Is there any benchmark programs for CUDA ?
and how much is actual difference in GFLOP-s between 8800GTX and 8800GT ?
The CUDA SDK contains some samples that you could use as benchmarks, but in our experience performance depends very much on your application.
The GeForce 8800 GT has fewer stream processors than the 8800 GTX (112 vs. 128), but a higher clock speed (1.5GHz vs. 1.35). It has a 256-bit memory interface (compared to 384-bit on the GTX), which can be a factor in bandwidth-limited applications.
The GeForce 8800 GT also supports compute capability 1.1 (atomic operations), which can be a big benefit in some algorithms.
In terms of price-performance, it’s hard to beat!
I’ve got one more question.
Can CUDA work in SLI mode ? and if so it means that dual 8800GT can beat 8800GTX approximately for the same price.
No CUDA does not work in SLI mode. You can use multiple GPUs, but you have to manage them.
OK, thanks for help.
I’ll buy 8800GT for testing and change it with ULTRA models when double precision will be available.
Please note that there is not yet a released display driver for Linux which supports the 8800GT.
also note (and please correct me if im wrong) this card isn’t supported in the sdk yet, only inver 1.1 it will be (which they say is do out very soon).
It’s the other way around - current CUDA driver (CUDA 1.0) doesn’t support 8800GT.
Paulius
oh ok thanks stand corrected :)
You can use multiple GPUs, but you have to manage them.
What could this possibly mean by managing multiple GPUs? So can we scale the performance by using different devices(8800GTX + 8800GTS) or the same GPUs?
Please search the forum for “sli” or “multi gpu”. There have been several threads discussing this.
Paulius
http://forums.nvidia.com/index.php?showtop…58&hl=multi-gpu
http://forums.nvidia.com/index.php?showtop…22&hl=multi-gpu
Question:
Well I guess I get an idea of it. However, could you explain the differences between SLi circuity and the Multi-GPU circuitry, coz we all know that SLi model doesn’t ultilize the 2nd GPU’s video memory at all.
There isn’t any special multi-GPU circuity. You start a CPU thread for each GPU in your computer, and each CPU thread manages one graphics card. You can run whatever kernels you want on each card independently.
Actually, SLI does use the video memory of both GPUs. SLI has two modes: Split-Frame Rendering (SFR) and Alternate-Frame Rendering (AFR). In the former, each GPU renders a portion of the graphics frame. In the latter, GPUs alternate rendering each frame. In both modes, only one GPU is responsible for pushing the frames out to the display (since a single monitor can’t be connected to both GPUs at once). The SLI connector is simply a “pixel bridge” which the secondary GPU uses to transfer either part of an image (in SFR) or a whole image (AFR) to the primary GPU to be displayed.
Since both GPUs are rendering the scene, both GPUs need all of the scene’s geometry and textures in their memories. Therefore the memory of all GPUs is used in SLI.
Now on to multi-GPU in CUDA. This is not SLI. Because the graphics API constrains the type of computation being done on the GPU, SLI can make assumptions about how to parallelize the application across multiple GPUs. In GPU computing with CUDA, it’s difficult to come up with assumptions like this that apply to all applications that a programmer might write. Therefore, to run on multiple GPUs, the designers of CUDA decided it would be best to enable CUDA programmers to manage the GPUs in the system themselves.
You can do this by creating as many host threads as you have GPUs, and running a separate CUDA context in each thread. The functions cudaGetDeviceCount(), cudaSetDevice(), cudaGetDevice(), cudaGetDeviceProperties(), and cudaChooseDevice() facilitate this. Please see the programming guide and the “multigpu” sample in the CUDA SDK for more details. There are more multi-GPU samples coming in upcoming SDK releases.
Mark
Hi,
I read on a Linux forum that NVIDIA has a beta version of a display driver for
Linux which supports the 8800GT. If so, can you give me the link to download
the driver?
Thank you.
The CUDA_1.1 release already supports the 8800GT.