- Last Active
Just to inject some more 7448 info... Make of it what you will...
Freescale are claiming that now that the 7448 can use out-of-order altivec instructions, some altivec code runs 20% faster on the 7448 than the 7447A. The 7447A Altivec core was already faster than the G5 core by a little bit so that's one in the knackers for IBM.
Improved memory subsystem - 23% improvement in altivec memcpy. Nice
twice the L2 cache - very nice. ECC - nice. slight latency increase - ok.
Dynamic Frequency Scaling much improved allows any frequency setting in just one clock cycle with no overhead - tell that to Apple with the dismal G5 power scaling in 'Automatic' mode. Previously used TEST pins on the 7447A have been used for allowing this to be controlled by hardware in the 7448 as well as by software.
Of note also is that the presentation which Freescale gave in June, mentions the Tundra Tsi108 bridge chip which now supports DDR2-400Mhz memory which they claim reduces memory power consumption by 50%. Tundra are what became of Moto's bridge chip business and the origins of the Tsi108 can be traced back to chips like Moto's MPC10x back in the CHRP days.
If Apple sticks that all together in a Powerbook then it's a nice curtain call. Even if it's just 1.7Ghz, I'd bet on it running quite a bit quicker than the 7447As at 1.67.
The Genesi/Pegasos hardware guys developing the ODW PPC workstation however were stating in May that their new boards would be coming in at 1.8Ghz, so I'd find it hard to believe that Apple would stick to 1.7Ghz.
Pentium M beater?