AMD have released some cache details for Orochi (Bulldozer uarch) via a patch to Open64:
L1D is 16KB, 4-way and uses 64B cache lines, with a 18 clock miss penalty. Barcelona's 64KB, 2-way with a 3 clock access latency and 11 clocks for a miss. Any clues as to whether that's good or bad, or what it could mean for L2 (since L2's shared), since it's quite a big change from K10?
"Minden negyedik-ötödik magyar funkcionális analfabéta – derült ki a nemzetközi felmérésekből."
