We're talking about ARM, so PCG is not applicable here. Current state is:
EKOPath ARM Cortex 50 (planning) - High performance ARM compiler targeting future ARM server devices.
Source: PathScale Labs
And why do you mention Geekbench?
Just an example: For some Intel systems there are SPEC submissions, where GCC has been used. They were submitted by AMD ;) and no further single benchmark optimizations were done, just the standards. This might cost 10-20%.
But on Intel Xeon X7350 systems (16 cores), the ICC SPECint_rate_base scores are ~54% higher (a single one even 60%).
So with just GCC available, what tricks might AMD have used to boost it's estimated result? Avoton has an additional advantage thanks to Intel's compiler of maybe 40-50% vs. GCC on ARM. The remainder of it's 33% advantage in SPECint_rate (plus a lower TDP) is due to Intel's process advantage and possibly it's architecture. For the latter comparison we'd need more data.
Market Data 
Markets 
AI
