You missed the part in the story about Psychohistory being constantly tweaked from the top BY A MIND READING ROBOT.
It is much easier to take a 'thinking' sample of a population if you can walk into a crowded area, such as those on the most populated planet in the galaxy, AND READ EVERYONES THOUGHTS.
Indeed. It has been my experience that when crypto writers move their libs from C to well optimized x86 assembly language they get at least 2x performance boost.
These guys are getting 4x, but only on a fairly powerful GTX 480 GPU. How will a typical mobile GPU's compare? Probably even slower than the CPU, right? This article makes me sad.
Doing a bit of math on the C-Ray benchmark numbers, which is a floating point heavy benchmark, it appears that a Bulldozer module (2 cores with 2 alu units and only 1 fpu unit) beats out a sandy bridge core in performance per clock. Essentially, each module of the bulldozer completes 0.133% of the C-Ray test workload per billion clock cycles while each sandy bridge does 0.122% of the workload per billion clock cycles.
Each bulldozer module is/willbe "advertised" as 2 cores, so for floating point work 2 sandy bridge cores will easily trump 2 bulldozer cores.
This is not surprising revelation, as many had guessed that bulldozer would be weak on FPU work while simultaneously being extremely strong at integer work.
The upshot of this new era is that you arent really buying cores anymore, you are either buying FPU's or ALU's. An 8 core bulldozer will have 4 FPU's that are each stronger than the 4 FPU's in a 4 core sandy bridge, but what remains to be seen is the price difference.
The only legit (that I know of) in-the-wild Bulldozer benchmark is a 1.8 GHz dual chip (2 x 16 = 32 cores) server chip setup using the Phoronix benchmark suite (search results)
It is likely that these sample chips are as much a test of the new 32nm fab as they are a test of the new cpu architecture, and definitely not a test of how quickly they can be clocked.
AMD was selling their chips for the same prices prior to the 2600K release, when Intel had only two chips that were even in the ballpark on performance for the cost.
Now suddenly you claim that AMD has to sell for those same prices because they cant compete with the 2600K? Do you often make it up as you go along?
Most parallel problems can be defined in terms that require no locks within the inner loops, such as the class of problems mentioned in the grandparent (image recognition..)
I find that people that dont know shit about algorithms always think that the "hard" parts of parallel strategies somehow magically apply to most highly parallel problems... which is stupid.. but there you are.
Don't bother replying until you have mastered a functional language to the point where the reason I am asking you to master one dawns on you.
What's holding it up? It's the lack of hi-def eyepieces isn't it?
Its the lack of eyes that can focus on things within an inch or two in front of them. Nearly every young person thinks that they can, until you have them try.
So here you are, telling us about the terrible information sources that people listen to, when you yourself are just as bad of an information source.
I can only assume that you are one of those people that thinks that anyone who dislikes Obama, for any reason at all, is a racist. Why can I only assume this? Well, its because you just declared it to be so. You didn't leave any room for you to wiggle out of it either. The best save you can make here is to say that you made a mistake and gave grossly erroneous information, which is only marginally better than the alternative.
Then why is it that Opera runs significantly better on low memory systems than Firefox-with-no-extension does?
Now, add extensions to Firefox. Its horrible in low-memory scenarios because *it* is the bloated one. The fact that Firefox sometimes uses significantly less memory when there is actually gobs of memory available is also evidence of a problem with Firefox.
But it should be noted that these mobile CPU's are also heading towards more power, not less.
At some point batteries will be so good that power needs become trivial beyond the "how much does it cost to charge it?" question, which isnt a question anybody is currently asking (people only ask "how long will a charge last?") So in the end it will be all about performance.
Intel has a problem in that its sort of the middle-solution in his space. Atom's arent low enough in power draw to compete with ARM's, they cannot offer the graphics performance of the AMD Bobcats or nVidia's ARM solutions, and they also can't compete with anyone on price in the mobile space now. The effects of Bobcat will take about a year to really sink into the market, as nobody choosing x86 for their new product will be considering Atom unless there is some sort of backroom bribe/kickback involved.
The problem with your "younger and younger" argument is that the slope is slippery in every direction, and not just the direction that you cherry picked. Older and older.. Fatter and fatter.. Thinner and thinner.. Whiter and whiter.. Blacker and blacker.. Hairier and Hairier.. Uglier and uglier.. Shorter and shorter.. taller and taller.. Flatter and flatter.. Overbites.. Underbites.. we could go on and on..
So, do pictures of shirt girls turn people into midget fetishists?
Intel NEEDS those specialized instructions added on to keep pace.
Note that Intel's compilers refuse to use those instructions when their output runs on AMD's and, unfortunately, the popular scientific libraries are all compiled with ones of Intel's compilers (ICC or their Fortran compiler) and only use the SIMD paths if they see "GenuineIntel" output from CPUID.
One of the most renowned software optimization experts studies this in detail in his blog.
This is certainly due to Intel not really refreshing its server lines at all, focusing mainly on the desktop space, while AMD has steadily updated its server lines.
Lets not forget that AMD is also about to unveil its bulldozer cores, while Intel has recently updated its desktop chips. Until this year, Intel had an extremely hard time competing in performance per $ in the desktop space, and expect that even if bulldozer doesnt match i7 levels it will again regain the performance per dollar crown that it had up until this year.
Certainly if you are spending $1000 on a CPU, you would go Intel today. But if you are spending $250 on a CPU, the choice today isnt so clear at all.
For nice fast RAM access, doesn't the new AMD Fusion GPU share the same silicon with the CPU anyway?
Indeed, and the Fusion chips are trouncing Atom-based solutions in graphics benchmarks mainly for this very reason.
The problem tho is it cant readily be applied to more mainstream desktop solutions because then you have CPU and GPU fighting over precious memory bandwidth. For netbooks and the like, it works well because GPU performance isnt expected to match even midrange cards, so only a fraction of DDR2/DDR3 bandwidth is acceptable. Even midrange desktop graphics cards blow the doors off of DDR3 memory bandwidth, so this really isnt the route to go.
..and before you think of it, you cant just up the memory bandwidth of the CPU to desktop GPU levels because that goes hand-in-hand with increased latency (the reason that Intel failed to develop Larrabee.) GPU's dont suffer much from increased latency, but CPU's and their random access patterns suffer greatly from it.
For starters, how about reasonable terms and prices? The terms seems very awkward, the licensees only get to use a specific version of Android.
New versions of android may infringe more than the one they are are acquiring a license for. Microsoft is offering B&N a chance to license the specific IP rather than a blanket coverage of "Android version X" but is asking for an NDA (a pretty standard thing) which B&N refuses to sign onto.
This for a price that is double the per unit price of Windows Phone 7.
..and is infinite the amount of the per unit price Microsoft charges in licensing fees for the pound of american cheese I just purchased. If WP7 also ran android (and thus used the exact same set of IP,) you would have some sort of point. Android clearly has more features than WP7.
Why is it that the same people that say that WP7 is so horrible, are also the ones so damn surprised when licensing it is cheaper than licensing the be-all-end-all of mobile OS's known as Android?
Presumably to encourage people to innovate, create their own inventions and thereby profit instead of failing to do so and instead setting up a racket to extort the work of others.
One method of profiting is to license your inventions to others.
B&N may have a case that the patents in question do not cover their device, but that is different from their claim that patents shouldnt be used to "extract" license fees from them. They might also have a case that the license fees are discriminatory, but that seems unlikely considering that other companies making android devices (some of B&N's competition) are paying those very same fees or cross-licensing their own IP with Microsoft.
'Microsoft intends to utilize its patents to control the activities of and extract fees from the designers, developers, and manufacturers of devices, including tablets, eReaders, and other mobile devices, that employ the Android Operating System.'
It would be a strange system where a patent holder couldn't do these things. What precisely does B&N think patents are for?
Slashdot Mirror
You missed the part in the story about Psychohistory being constantly tweaked from the top BY A MIND READING ROBOT.
It is much easier to take a 'thinking' sample of a population if you can walk into a crowded area, such as those on the most populated planet in the galaxy, AND READ EVERYONES THOUGHTS.
Indeed. It has been my experience that when crypto writers move their libs from C to well optimized x86 assembly language they get at least 2x performance boost.
These guys are getting 4x, but only on a fairly powerful GTX 480 GPU. How will a typical mobile GPU's compare? Probably even slower than the CPU, right? This article makes me sad.
All the .NET languages, including PowerShell, supports COM.
Admittedly COM is sort of out-of-place, but its all relatively seamless none-the-less.
If the HDMI is on one end, and the USB is on the other, is this thing battery powered?
Doing a bit of math on the C-Ray benchmark numbers, which is a floating point heavy benchmark, it appears that a Bulldozer module (2 cores with 2 alu units and only 1 fpu unit) beats out a sandy bridge core in performance per clock. Essentially, each module of the bulldozer completes 0.133% of the C-Ray test workload per billion clock cycles while each sandy bridge does 0.122% of the workload per billion clock cycles.
Each bulldozer module is/willbe "advertised" as 2 cores, so for floating point work 2 sandy bridge cores will easily trump 2 bulldozer cores.
This is not surprising revelation, as many had guessed that bulldozer would be weak on FPU work while simultaneously being extremely strong at integer work.
The upshot of this new era is that you arent really buying cores anymore, you are either buying FPU's or ALU's. An 8 core bulldozer will have 4 FPU's that are each stronger than the 4 FPU's in a 4 core sandy bridge, but what remains to be seen is the price difference.
Visual Programming is more than a dream... its a reality... just not in the general-purpose sense.
Plenty of domain-specific languages use a visual paradigm of stringing things together, and one of the most successful is LabVIEW.
The key to VP is DOMAIN-SPECIFIC. In a general-purpose form, it just isnt workable.
The only legit (that I know of) in-the-wild Bulldozer benchmark is a 1.8 GHz dual chip (2 x 16 = 32 cores) server chip setup using the Phoronix benchmark suite (search results)
It is likely that these sample chips are as much a test of the new 32nm fab as they are a test of the new cpu architecture, and definitely not a test of how quickly they can be clocked.
AMD was selling their chips for the same prices prior to the 2600K release, when Intel had only two chips that were even in the ballpark on performance for the cost.
Now suddenly you claim that AMD has to sell for those same prices because they cant compete with the 2600K? Do you often make it up as you go along?
But back in the Athlon 64 days, the numbers used to have actual meaning and were related to Intel's processor performance.
An Athlon 64 3800+ performed on par with a 3800 MHz Pentium 4
Indeed, and note that Hawking is talking not only of time, but of space too. Space-time could not exist before inflation.
Most parallel problems can be defined in terms that require no locks within the inner loops, such as the class of problems mentioned in the grandparent (image recognition..)
I find that people that dont know shit about algorithms always think that the "hard" parts of parallel strategies somehow magically apply to most highly parallel problems... which is stupid.. but there you are.
Don't bother replying until you have mastered a functional language to the point where the reason I am asking you to master one dawns on you.
What's holding it up? It's the lack of hi-def eyepieces isn't it?
Its the lack of eyes that can focus on things within an inch or two in front of them. Nearly every young person thinks that they can, until you have them try.
Racists (which is what "birthers" are) ...
So here you are, telling us about the terrible information sources that people listen to, when you yourself are just as bad of an information source.
I can only assume that you are one of those people that thinks that anyone who dislikes Obama, for any reason at all, is a racist. Why can I only assume this? Well, its because you just declared it to be so. You didn't leave any room for you to wiggle out of it either. The best save you can make here is to say that you made a mistake and gave grossly erroneous information, which is only marginally better than the alternative.
Then why is it that Opera runs significantly better on low memory systems than Firefox-with-no-extension does?
Now, add extensions to Firefox. Its horrible in low-memory scenarios because *it* is the bloated one. The fact that Firefox sometimes uses significantly less memory when there is actually gobs of memory available is also evidence of a problem with Firefox.
That is exactly it.
But it should be noted that these mobile CPU's are also heading towards more power, not less.
At some point batteries will be so good that power needs become trivial beyond the "how much does it cost to charge it?" question, which isnt a question anybody is currently asking (people only ask "how long will a charge last?") So in the end it will be all about performance.
Intel has a problem in that its sort of the middle-solution in his space. Atom's arent low enough in power draw to compete with ARM's, they cannot offer the graphics performance of the AMD Bobcats or nVidia's ARM solutions, and they also can't compete with anyone on price in the mobile space now. The effects of Bobcat will take about a year to really sink into the market, as nobody choosing x86 for their new product will be considering Atom unless there is some sort of backroom bribe/kickback involved.
The problem with your "younger and younger" argument is that the slope is slippery in every direction, and not just the direction that you cherry picked. Older and older.. Fatter and fatter.. Thinner and thinner.. Whiter and whiter.. Blacker and blacker.. Hairier and Hairier.. Uglier and uglier.. Shorter and shorter.. taller and taller.. Flatter and flatter.. Overbites.. Underbites.. we could go on and on..
So, do pictures of shirt girls turn people into midget fetishists?
Intel NEEDS those specialized instructions added on to keep pace.
Note that Intel's compilers refuse to use those instructions when their output runs on AMD's and, unfortunately, the popular scientific libraries are all compiled with ones of Intel's compilers (ICC or their Fortran compiler) and only use the SIMD paths if they see "GenuineIntel" output from CPUID.
One of the most renowned software optimization experts studies this in detail in his blog.
Indeed, AMD is still crushing Intel's 4-chip solutions in performance
This is certainly due to Intel not really refreshing its server lines at all, focusing mainly on the desktop space, while AMD has steadily updated its server lines.
Lets not forget that AMD is also about to unveil its bulldozer cores, while Intel has recently updated its desktop chips. Until this year, Intel had an extremely hard time competing in performance per $ in the desktop space, and expect that even if bulldozer doesnt match i7 levels it will again regain the performance per dollar crown that it had up until this year.
Certainly if you are spending $1000 on a CPU, you would go Intel today. But if you are spending $250 on a CPU, the choice today isnt so clear at all.
For nice fast RAM access, doesn't the new AMD Fusion GPU share the same silicon with the CPU anyway?
Indeed, and the Fusion chips are trouncing Atom-based solutions in graphics benchmarks mainly for this very reason.
..and before you think of it, you cant just up the memory bandwidth of the CPU to desktop GPU levels because that goes hand-in-hand with increased latency (the reason that Intel failed to develop Larrabee.) GPU's dont suffer much from increased latency, but CPU's and their random access patterns suffer greatly from it.
The problem tho is it cant readily be applied to more mainstream desktop solutions because then you have CPU and GPU fighting over precious memory bandwidth. For netbooks and the like, it works well because GPU performance isnt expected to match even midrange cards, so only a fraction of DDR2/DDR3 bandwidth is acceptable. Even midrange desktop graphics cards blow the doors off of DDR3 memory bandwidth, so this really isnt the route to go.
Because Microsoft does not own Android
Microsoft claims to own parts of Android. Thats the whole purpose of the lawsuit.
Did you forget that thats what the lawsuit is about?
For starters, how about reasonable terms and prices? The terms seems very awkward, the licensees only get to use a specific version of Android.
New versions of android may infringe more than the one they are are acquiring a license for. Microsoft is offering B&N a chance to license the specific IP rather than a blanket coverage of "Android version X" but is asking for an NDA (a pretty standard thing) which B&N refuses to sign onto.
This for a price that is double the per unit price of Windows Phone 7.
Why is it that the same people that say that WP7 is so horrible, are also the ones so damn surprised when licensing it is cheaper than licensing the be-all-end-all of mobile OS's known as Android?
considering that the patents are questionable and don't even cover android
Thats for a court, and not you, to decide. If they so obviously don't cover B&N's devices then they have nothing to worry about.
What other way is Microsoft supposed to enforce their IP besides filing suit against an infringer that refuses to license?
Presumably to encourage people to innovate, create their own inventions and thereby profit instead of failing to do so and instead setting up a racket to extort the work of others.
One method of profiting is to license your inventions to others.
B&N may have a case that the patents in question do not cover their device, but that is different from their claim that patents shouldnt be used to "extract" license fees from them. They might also have a case that the license fees are discriminatory, but that seems unlikely considering that other companies making android devices (some of B&N's competition) are paying those very same fees or cross-licensing their own IP with Microsoft.
'Microsoft intends to utilize its patents to control the activities of and extract fees from the designers, developers, and manufacturers of devices, including tablets, eReaders, and other mobile devices, that employ the Android Operating System.'
It would be a strange system where a patent holder couldn't do these things. What precisely does B&N think patents are for?