Nvidia do architecture 64bit-Denver-soc from the cloth

0
281

Nvidia has released more details of his upcoming Denver-soc released. This 64bit-ARMv8-a soc will get the decision about a mechanism that dynamic code optimization is baptized. According to Nvidia, the Denver-soc the fastest 64bit processor on Android hardware.

The plans for a 64bit version on the basis of the Tegra K1 processor in January during the CES revealed. It now has Nvidia in a blogposting and white paper for more details about the architecture of the dualcore-Denver-soc published.

Particular to the design is that no use is made of out of order execution, as is the case with almost every modern processorachitectuur, but of in order execution. With OoO execution, instructions are not always in the order in which they arrive. In cases where two successive instructions on each other have to wait, the processor independent instructions give priority to, what the efficiency is good. Modern soc’s, among other Samsung, Qualcomm and Intel make use of OoO-exeuction.

To achieve similar results makes Nvidia use of softwareroutines that the incoming instructions and corresponding branch results to analyze and convert to optimized microcode. The optimized instructions are then in a 128MB block of ram you have placed, while a small cache on the chip, a look-up table that the processor, the optimized code quickly should be able to find.

Thanks to that technique, by Nvidia dynamic code optimization, performance similar to an out of order architecture, without that energy fix function logic is needed in the chipontwerp, according to Nvidia. An additional advantage is according to Nvidia that its software-based solution to a larger set of instructions at the same time can analyze the hardware in an OoO processor.

The design is not without disadvantages. So does the software optimize the instructions of overhead, and in addition, the Denver-core co by the addition of extra cache is a lot bigger than the Cortex A15 core in the 32bit version of the K1 soc. Where that version of the chip has five Cortex-A15 cores and knows how to accommodate, to fit on the same surface with the 64bit variation only two Denver cores. The chips are pin compatible.

Nvidia claims that the Denver-soc with a clock speed of 2.5 GHz provides performance equivalent or better than that of Apple’s A7 soc. According to Nvidia can also Celerons from Intel’s Haswell series in some of the tests are bijgebeend on ipc-plane. This would make it the fastest 64bit-ARM-soc for the Android market. The first products incorporating the 64bit version of the Tegra K1 soc is available, later this year come on the market.

Comments
(68)