LinkBack URL
About LinkBacksRead more.Why we need more funding for GPGPU. It beats the pants off a quad-core CPU when done right.
Read more.Why we need more funding for GPGPU. It beats the pants off a quad-core CPU when done right.
done right is the key. there are a whole bunch of problems with gpgpu, not least of which is the low accuracy of numbers (32-bit floats only, with no guarantees how things will be rounded)
Only 10x?
The R580 can do operations 40-50x faster than a CPU provided the problem is something that a shader can solve.
its interesting because only a couple of years back we where told what an amazing difference these Physics Co Proccessors would make. Now the problem comes when you can't express the problem in a way that makes it easy to leverage these co-processors.
Its hard enough most of the time to make something that can be parralised accross two proccessors, let alone 16.
Also as hex said, rounding is odd with GPUs, take something we all should know, 0.1 + 0.2 != 0.3, but with GPUs, it could be a slightly different rounding.
Take a look at:
Programming in the Age of Concurrency: The Accelerator Project
This allows you to use the Pixel Shader, via direct-x from .Net code.
Then try and think how you could use that for the average developement task.
throw new ArgumentException (String, String, Exception)
dangel (12-02-2008)
Very interesting - I could see some applications for the tech in the work I do.. I did pick up a book on shader programming but I must admit i'd much rather have a tool that scaled the code onto the hardware automajically..
Crosshair VIII Hero (WIFI), 3900x, 32GB DDR4, Many SSDs, EVGA FTW3 3090, Ethoo 719
I'm currently in the process of optimizing OpenFOAM with CUDA, it really does require some thought and understanding of the architecture and application domain.
Some algorithms happily lend themselves to the parallel architecture of the GPU and are fairly good to execute on the GPU. Others clearly dont given the limited actual buffer space despite the huge transfer speeds capable on the ring-bus memory system. Also with the G9x series coming out in the near future (Geforce 9xxx), 64-bit precision wont be a huge issue either.
I'm finding alot of algorithms dont even need to fully occupy (WARP as its known in CUDA world) the ALU to benefit entirely from the GPU - a lot of operations in CFD applications are bandwidth restricted rather than processing power restricted. For example, benchmarking OpenFOAM on a quad core system that shares the bus (e.g. current Intel ones) show a nonlinear increase in performance at 3 cores, as the cores fight for bus access.
Here's an example of what algorithms clearly perform well on the GPU architecture to date : NVIDIA CUDA SDK Code Samples
As for 64-bit on current hardware (G82 etc), its so dam powerful that should you manage to spawn off enough threads per seconds to occupy the ALUs, it would easily overcome the extra cycles required to calculate at 64-bit precision.
As for Microsoft Accelerator, you'll want to skip that, it optimizes too generically for the architecture due to DX limitations. Interfaces such as CUDA (NVIDIA) / Close to Metal (ATI research) allow far better access to the chip, with less performance penalties per thread.
Finally, there are papers on IEEE 754 rounding on GPGPU architectures, which can help mitigate the total effect on your application, have a look within the ACM for these.
Last edited by javalord; 12-02-2008 at 02:10 PM.
Thats not necessarily true when you place it into context in a real world application. Theoretically and practically yes you can optimize a algorithm to run 40x - 50x faster, but thats only one part of the application in question.Originally Posted by kalniel
What application? If you're making the point that some applications can't be speeded up 40-50x then well, of courseBut it's the same here - GPGPU will not speed up every application by 10x either.
we're doing some gpgpu research imminently, and have a 1u nvidia tesla on order. the gut feeling is "it's great for monte carlo simulations" (i.e. codes where you want a huge number of unconnected threads doing the same thing and never talking to each other). let's see how that pans out in reality
but with all these alternative techniques (gpgpu, accelerator cards like mdgrape or clearspeed, fpgas, etc) there still needs to be sufficient payoff for the coding time, effort, cost, and running costs which all jump up where you start deviating from the norm
Would be 320 x2, or 640 stream processors in one of them.ATI's Radeon HD 3870 X2, released last month, can hit around 1TFLOPS (one trillion floating-point operations per second). Its 320 stream processors combine for massive parallel computation.
it's about time SOUND CARDS caught up with all the FLOP-age available.
I want a HUGE leap forward in accuracy, positional and spacing, and enviroments,.
It's all too fake at the moment
that or its time to take the sound card technology they have, minaturise it, bring it fully into pci express (100% ditching pci) and maybe start adding extra features into the on pcb space that most sound cards seem to have...
For example - soundcard with onboard physx card? or soundcard with onboard nic/wireless.....maybe even a mini cpu (ala killer nic cards) on soundcards that can run an antivirus or such forth on it.......
id love to see sound cards move towards multi purpose
Consider how many people actually have decent speakers to make use of "quality" sound, I personally think the sound card market pretty much reached the point of diminishing return, unless there are some kind of interface that transfer sound directly to our brain, otherwise the difference in sound quality will be quite minimal.it's about time SOUND CARDS caught up with all the FLOP-age available.
I want a HUGE leap forward in accuracy, positional and spacing, and enviroments,.
It's all too fake at the moment
Alternatively, we could have some sort of sound system that looks like a halo, with like 360 tiny speakers. That way true surround sound and could max out the volume without other people hearing it (phased array).
Or at the very least i think it is possible to make sensors that clip to your ears and the software will calculate the position of the speakers in related to you, and the sound card will calculate positional sound accurately
Back in Audigy2 age, there is(was?) a THX console where you can set the distance and angle of the speaker. Creative only put that console on the CD so I couldn't get hold of it anymore :S
Back to topic.
I'll buy whatever graphics card(s) that will do hardware-assisted x264 encoding. Even overclocked quad-core can't do real-time x264 encoding with a couple of cpu-intensive filters.
Javalord - I am interested to know what success you've had with OpenFOAM and using CUDA.
Is there a place to look at what you have done?
Hello everybody! As a proud owner of Tesla C1060 card I want to know if anybody can share some experiences of using OpenFoam with CUDA?
Anyway, I plan to hack it and if anyone is interested to join, then please let me know.
....
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote
