What matters is not the diameter of the PSF itself, but the MTF (its Fourier's transform). The optical cut-off frequency is at 1/(lambda*F#) \approx 1010 lpmm (line pairs per mm) at 550nm (green color). The sensor cut-off is at 1/pitch \approx 833lpmm and the Nyquist frequency is at half that (416lpmm). So there is still a bit of information to suck still. Then it really depends on the shape of the PSF and how it lowers the MTF profile. These restrictions are physical and, thus, the same than for a DSLR.
I am not saying than this camera is nowhere near that of a DSLR; but your argument does not work here.
a network of deep space communications satellites colliding signals to create constructive interference to boost communications
After reading the article, I don't think they were using spaceborne emitters to build the constructive interference around the satellite location, but only ground based stations. The amount of timing precision required to trigger long-distance emitters and get this coherence would have been amazing. Doing so on the ground is still great but nowhere as difficult.
Short-wave Infrared goes through smoke and fog (cf. this). Then it is just a matter of attenuation, you just have to pump more power. What is fun is indeed snooping on the communication via the light scattered away from the beam.
Well, yes. But I don't think that we can say "terrible" performance for conditional execution. Very simply, if you have a condition "if(test){... } else {... }", the warp (group of threads) will go in the true-block if at least one of them ticks (test==true). During this portion of the execution, the threads which did not tick are disabled and are indeed waiting. And vice-versa for the false-block. If none of the threads tick, or if they all do, then the unnecessary block will be avoided (this is what we hope to have when we write code for GPUs). But at worst, you will go through both block and have half of your threads doing nothing (of course, this also depends on the balance between the amount of work between each block, here I am assuming 50/50 just to keep things simple).
Where we are severely loosing performance is when we have a condition to end a loop which is different across threads in a warp. Then some of them might spend a lot of time waiting for the last one in the warp to complete.
No they are not. The threads in a modern GPU are not all free to execute different instructions. A GPU is a SIMT architecture : Single Instruction, Multiple Threads; each warp of threads (group of approx. 16 to 32 threads) will execute the same instruction at the same time on whatever data each one is holding (some threads can also be deactivated in the group, for this instruction). So the physical architecture for each of the thread in a GPU is much simpler than for the threads of this processor (because of factorization of all the instruction queue and related mechanism, much simpler synchronization, etc.).
The main problem would be the memory bandwidth then. GPU can siphon through a lot of data because the architecture assumes that nearby threads are very likely to read contiguous data. This architecture however, allows for each core to have its own instruction queue, it should be hard to predict which thread is going to access which portion of the memory so that we can fetch it into a single request. I fail to see how you can scale the bus/controller/etc to match the bandwidth requirement (outside of few dozens MB of cache at best).
Some of the tasks you mentioned (image processing, deep learning, etc.) are already well adapted to GPUs. I doubt that this new processor will be able to beat them on this.
You have no idea how bad scientific code is : bad memory accesses, poor algorithms designs, dumb data structures, no consideration of possible bottlenecks, no cluster architecture knowledge... And wait until you see people allocating either way too much or far less resources than required for their jobs in these clusters...
Small correction here : foie gras is traditionally made from forced-fed ducks (the version with geese liver does not taste remotely the same but is cheaper). Also, foie gras is truly delicious, to the contrary of any MS platform/product.
Bold Futura, subsidiary of Tagruato Corp., has already issued this press release detailing the incident. Nothing more to see here, let's go home and drink some Slusho!
That would not be sufficient. It is not just about the libraries : powerful AMD cards are usually loud and power hungry, plus they do not have Linux/Unix support which means that they are not cluster-friendly. The message with ATI/AMD has always been clear : they only care about the gaming market and will focus it only on Windows platform.
Note that both the Titan and the Titan Z have better DP performance than the Titan X (1TFlops and 1.5 TFlops IIRC). I am hopping that they will stop crippling the DP on their "gaming" board though (or at least doing it to a lesser extent than the current 1/20~1/32x). Also, it is nice to see that the global memory bandwidth will go 4x from this generation (~250GB/s).
Slashdot Mirror
Courage
What matters is not the diameter of the PSF itself, but the MTF (its Fourier's transform). The optical cut-off frequency is at 1/(lambda*F#) \approx 1010 lpmm (line pairs per mm) at 550nm (green color). The sensor cut-off is at 1/pitch \approx 833lpmm and the Nyquist frequency is at half that (416lpmm). So there is still a bit of information to suck still. Then it really depends on the shape of the PSF and how it lowers the MTF profile. These restrictions are physical and, thus, the same than for a DSLR.
I am not saying than this camera is nowhere near that of a DSLR; but your argument does not work here.
a network of deep space communications satellites colliding signals to create constructive interference to boost communications
After reading the article, I don't think they were using spaceborne emitters to build the constructive interference around the satellite location, but only ground based stations.
The amount of timing precision required to trigger long-distance emitters and get this coherence would have been amazing. Doing so on the ground is still great but nowhere as difficult.
These rankings are only relevant if you comparing US/UK/AUS universities, not other places in the world.
Short-wave Infrared goes through smoke and fog (cf. this). Then it is just a matter of attenuation, you just have to pump more power.
What is fun is indeed snooping on the communication via the light scattered away from the beam.
... Bubble Bobble!
More like 20 euros/month, 50GB in 4G (throttling over the cap)...
I don't see how they would justify a pricey 5G.
$ lscpu
Model name: Intel(R) Core(TM) i5-4200U CPU @ 1.60GHz
No problem to render 1080p videos here, or to process high-res images with the GPU... but, I am not using Windows nor Flash.
... stop calling artificial intelligence because, most of the time, it is not intelligent, it merely reproduces what it was taught to do.
Well, yes. But I don't think that we can say "terrible" performance for conditional execution. Very simply, if you have a condition "if(test){ ... } else { ... }", the warp (group of threads) will go in the true-block if at least one of them ticks (test==true). During this portion of the execution, the threads which did not tick are disabled and are indeed waiting. And vice-versa for the false-block. If none of the threads tick, or if they all do, then the unnecessary block will be avoided (this is what we hope to have when we write code for GPUs). But at worst, you will go through both block and have half of your threads doing nothing (of course, this also depends on the balance between the amount of work between each block, here I am assuming 50/50 just to keep things simple).
Where we are severely loosing performance is when we have a condition to end a loop which is different across threads in a warp. Then some of them might spend a lot of time waiting for the last one in the warp to complete.
No they are not. The threads in a modern GPU are not all free to execute different instructions. A GPU is a SIMT architecture : Single Instruction, Multiple Threads; each warp of threads (group of approx. 16 to 32 threads) will execute the same instruction at the same time on whatever data each one is holding (some threads can also be deactivated in the group, for this instruction). So the physical architecture for each of the thread in a GPU is much simpler than for the threads of this processor (because of factorization of all the instruction queue and related mechanism, much simpler synchronization, etc.).
The main problem would be the memory bandwidth then. GPU can siphon through a lot of data because the architecture assumes that nearby threads are very likely to read contiguous data. This architecture however, allows for each core to have its own instruction queue, it should be hard to predict which thread is going to access which portion of the memory so that we can fetch it into a single request. I fail to see how you can scale the bus/controller/etc to match the bandwidth requirement (outside of few dozens MB of cache at best).
Some of the tasks you mentioned (image processing, deep learning, etc.) are already well adapted to GPUs. I doubt that this new processor will be able to beat them on this.
You have no idea how bad scientific code is : bad memory accesses, poor algorithms designs, dumb data structures, no consideration of possible bottlenecks, no cluster architecture knowledge...
And wait until you see people allocating either way too much or far less resources than required for their jobs in these clusters...
Small correction here : foie gras is traditionally made from forced-fed ducks (the version with geese liver does not taste remotely the same but is cheaper).
Also, foie gras is truly delicious, to the contrary of any MS platform/product.
I hope they don't click the red cross... or we are all fucked...
Only 9EB?
That doesn't sound like much...
It is not a "real" gigapixel camera...
(today's largest are about 250Mp)
But I don't have the time to program...
Bold Futura, subsidiary of Tagruato Corp., has already issued this press release detailing the incident.
Nothing more to see here, let's go home and drink some Slusho!
And what about a machine that takes your old, messy and large entropy and reduce to a new, tiny and shiny one?
That would not be sufficient. It is not just about the libraries : powerful AMD cards are usually loud and power hungry, plus they do not have Linux/Unix support which means that they are not cluster-friendly.
The message with ATI/AMD has always been clear : they only care about the gaming market and will focus it only on Windows platform.
is not "holographic"...
But hey, asking Microsoft to stop abusing something is a lost fight, right?
That will be difficult, as I don't mine bitcoins with them...
I need one!
Or more precisely, I need around 256 GPUs... Mark? Maaaaaaaark!
Note that both the Titan and the Titan Z have better DP performance than the Titan X (1TFlops and 1.5 TFlops IIRC). I am hopping that they will stop crippling the DP on their "gaming" board though (or at least doing it to a lesser extent than the current 1/20~1/32x).
Also, it is nice to see that the global memory bandwidth will go 4x from this generation (~250GB/s).