A group of researchers of the University of Princeton just of surprise to the community with the creation of a chip Open Source based in the architecture SPARC that has with nothing less that 25 nuclei. So far everything more or less common although this project truly stands out for one more interesting feature since it is scalable, something which would use it in the future to a computer with 200,000 cores.
This chip has been baptized by its creators under the name of Python and is designed with the sole purpose of being particularly flexible and easily scalable to that, once is a commercial product, from the first moment can be used many of these CPUs together in servers and data centers that require a high degree of parallel data processing.
Researchers from the University of Princeton us surprise with an interesting idea of CPU flexible and scalable.
This functionality is given exactly by the use of the SPARC architecture since, unlike those used currently by different companies like ARM that may be much more projection, for this project, its creators decided to be based on OpenSparcdesign, which in turn is nothing more than a modified version of the OpenSparc T1 Oracle processor.
Entering a little more in detail, we find with that them developers of this chip as peculiar have used a topology of matrix of 5 x 5 nuclei, each one of which operates to 1 GHz. A project that is especially interesting for the scalability that appears when need to share memory DRAM or directly the storage available.
For the communication between cores is bet by the integration of a small router in each one, something that to your time makes of this design one of them more interesting in the field Open Source. We are talking about a CPU equipped with 460 million transistors, a figure that is still far from the 8,000 million which have such processors from Intel or NVIDIA, which among other things is due to the manufacturing processes used since, while major manufacturers work in 14 or 16 nm researchers at Princeton University have used a 32 nm scale.
More information: pcworld
