gets what is advertised.
ROFLMAO
Firstly the advertised speed for DSL is generally the speed on a "perfect" line. Unless you have a FTTP setup or similar scenario where all copper lines are very short speeds are often much lower than advertised.
Secondly while there may not be contention on the local loop there is almost certain to be contention somewhere either in the ADSL backend network or in the network that joins it to the internet.
"A lot" is relative, Given the small ammounts of data involved and the short range it should easilly be possible at under 100ma. USB ports are supposed to be able to provide 500ma.
So either the GP is lying or he has machines with REALLY REALLY shitty USB ports or the wireless receiver is using far more power than nessacery.
Unfortunately I suspect that due to contractual limitations the content available to international viewers will only be a subset of that available to british ones.
Geographic restrictions aren't really a software issue. IIRC amazon mp3 for example has geographic restrictions yet doesn't need any special client software at all.
Geographic restrictions on digital content come primerally from the fact that media is generally licensed on a per-country basis. To follow the terms of their licenses from the content providers the service must enforce geographical restrictions.
Sorry I wasn't clear enough, assume you are not making gigantic sequential accesses, but just say 64K writes that are otherwise "randomly" distributed. But 64K might be large enough to get you the sequential write performance speed. If you stripe that 64K access over 4 SDDs it might turn out that the performance you see for 16K "random" writes to each SDD is less than 1/4th the performance you were getting doing 64K "random" writes to one SSD. Which results in lower over all performance, which was my original point
I'd think you could mostly avoid that by using a large stripe size for your striping so that small accesses go to just one drive and only large stuff gets interleaved.
What's I'd like to see is a push toward faster and cheaper, at the expense of capacity.
Performance and capacity of SSDs tends to be at least somewhat linked, assuming the flash chips themselves are equal and you have a good controller more chips gives you both more speed and more capacity.
The average out-of-the-box OS should only take on the order of 10GB of space with a few more GB for common applications
It probably "should" (i'm not a fan of the way software seems to keep growing in size with very little improvement in functionality) but it doesn't and in the short term there isn't really anything anyone except Microsoft can do about that.
The fact is that most users run windows and most buisness users also run office. Therefore any mainstream laptop/desktop needs at least to be able to accomodate the latest versions of windows, office, a virus checker and have a reasonable ammount of space left over for updates, user data, domain specific applications etc.
so lets add things up (based on official minimum requirements here)
Win7: 20GB (assuming that a machine with a SSD will be high-end in other ways too a 64-bit install is probably desirable)
Office 2010: 3.5GB (for "professional plus" but other editions were only slightly lower)
Antivirus: 0.5GB (for avg free)
So we are talking about 24GB just for the "basics". Add a bit more for optional components/OEM crapware/user data/domain specific applications and different definitions of gigabyte (Afaict MS use 2^30 while drive manufacturers use 10^9)and you are probably talking about 32GB as the absolute minimum a new machine can reasonably ship with and 48GB-64GB as more realistic.
Linux (at least debian, but I think other distros are similar) is much better in this regard with a standard desktop install including openoffice still being under 5GB but it remains a niche option.
One thing I have noticed is that office paper and non-fiction books are about the whitest paper commonly seen. Fiction books are less white and newspapers can't really be called white at all. Whiteness makes a document look more professional and is obviously an advantage when printing photos etc but i'm not at all convinced it's needed to make text comfortable to read.
I've even known people who need to have things printed on special paper and/or use filter sheets or filtered glasses because office paper was too white for them to see properly.
Would you consider the cost of beefing up internet connections and the equipment behind them to mitigate the impact of an ongoing DDOS to be legitimate damages? There is certainly no way a company that offers internet services can reasonably make themselves DDOS proof, they can only beef up their systems to handle more traffic than the attacker can generate (a quantity which is entirely dependent on the specific attacker)
if you do it right Java/C++ performance is so close as to not matter.
The thing is java's language design seems to put design for performance and memory efficiency at odds with design for readability.
Specifically Java encourages you to create lots of tiny objects each with a seperate heap allocation, a pointer to a virtual method table, a pointer pointing at the object and probably other overheads too.
You can avoid this by various tricks but those tricks will all make the code less readable.
C and C++ on the other hand encourage you to create arrays of structures which have virtually zero overhead since they are allocated in a single block of memory and afaict C++ doesn't create anything like a virtual method table unless you specifically use things that need it as part of your object.
Another thing to consider when benchmarking is that GC changes WHEN the price for your actions is paid. With traditional memory allocation/deallocation you pay for cleaning stuff up as part of the process of using it. With GC you pay for it as a seperate GC process that happens at times outside of your control (either when the vm thinks it has time to spare or when it hits it's memory limit, whichever comes first).
Nothing has changed.
What has changed is that with core 2 stuff you have the option of a nvidia chipset with integrated graphics that were better than intel integrated graphics while being physically smaller and lower power consumption than a discrete solution with it's own memory.
With current gen intel stuff that option is gone (though admittedly from a users point of view the fact that intel integrated graphics are better than they used to be somewhat makes up for it).
Afaict Nvidia was afraid that intel would take it even further and remove fast PCIe support from some of their chips so they persuaded the FTC to include a clause about that in the antitrust settlement.
it's still a year before they take them out
Umm at least according to the migration document they will stop taking orders for xserves at the end of january 2011. That is less than 3 months away.
The "migration strategy" doesn't really tell you anything that a deployment guy with half a brain couldn't already figure out which is that apple has two other server options both of which are vastly inferior for many situations that involve xserves in a datacenter.
The mac pro can offer similar performance to the xserve but can't offer anywhere the density (yes mac pros can be configed with more grunt than xserves but this can't make up for the fact you only get two of them in 12U).
The mini is also lagging the xserve on density at least if you belive apples figure of two in 1U but more importantly it's horribly lacking in storage options. Your only choices are laptop drives (small, slow) or SSDs (fast but even smaller than laptop drives) and if you want SSDs you will have to fit them yourself since apple doesn't sell the mini with them. There is also the increased admin cost of the higher machine count.
And from the migration document it appears neither of them offers server features like lights out management and redundant power supplies.
You get much more oomph from four Mac Minis than you do from an XServe.
On what do you base this claim?
CPU wise a mini maxes out at 2.66GHz dual core, an xserve maxes out at 2.93GHz 8 cores (two sockets each with a quad core processor).
Ram wise a mini maxes out at 8GB, an xserve maxes out at 48 GB (note: ram figures are maximums availible from apple, it may be possible to fit more especially in the xserve).
On storage the mini takes up to two drives while the xserve takes up to three but the mini's drives have to be laptop drives (and I think they have to be thin laptop drives) while the xserve takes full desktop drives.
So on CPU the xserve would appear to beat four minis, on ram the minis beat the xserve on paper but running four OS images will mean higher ram overheads. On storage it's a bit of a tossup depending on your needs but having more storage in the same box definiately gives more flexibility. The xserve also has far better options for external storage than the minis.
If you can print a car [slashdot.org]
You can't print a car in any meaningful sense, you can print the bodywork but the important parts of the car still have to be made by more traditional methods.
I don't think apple or MS ceasing to exist tomorrow is a realistic fear unless there is a nuclear war or something. Even if MS did dissapear though i'm sure the governement could introduce an interim law that allowed vendors to keep supplying windows hardware until a permanent soloution could be found. With apple you would have to run your OS on hardware where it has never had any significant testing.
A much more immediate concern though is supply chain foul-ups. If the supply chain for a MS OEM fouls up you just go to another one. If the supply chain for apple fouls up you are SOL.
Given the choice between single vendor software on single vendor hardware VS single vendor software on multi-vendor hardware i'd definately prefer the latter.
and most especially the _testing_ costs for multiple regions.
Heh you still need to test on multiple language versions of your OS even if all your text is 7-bit ascii. For example you need to figure out where you will be using the local conventions for decimal seperators and where you will be using the dot and make sure you use the right conversion routines in the right place. Failure to do this will lead to software that works fine on english systems but may break on continental european ones.
Another trick MS has pulled is making volume license windows sales upgrade/downgrade only and OEM ones non-transferable. So either you pay full retail upfront (few people do) or every time you replace a computer you pay for windows again even if you don't particulally want a new version (witness the number of machines that were available with vista or win7 downgraded to XP). Worse I belive OEM downgrades have now ceased so if you want XP now afaict you need to buy a machine with win7 and then buy a volume license copy to let you downgrade it.
This gives MS a practically guaranteed revenue stream.
I'd be pretty peeved if someone did this to my property, leaving it there is a huge liability, removing it and making good properly would be quite a hassle.