Also you save lots of power, because you dont have vast amounts of cache (though this may effect performance for some cases), and the architecture is much simpler, with only the instructions needed (not decades of x86 legacy). last time i was in an HPC machine room, there was a pile of not very old xeon servers in the corner, still fast machines, but not worth the energy cost to run them compared to the new servers.
>Adapteva claims more than 9. So they're twice as good as IBM? Really?
Epiphany has a customised core that only has instructions useful for floating point,and fetching data. They have also changed from hierarchy of caches to a more network like method of moving data around. Its a scale of general purposeness, there things that a full CPU can do well that will choke an ephiphany. there are things an epiphany can do well that will choke a GPU.
i agree that the 64core versions is far more exciting than the 16core version. i guess maybe they think there is a lot higher risk there (they have already made and tested a small number of 16core chips, http://www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf )
>If you're a small startup, then you'll be bound to 100nm processes (at best), and have to use automated layouts (not the hand-optimized ones e.g. Intel uses). Both reduce performance, increase power intake.
they are planing to use "GlobalFoundries’ 28nm SLP technology" www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf
raw GFLOPS is not everything. GPUs need very SIMD work to reach the theorical limits. for a lot of graphcis work thats fine. for some scientific work its good too. for other promblems, where the code has lots of branches in it, you end up with cores waiting for other cores to do different branches (look up wavefront for more info).
the $99 on the kickstart gets you a dual core arm A9, with a 16core epiphany processor (a dual core arm A9 devboard costs in the region of $100 normally). no one is saying that the 16core version is going to beat an i7 or a GPU, it just puts a lot of CPU power into a tiny area of silicon, that uses a tiny amount of power. its a step on the road to 1024 and 4096 core versions, that will sit on PCI-E cards, and challange the performance of a GPU for a fraction the power consumption (most of the cost of HPC).
>As for apples to apples, It's vaporware specs read similar to the old printout of the vaporware specs for the Propeller2 from microchip inc on my desk.
they have working 16 core silicon. they shared the cost of a 65nm wafer with other companies small run asics. this lowers the entry cost to making silicon, but gives a crazy high per unit cost. if they raise enough money to do a full wafer at 28nm, then it becomes cost effective. there are intersting details and numbers on page 3 of http://www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf
the devboard has a Dual-core ARM A9, so more like a pandaboard. even if you ignore the co-processor they are offering a lot for $99.
its interesting to compare the epiphany processor to a GPU. both give you lots of cores, GPUs get up ino the hundreds, epiphany is meant to scale to 4000. But a GPU is highly opitmised for graphics, and applying identical operations to millions of data values. in a GPU groups of core (typically 32) operate as a wavefront, if the code branches on an if stament, then the cores that get the else branch have to wait until the ones that follow the if finish.
epiphany has independant cores. you can send them each a different program. so for a much wider set of algorithms you can get efficient speedups. in a way it is more like the xeon phi, but without making each core a full x86 compatible processor.
I swapped all my bulbs to CFL years ago. when my energy company send me 3 free CFLs a while back they just went in a draw, because i had no incandescents left to replace. CFLs cost about £1 these days, and save ~£5-20 per year per bulb (depending on usage).
so 30 Billion watts for something like 2 billion internet users. That's not to shoddy. Probably similar to the amount of power used at the client end (though that ought to be falling as people move from desktops to laptops and tablets). Global power usage is 15 Tera Watts, so data centres are about 0.2% of energy use.
So, what would do more to save the world, made data centers twice as efficient, or make transport 5% more efficient. Unfortunately if the former that's easier.
ITER is a prototype for a 0.5GW machine. Now first of their kind machines cost stupidly more than general production machines (R&D is not cheap). So the hope would be that you can one day build a 1GW reactor for something less than $10 Billion. Then it is cost effective even if you are paying interest at 10%. (for nuclear (and renewables) you can neglect fuel costs).
I think this mostly due to the PowerVR SGX graphics engine (remember the gma500 poulsbo). for the gma500 intel made a binary linux driver that did not impress anyone. I guess for clovertrail they are just not bothering with releasing a binary driver.
So it might work fine as a CPU, but have no graphics acceleration. however for a tablet chip that cannot play video or composite a desktop in software, it might be effectively useless.
Nuclear power has an very low deaths per kWh, even when you include chernobyl, 3mile island and fukushima ( http://nextbigfuture.com/2011/03/deaths-per-twh-by-energy-source.html ). chernobyl is a terrible design (as the coolant boils, the reaction goes faster. fail), nothing like that could happen in any modern (by which i mean anything made in last few decades).
Switching to any other form of power generation will cost lives.
From a environmental point of view, suppose japan can build enough wind and solar to replace nuclear (big job on the scale of a war effort), if they did that along side nuclear they would be reducing carbon emissions. if you do it instead of nuclear then you are standing still. Now take a look at this http://www.esrl.noaa.gov/gmd/ccgg/trends/ and have a read of IPCC, and explain how we are going to not hit 400 ppm.
lets take 70,000 kg as a typical take off mass (wikipedia 737 page has numbers from 50,000 to 80,000). and we are saving 15kg so we have a weight saving of about 0.02%, and presumably a similar fuel/emission saving.
claims that growth in air travel will be offset by improved efficiency do not look very strong if this sort of thing is considered press release worthy.
> If you hate Gnome 3, it is better to switch to XFCE, or LXDE, or KDE than to cling to an obsolete OS.
i respectfully disagree, GNOME2 (or MATE) has a lot of features that are missing in XFCE and LXDE. Its also had millions of man hours more testing, and so is very stable and works deep into corner cases. In a few years time XFCE or LXDE may catch up with GNOME2 in terms of features (if thats what they want to do (i am not sure it is)), or GNOME3 might be customisable to please the grumpy old men^{tm}. Until then i dont see any harm in sticking with what works. There are several distros where it easy to run MATE with a new kernel and everything else.
Actually MATE development is quite active, it has picked up some interesting features, and is gradually porting to newer libraries. it can't be to long before people get grumpy with the changes in MATE and create a new fork:-)
i suspect that quite a few people downloading debian (and linux distros in general) do not realise this. if the architectures were labelled more clearly a lot more people would probably be using the 64 bit version, and enjoying a significant performance increase.
Most benchmarks i have seen have much better speed up than 4%. eg http://www.scribd.com/doc/363677/Benchmarks-AMD64-in-32bit-mode-vs-64bit-mode-Ubuntu
Those LFS benchmarks are comparing build times. ie building a 64bit exectuable on a 64bit system is 4% faster than building a 32bit executable on a 32bit system. by that logic you should compile everything with -O0 or -O1 because thats much quicker to build.
The plausible explanation that i have heard for the big speed up is that for x86_64 you can assume that the CPU has things like SSE2, where as if you build i386 packages you might want to support older chips pentium 2 or 3, or chips like the via that lack some newer instructions. if you build everything with -march=native then you can enable the extended instruction sets on 32bit, but given the context of the discusion debian can't enable SSE2 on their i386 packages.
Slashdot Mirror
you need to fill the SIMD units to get theoretical performance on a i7 (or similar). with epiphany it may be easier to get close to the theoretical FLOPS
http://www.adapteva.com/white-papers/ten-myths-debunked-by-the-epiphany-iv-64-core-accelerator-chip/
Also you save lots of power, because you dont have vast amounts of cache (though this may effect performance for some cases), and the architecture is much simpler, with only the instructions needed (not decades of x86 legacy). last time i was in an HPC machine room, there was a pile of not very old xeon servers in the corner, still fast machines, but not worth the energy cost to run them compared to the new servers.
all the compiler stuff is merged into GCC as of 4.7 http://gcc.gnu.org/gcc-4.7/changes.html
>Adapteva claims more than 9. So they're twice as good as IBM? Really?
Epiphany has a customised core that only has instructions useful for floating point,and fetching data. They have also changed from hierarchy of caches to a more network like method of moving data around. Its a scale of general purposeness, there things that a full CPU can do well that will choke an ephiphany. there are things an epiphany can do well that will choke a GPU.
your wish on the clusters has been answered. http://www.kickstarter.com/projects/adapteva/parallella-a-supercomputer-for-everyone/posts/323994
i agree that the 64core versions is far more exciting than the 16core version. i guess maybe they think there is a lot higher risk there (they have already made and tested a small number of 16core chips, http://www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf )
>If you're a small startup, then you'll be bound to 100nm processes (at best), and have to use automated layouts (not the hand-optimized ones e.g. Intel uses). Both reduce performance, increase power intake.
they are planing to use "GlobalFoundries’ 28nm SLP technology" www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf
raw GFLOPS is not everything. GPUs need very SIMD work to reach the theorical limits. for a lot of graphcis work thats fine. for some scientific work its good too. for other promblems, where the code has lots of branches in it, you end up with cores waiting for other cores to do different branches (look up wavefront for more info).
the $99 on the kickstart gets you a dual core arm A9, with a 16core epiphany processor (a dual core arm A9 devboard costs in the region of $100 normally). no one is saying that the 16core version is going to beat an i7 or a GPU, it just puts a lot of CPU power into a tiny area of silicon, that uses a tiny amount of power. its a step on the road to 1024 and 4096 core versions, that will sit on PCI-E cards, and challange the performance of a GPU for a fraction the power consumption (most of the cost of HPC).
>As for apples to apples, It's vaporware specs read similar to the old printout of the vaporware specs for the Propeller2 from microchip inc on my desk.
they have working 16 core silicon. they shared the cost of a 65nm wafer with other companies small run asics. this lowers the entry cost to making silicon, but gives a crazy high per unit cost. if they raise enough money to do a full wafer at 28nm, then it becomes cost effective. there are intersting details and numbers on page 3 of http://www.adapteva.com/wp-content/uploads/2011/06/adapteva_mpr.pdf
the devboard has a Dual-core ARM A9, so more like a pandaboard. even if you ignore the co-processor they are offering a lot for $99.
its interesting to compare the epiphany processor to a GPU. both give you lots of cores, GPUs get up ino the hundreds, epiphany is meant to scale to 4000. But a GPU is highly opitmised for graphics, and applying identical operations to millions of data values. in a GPU groups of core (typically 32) operate as a wavefront, if the code branches on an if stament, then the cores that get the else branch have to wait until the ones that follow the if finish.
epiphany has independant cores. you can send them each a different program. so for a much wider set of algorithms you can get efficient speedups. in a way it is more like the xeon phi, but without making each core a full x86 compatible processor.
libreoffice does a hybrid PDF, that embeds the ODF file as well. so when you import it back into libreoffice everything is preserved.
+1
I swapped all my bulbs to CFL years ago. when my energy company send me 3 free CFLs a while back they just went in a draw, because i had no incandescents left to replace. CFLs cost about £1 these days, and save ~£5-20 per year per bulb (depending on usage).
so 30 Billion watts for something like 2 billion internet users. That's not to shoddy. Probably similar to the amount of power used at the client end (though that ought to be falling as people move from desktops to laptops and tablets). Global power usage is 15 Tera Watts, so data centres are about 0.2% of energy use.
So, what would do more to save the world, made data centers twice as efficient, or make transport 5% more efficient. Unfortunately if the former that's easier.
also 30% of the control group got tumours. obviously not eating GM gives you cancer.
Mark Lynas ( https://twitter.com/mark_lynas ) picked some interesting points out of the paper (and links to a mirror of the paper).
30% of the 20 control rats also got tumours.
but with free software (firefox, gimp, gnash...) we can keep windows XP alive for ever. hooray, screw MSTF. oh, wait a minute...
$2 Billion is pretty small on scale of what the planet spends on energy. For rough maths a GW of power, for 1 year at 10c/kWh is ~ $1Billion ( http://www.wolframalpha.com/input/?i=1+GW+*+1+year+*+(0.1USD+%2F+kWh) ).
ITER is a prototype for a 0.5GW machine. Now first of their kind machines cost stupidly more than general production machines (R&D is not cheap). So the hope would be that you can one day build a 1GW reactor for something less than $10 Billion. Then it is cost effective even if you are paying interest at 10%. (for nuclear (and renewables) you can neglect fuel costs).
i thought Linux supported more archs than netbsd these days. (maybe thats only if you dont include toasters)
I think this mostly due to the PowerVR SGX graphics engine (remember the gma500 poulsbo). for the gma500 intel made a binary linux driver that did not impress anyone. I guess for clovertrail they are just not bothering with releasing a binary driver.
So it might work fine as a CPU, but have no graphics acceleration. however for a tablet chip that cannot play video or composite a desktop in software, it might be effectively useless.
Nuclear power has an very low deaths per kWh, even when you include chernobyl, 3mile island and fukushima ( http://nextbigfuture.com/2011/03/deaths-per-twh-by-energy-source.html ). chernobyl is a terrible design (as the coolant boils, the reaction goes faster. fail), nothing like that could happen in any modern (by which i mean anything made in last few decades).
Switching to any other form of power generation will cost lives.
From a environmental point of view, suppose japan can build enough wind and solar to replace nuclear (big job on the scale of a war effort), if they did that along side nuclear they would be reducing carbon emissions. if you do it instead of nuclear then you are standing still. Now take a look at this http://www.esrl.noaa.gov/gmd/ccgg/trends/ and have a read of IPCC, and explain how we are going to not hit 400 ppm.
lets take 70,000 kg as a typical take off mass (wikipedia 737 page has numbers from 50,000 to 80,000). and we are saving 15kg
so we have a weight saving of about 0.02%, and presumably a similar fuel/emission saving.
claims that growth in air travel will be offset by improved efficiency do not look very strong if this sort of thing is considered press release worthy.
> If you hate Gnome 3, it is better to switch to XFCE, or LXDE, or KDE than to cling to an obsolete OS.
i respectfully disagree, GNOME2 (or MATE) has a lot of features that are missing in XFCE and LXDE. Its also had millions of man hours more testing, and so is very stable and works deep into corner cases. In a few years time XFCE or LXDE may catch up with GNOME2 in terms of features (if thats what they want to do (i am not sure it is)), or GNOME3 might be customisable to please the grumpy old men^{tm}. Until then i dont see any harm in sticking with what works. There are several distros where it easy to run MATE with a new kernel and everything else.
Actually MATE development is quite active, it has picked up some interesting features, and is gradually porting to newer libraries. it can't be to long before people get grumpy with the changes in MATE and create a new fork :-)
i suspect that quite a few people downloading debian (and linux distros in general) do not realise this. if the architectures were labelled more clearly a lot more people would probably be using the 64 bit version, and enjoying a significant performance increase.
Most benchmarks i have seen have much better speed up than 4%. eg http://www.scribd.com/doc/363677/Benchmarks-AMD64-in-32bit-mode-vs-64bit-mode-Ubuntu
Those LFS benchmarks are comparing build times. ie building a 64bit exectuable on a 64bit system is 4% faster than building a 32bit executable on a 32bit system. by that logic you should compile everything with -O0 or -O1 because thats much quicker to build.
The plausible explanation that i have heard for the big speed up is that for x86_64 you can assume that the CPU has things like SSE2, where as if you build i386 packages you might want to support older chips pentium 2 or 3, or chips like the via that lack some newer instructions. if you build everything with -march=native then you can enable the extended instruction sets on 32bit, but given the context of the discusion debian can't enable SSE2 on their i386 packages.