Reverse Hyperthreading
Reverse Hyperthreading
Have anyone considered this concept? Rather than dividing a CPU core in half, what if two cores worked together for double the performance? Currently, cores process one instruction per program, but imagine if Intel, AMD, or Qualcomm created firmware to combine two physical cores into a RAID 0 setup. This would let games and programs that normally used just one core use two. An 8-core processor could be split in half instantly, though most people don’t really use more than four cores anyway. (Besides, there’d need to be a fast controller to manage the instructions.)
AMD attempted to revive Bulldozer, but it ultimately ended in failure.
It wasn't exactly what you expected, but it was mostly fine. We had 8 X86 cores with 4 FPUs each, though 2 cores shared one FPU. This setup didn’t use SMT or HT—just one X86 core handling multiple threads. You don’t need a dedicated FPU; processors at the time only had one, without an FPU built-in.
It seems like this discussion revolves around AMD's FX processor and its performance characteristics. The conversation touches on clock speeds, transistor counts, and the impact of design choices on speed. There’s a focus on how certain components, like the 1090T chip, compare to other processors, and how AMD’s approach to chip architecture influenced market perception. The tone reflects both technical analysis and personal experiences with overclocking and hardware performance.