At Phoronix there's been many GCC 4.8 benchmarks already, ahead of the compiler's official release in March or April. Most of these extensive GCC 4.8 benchmarks have been from Intel/AMD x86_64 hardware and not ARM. However, being curious about the ARMv7 GCC 4.8 performance, I did some cursory benchmarks of GCC 4.7 vs. GCC 4.8 with a Samsung Exynos 5 Dual (1.7GHz dual-core Cortex-A15) SoC found in the popular Samsung/Google Series 5 Chromebook.
From the Samsung Series 5 Chromebook, which was loaded up with an Ubuntu 13.04 "Raring Ringtail" ARM snapshot, GCC 4.7.2 and GCC 4.8.0 2013-02-09 were compared for a variety of open-source, computationally-focused benchmarks. All of this ARM Linux benchmarking is handled in a fully automated and reproducible manner using the open-source multi-platform Phoronix Test Suite software.
Benchmarking results in full along with the verbose system hardware/software details can be found on OpenBenchmarking.org within the 1302257-FO-GCC48EXYN99 result file. With the results being hosted on our collaborative OpenBenchmarking.org cloud platform, it's as easy as running phoronix-test-suite benchmark 1302257-FO-GCC48EXYN99 to see how your system's performance compares to this Exynos 5 dual-core notebook setup with Ubuntu 13.04.
MAFFT, a scientific/biology test, was slightly faster with GCC 4.8.
The C-based SciMark micro-benchmarks seem to be pushing slightly higher using the GCC 4.8.0 development snapshot from early February.
GCC 4.8.0 though seems to be taking dramatically longer to build software on this ARM platform than with the GCC 4.7.2 stable release.
C-Ray multi-threaded ray-tracing is improved on the ARM A15 setup with the forthcoming GNU Compiler Collection release.
Primesieve and Smallpt are also doing better on GCC 4.8 over GCC 4.7.2.
The rest of the tests didn't yield much difference between major GCC releases on this ARM Cortex-A15 SoC. Again, the rest of the data can be found at 1302257-FO-GCC48EXYN99.