Nvidia's Kal-El Quad-Core ARM Chip Is Actually Slower than Intel's Core 2 Duo T7200
During the 2011 Mobile World Congress, which took place last week in Barcelona, Spain, Nvidia has shown us a benchmark of its upcoming quad-core ARM-based chip, code-named Kal-El, which placed it on top of an Intel Core 2 Duo T7200 processor. However, recent findings tend to disagree with this result and imply that Kal-El is actually slower than the Intel CPU.To showcase the power of the next-generation Tegra architecture, also known as Kal-El, Nvidia ran a synthetic ALU and FPU benchmark, dubbed CoreMark 1.0, on an Android platforms.
The results showed that Kal-El is about 10% faster than Core 2 Duo T7200 processor which features 4MB of L2 cache and runs its two x86 cores at 2GHz.
However, the small print text which was written right under the CPU scores and detailed the specific setting used by Nvidia for performing this test, showed that Kal-El was run using a significantly optimized version of CoreMark compared to that used for the Intel system.
More specifically, the Kal-El version was compiled using one of the most recent versions of GCC (4.4.1) with the O3 flag enabled while the Core 2 Duo T7200 version was complied with GCC 3.4.6 and the regular O2 flag.
Noticing these differences, the Ilsistemista website got a hold of a Core 2 Duo T7200-powered Dell laptop and ran the CoreMark 1.0 benchmark with both versions of GCC and with different architecture optimizations.
According to their findings, when using GCC 4.4.1 and more aggressive compiler flags, the performance of the Intel system apparently surged by 41%, which would place the Core 2 Duo system ahead of Nvidia's chip.
Kal-El is Tegra 2's successor in the mobile space and it features four ARM Cortex A9 cores with NEON support as well as a “12 core” GeForce GPU. (via Tech Report)