Showing posts with label NVIDIA. Show all posts
Showing posts with label NVIDIA. Show all posts

Wednesday, April 6, 2011

AMD Bulldozer versus Intel Sandy Bridge versus NVIDIA CUDA

AMD Bulldozer versus Intel Sandy Bridge versus NVIDIA CUDA
By William Jiang, MLS

The new battleground for microprocessors is now the realm of graphics, and there are three major combatants: AMD, NVIDIA, and Intel. Intel produced the Sandy Bridge series of processors which integrate the Computer Processing Unit (CPU) and Graphical Processing Units (GPUs). This integration of CPU and GPU seems to be a trend for the future. AMD’s flavor of this hybrid CPU and GPU is called an APU or Accelerated Processing Unit. AMD’s new processor architecture named Bulldozer has come out with their first chip aimed at the mobile segment: Llano.

The “system on a chip” was not always in existence. Since the Intel 8066 chip came out in 1978 video cards handled video, and the CPU was divided into the CU (Control Unit) and ALU (Arithmetic Logic Unit) and never did the two meet. NVIDIA emerged as the video card power more than a decade ago, and relatively recently NVIDIA started the CUDA technologies which gamers and people involved in multimedia use for drastic improvements in performance. As of yet, a strong CPU with a strong NVIDIA video card combination’s performance can not be touched by an integrated CPU/GPU as of yet, so NVIDIA has some breathing room. However, as the CPU/GPU integration becomes tighter, NVIDIA may need to find a strategic partner in AMD or Intel.

Wednesday, December 29, 2010

NVIDIA and GPUs VS Intel Sandy Bridge By William Jiang, MLS

In the past, all the work of computing depended on the CPU. The The more powerful the CPU, the faster programs ran. However, besides the CPU there is now the GPU (Graphical Processing Unit on the video card) revolution that NVIDIA current leads with a technology called CUDA. Desktop computers can run at speeds that rival supercomputers of 5 years ago!


the new generation of NVIDIA Echelon GPU’s run at about 10 teraflops (http://gpuscience.com/news/sc10-nvidia-details-10-teraflops-echelon-processor/)

most supercomputers still running the teraflop range. So, this is an example of technology advancing significantly until what used to be a supercomputer not too long ago that cost millions of dollars is now available on the desktop for a few thousand. the graphic below shows how powerful processors can increase their power of CPUs exponentially by using programs that use the GPU like the Mercury playback engine in Adobe Premiere CS5. Below we see that the rendering time of a video decreases by almost 10x. (http://www.tomshardware.com/reviews/xeon-x5680-5600-series-westmere-ep,2692-7.html)





Of course the giant Intel does not want to NVIDIA to have this much power in the computer market. So what Intel is doing with the Sandy Bridge processor is they are putting a middling powerful GPU right on the chip. NVIDIA GPUs still has more power, but with Intel’s economy of scale and most people going with Intel, the behemoth may make inroads into the GPU market and the crush NVIDIA. This senario may be how Microsoft crushed Netscape. With up to eight cores with an integrated GPU Intel is looking powerful.
read more about the Sandy Bridge architecture here: http://en.wikipedia.org/wiki/Sandy_Bridge_(microarchitecture)