3 ms·

This overview, while a great start, doesn't really dive into the details of how modern GPUs work. Since 2007, many of the limitations that held GPUs back from b

by ak217 12y ago

This overview, while a great start, doesn't really dive into the details of how modern GPUs work. Since 2007, many of the limitations that held GPUs back from being general-purpose computers have been removed (by relentless efforts of NVIDIA and to a lesser extent ATI/AMD, spurred in large part by NVIDIA's traction in the supercomputing space, for example http://en.wikipedia.org/wiki/Titan_%28supercomputer%29 http://en.wikipedia.org/wiki/Titan_%28supercomputer%29). My go-to source for a lot of these developments is AnandTech (http://www.anandtech.com/tag/cuda http://www.anandtech.com/tag/cuda) but I'm sure there are plenty of other resources others can point to. Another fascinating bit is that NVIDIA and ATI/AMD have developed what are now the largest general-purpose processors in the world (over 5 billion transistors per chip and counting - available in consumer GPUs for under $300, as opposed to Intel's largest Xeons that top out at 4 billion and cost $2000+) but are being held back at the 28nm process because their fab partner (TSMC) is oversubcribed by smaller, higher demand ARM chips that go into phones.

rgbrenner 12y ago
> but are being held back at the 28nm process because their fab partner (TSMC) is oversubcribed by smaller, higher demand ARM chips that go into phones. TSMC plans to begin 16nm finfet production early 2015.. although they're doing it so they can supply apple and keep up with samsung (who also supply apple and have plans for 14nm/16nm). nvidia's parker is suppose to use finfet and they're a customer of TSMC.. but parker will be a 64-bit arm cpu for servers/mobile devices.
- ChuckMcM 12y ago
  This from January of last year : http://www.dailytech.com/TSMC+Were+Far+Superior+to+Intel+and+Samsung+as+a+Partner+Fab/article34148.htm http://www.dailytech.com/TSMC+Were+Far+Superior+to+Intel+and... suggested 16nm at the end of this year. Vs Intel doing 14nm.
mej10 12y ago
What are you meaning by general-purpose here? Do you not have to use a different programming model anymore?
- dantillberg 12y ago
  The parent is referring to how CUDA's introduction enabled developers to write and compile C-ish code to run on a GPU, while previously programmers could take advantage of the power of GPUs for non-render computations by hacking the pixel shaders and such, bending the graphics hardware to do something it was not designed for. You still have to write your program in a very different way in order to run efficiently on GPUs as opposed to CPUs.
  - mej10 12y ago
    The comparison to a Xeon is what confused me. I didn't think such a development had taken place.
- m_mueller 12y ago
  I'd sum it up like this: GPGPU can be made to run any computational code - this doesn't mean that it's necessarily faster though (otherwise we could just forget about CPUs couldn't we?). A few things need to be true in order for GPUs to execute something with a speedup compared to CPU: 1) Code needs to have at least 1k, better 10k+ parallel 'threads'. 2) These threads should be largely data parallel (branching is possible but hurts performance more significantly than on CPU). 3) Registers and memory per thread are limited, around 30-60 registers and 400-800k memory are the limits to achieve a reasonable saturation. If you disregard this, spilling of memory will occur (or the memory will just run out, there's no swapping so it will just crash). 4) Because of (1) and (3), GPUs like so called 'tight loops', i.e. many parallel but smallish kernels.
  - sklogic 12y ago
    2) Some GPUs do not penalise branching 3) Some GPUs have MMUs (and share their paging with the host CPUs)
    - m_mueller 12y ago
      2) You mean Xeon Phi / Knights Corner? Yes, I wouldn't call these GPUs though, they have to be regarded a bit differently. They have quite a few problems in other areas btw., so far the results of theses systems are not very promising. 3) By MMUs do you mean unified memory? Well yes, but for now this is so slow that you don't really want to use it. This might change on Power systems with nvlink and for Knights Landing generation Intel accelerators, but that's still in the future / not publicly available.

by ak217 12y ago

rgbrenner 12y ago

> but are being held back at the 28nm process because their fab partner (TSMC) is oversubcribed by smaller, higher demand ARM chips that go into phones. TSMC plans to begin 16nm finfet production early 2015.. although they're doing it so they can supply apple and keep up with samsung (who also supply apple and have plans for 14nm/16nm). nvidia's parker is suppose to use finfet and they're a customer of TSMC.. but parker will be a 64-bit arm cpu for servers/mobile devices.

ChuckMcM 12y ago
This from January of last year : http://www.dailytech.com/TSMC+Were+Far+Superior+to+Intel+and+Samsung+as+a+Partner+Fab/article34148.htm http://www.dailytech.com/TSMC+Were+Far+Superior+to+Intel+and... suggested 16nm at the end of this year. Vs Intel doing 14nm.

mej10 12y ago

What are you meaning by general-purpose here? Do you not have to use a different programming model anymore?

dantillberg 12y ago
The parent is referring to how CUDA's introduction enabled developers to write and compile C-ish code to run on a GPU, while previously programmers could take advantage of the power of GPUs for non-render computations by hacking the pixel shaders and such, bending the graphics hardware to do something it was not designed for. You still have to write your program in a very different way in order to run efficiently on GPUs as opposed to CPUs.
- mej10 12y ago
  The comparison to a Xeon is what confused me. I didn't think such a development had taken place.
m_mueller 12y ago
I'd sum it up like this: GPGPU can be made to run any computational code - this doesn't mean that it's necessarily faster though (otherwise we could just forget about CPUs couldn't we?). A few things need to be true in order for GPUs to execute something with a speedup compared to CPU: 1) Code needs to have at least 1k, better 10k+ parallel 'threads'. 2) These threads should be largely data parallel (branching is possible but hurts performance more significantly than on CPU). 3) Registers and memory per thread are limited, around 30-60 registers and 400-800k memory are the limits to achieve a reasonable saturation. If you disregard this, spilling of memory will occur (or the memory will just run out, there's no swapping so it will just crash). 4) Because of (1) and (3), GPUs like so called 'tight loops', i.e. many parallel but smallish kernels.
- sklogic 12y ago
  2) Some GPUs do not penalise branching 3) Some GPUs have MMUs (and share their paging with the host CPUs)
  - m_mueller 12y ago
    2) You mean Xeon Phi / Knights Corner? Yes, I wouldn't call these GPUs though, they have to be regarded a bit differently. They have quite a few problems in other areas btw., so far the results of theses systems are not very promising. 3) By MMUs do you mean unified memory? Well yes, but for now this is so slow that you don't really want to use it. This might change on Power systems with nvlink and for Knights Landing generation Intel accelerators, but that's still in the future / not publicly available.