Nvidia is continuing to tease all its fans on Twitter by releasing interesting pictures of its new and much anticipated GF100 Fermi-based video cards. online pharmacy no prescription needed Not a log ago there was a picture of a single GF100 card running an Unigine game engine benchmark and now we have a new picture showing two GF100 cards running in SLI mode. On the image above you can see Tom Petersen, Director of Technical Marketing at NVIDIA (an Apple fan obviously), proudly showing off his new Maingear Shift PC, equipped with two GF100 cards. So when will all this waiting end and we are going to see the new GeForce video cards on the market and be able to buy them… it is still a mystery.
There are a lot of Nvidia-based video cards out there with support for the CUDA technology, over 150 million worldwide according to the company, but these cards do have very different parameters. Starting from GeForce 8×00 series Nvidia has added support for the CUDA technology, even for the low end GPUs and these, although not offering that much processing power, are still able to take advantage of the GPU acceleration that can be achieved thanks to this technology. But the problem you might face using CUDA-enabled software is not the speed you’ll get – this of course depends on the GPU power, but is related to the amount of video memory your video card has. If you are going to put your CUDA capable Nvidia-based GPU with a software that can take advantage of the technology in order to accelerate the processing speed of something you’ll also need more memory. For instance if you are going to work with Full HD video or other high-resolution audio video material you’ll probably need more video memory available in order for the video card to process all that data.
One recent example cheap Acomplia of bumping into a problem with not enough video memory available to process certain things with the help of CUDA was when I was testing the muvee Reveal 8 software and wanted to use CUDA on HD quality video, but on a lower-end GPU with a limited amount of video memory. Here is what I was able to achieve as output quality in terms of resolution depending on the amount of video memory available for the GPU using CUDA acceleration for the video encoding process:
- 512MB VRAM – 720p and 1080p resolution
- 256MB VRAM – maximum 960×540 resolution
- 128MB VRAM – maximum 640×360 resolution
- 64MB VRAM – CUDA option was not available
And these was on the same not very powerful GPU that uses shared video memory – Nvidia ION (GeForce 9400M class) with just 16 stream processors and with the help of a single core Intel Atom processor running at 1,6 GHz with HyperThreading enabled. So the capabilities of the GPU in terms of the output resolution were not limited by the number of the processing elements inside the GPU, but on the amount of video memory it has available…
So just as an useful advice for you, if you are planning to use your Nvidia video card for CUDA acceleration and want to take full advantage of the features available to you by the software you better go for at least 512MB of video memory, even if the GPU is not a high-end one!
After reporting about the problems with CUDA I’ve had with the recently released muvee Reveal 8 software, the company has responded very quickly with a release of a new build of the program that fixed the issue. So now you can download the 15 day Trial version of muvee Reveal 8 and put your PC to the test, using this quick and easy video guide for the test procedure to follow. Of course to benchmark your CUDA performance you’ll have to own an NVIDIA-based video card with CUDA support, otherwise you can just compare your CPU performance…
I’ve already got some test results prepared from different computer systems, but if you decide to try this benchmark on your system please don’t forget to share the results you got here in the comments, so that the database of results can be increased. Please test in both 720p and 1080p resolutions in AVC-HD format (H.264 compression) with and without CUDA enabled and post the CPU and Video Card you have in your computer along with the four results like in the examples below.
