...The cfft1d routine in the Intel Math Kernel library uses four threads to exploit the four available processors on the dual Opteron 280 processors and the dual 3.6 GHz HT Xeon processors. In practice, using the FFTW metric, our algorithm is able to achieve 6 GFLOPS of computational performance on a NVIDIA 7900 GPU.
Its ambitious i know but i'm going to try to start to learn to code myself. I don't have nearly the grey matter as the coders on some of these pages but maybe ill get far enough to give it a stab in the dark and maybe that'l be enought to finialy give this the starting push.
Well I wish you the best of luck Pepperammi, and hopefully you efforts might inspire some other talented coders... hint-hint-nudge-nudge... to take up this monumental task.
Best of luck.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
Its ambitious i know but i'm going to try to start to learn to code myself. I don't have nearly the grey matter as the coders on some of these pages but maybe ill get far enough to give it a stab in the dark and maybe that'l be enought to finialy give this the starting push.
Well I wish you the best of luck Pepperammi, and hopefully you efforts might inspire some other talented coders... hint-hint-nudge-nudge... to take up this monumental task.
Hm... Hints...
I think an FPGA based accelerator card would be more flexible and faster and cheaper than a GPU.
...and more funny :-)
I know this comparison isn't the best, but you can use more FPGAs altogether to improve the performance, and don't forget FPGAs are more flexible hardware elements than GPUs.
I know this comparison isn't the best, but you can use more FPGAs altogether to improve the performance, and don't forget FPGAs are more flexible hardware elements than GPUs.
That all means purchasing and installing more hardware (from all i'm finding out you'll probyl have to build and programm it yourself). That to me means its MORE expensive because i don't need to buy a new graphic card wheras i woulld have to shell out to buy an exspansion card. Most people aren't going to want to spend money for a project. They're getting enought out of me letting them have my spare computation.
As i tried to point out before. Most of us are already using cards that could be used. Also gpu's are becoming more and more flexable.
I wouldn't mind a few of those cards tho :) but i'm not going to go out and buy them.
Where would you buy them anyway?
Whist we're at it we might as well mention this
"BFG Ageia PhysX Accelerator"
Found whilst looking up those XILINX things to find out what their about.
This thing is apparently a pure physics card. Meant for gaming but isn't used in any games yet. AT £200 why would you buy one.
Anyway how do you thing something like this would run?
That all means purchasing and installing more hardware. That to me means its MORE expensive because i don't need to buy a new graphic card wheras i woulld have to shell out to buy a exspansion card. Most people aren't going to want to spend money for a project. They're getting enought out of me letting them have my spare copmutation.
Probably an FPGA card would be helpful in picture manipulation, 3D or phisics calculations, etc. I use an old S3 Virge and two TNT2 video card, so it's a real alternative to me. :)
Quote:
As i tried to point out before. Most of us are already using cards that could be used. Also gpu's are becoming more and more flexable.
Flexibility of GPUs never will reach FPGAs, but I did a GPU from FPGA that also handled a camera.
Quote:
Where would you buy them anyway?
Probably you can buy complete cards from XILINX (US) directly. ( Perhaps from Lattice, Altera or Intel? )
Whist we're at it we might as well mention this
"BFG Ageia PhysX Accelerator"
Found whilst looking up those XILINX things to find out what their about.
This thing is apparently a pure physics card. Meant for gaming but isn't used in any games yet. AT £200 why would you buy one.
Anyway how do you thing something like this would run?
Probably about as well as a GPU. IF the project's lucky and it's inner loops do one of the things the card does really fast a nice gain can be seen. Otherwise not much. The PhysX cards are focused on particle interactions, their core strength isn't directly applicable, they may or may not be useful. And at teh moment, they're strictly gimic hardware with no killer app to justify the purchase.
Whist we're at it we might as well mention this
"BFG Ageia PhysX Accelerator"
Quote:
Found whilst looking up those XILINX things to find out what their about.
This thing is apparently a pure physics card. Meant for gaming but isn't used in any games yet. AT £200 why would you buy one.
Anyway how do you thing something like this would run?
I don't know much about Ageia's card. It seems to be a bit expensive.
The location of the PPU and the memory is very similar to a normal processor ( one central memory, without CPU extension possibilities ).
Just thought of a strage
)
Just thought of a strage point.
That would it mean to the flops counting system that been introduced on SETI and is expected to cross ever to other projects?
I ask becasue here
http://gamma.cs.unc.edu/GPUFFTW/results.html <--pretty graph
RE: Its ambitious i know
)
Well I wish you the best of luck Pepperammi, and hopefully you efforts might inspire some other talented coders... hint-hint-nudge-nudge... to take up this monumental task.
Best of luck.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
RE: RE: Its ambitious i
)
Hm... Hints...
I think an FPGA based accelerator card would be more flexible and faster and cheaper than a GPU.
...and more funny :-)
... and more expensive.
)
... and more expensive.
RE: ... and more
)
I'm not sure.
Execution times of 1 million 1D FFT:
nVIDIA 7900GTX with GPUFFTW: 0,017 sec / 500 USD (card)
(32 bit precision)
XILINX XC3S1600e: 0,008 sec / 100 USD (card, FPGA ~65USD)
(16 bit precision)
XILINX XC3S1600e: ~ 0,024 sec / 100 USD
(32 bit precision)
I know this comparison isn't the best, but you can use more FPGAs altogether to improve the performance, and don't forget FPGAs are more flexible hardware elements than GPUs.
RE: I know this comparison
)
That all means purchasing and installing more hardware (from all i'm finding out you'll probyl have to build and programm it yourself). That to me means its MORE expensive because i don't need to buy a new graphic card wheras i woulld have to shell out to buy an exspansion card. Most people aren't going to want to spend money for a project. They're getting enought out of me letting them have my spare computation.
As i tried to point out before. Most of us are already using cards that could be used. Also gpu's are becoming more and more flexable.
I wouldn't mind a few of those cards tho :) but i'm not going to go out and buy them.
Where would you buy them anyway?
Whist we're at it we might as
)
Whist we're at it we might as well mention this
"BFG Ageia PhysX Accelerator"
Found whilst looking up those XILINX things to find out what their about.
This thing is apparently a pure physics card. Meant for gaming but isn't used in any games yet. AT £200 why would you buy one.
Anyway how do you thing something like this would run?
RE: That all means
)
Probably an FPGA card would be helpful in picture manipulation, 3D or phisics calculations, etc. I use an old S3 Virge and two TNT2 video card, so it's a real alternative to me. :)
Flexibility of GPUs never will reach FPGAs, but I did a GPU from FPGA that also handled a camera.
Probably you can buy complete cards from XILINX (US) directly. ( Perhaps from Lattice, Altera or Intel? )
RE: Whist we're at it we
)
Probably about as well as a GPU. IF the project's lucky and it's inner loops do one of the things the card does really fast a nice gain can be seen. Otherwise not much. The PhysX cards are focused on particle interactions, their core strength isn't directly applicable, they may or may not be useful. And at teh moment, they're strictly gimic hardware with no killer app to justify the purchase.
RE: Whist we're at it we
)
I don't know much about Ageia's card. It seems to be a bit expensive.
The location of the PPU and the memory is very similar to a normal processor ( one central memory, without CPU extension possibilities ).