楼主故意不读测试报告？
The SPE versus PPE performance gap is huge. Measuring the exact SPE performance shows that one SPE runs at more than 11 times the speed of the PPE and at more than twice the speed of the P4. Figure 6A shows a 1.5 times speedup for one SPE, which indicates that even with one SPE, the PPE is a major drag on performance. This is despite the fact that most of the PPE integration code is run in parallel on the SPE. The PPE overhead in this benchmark is mostly due to packing and unpacking SPE workloads, although there is some small overhead for SPE scheduling. The multi-SPE scheduling is less than optimal, as well. The multi-SPE performance continues to scale until the PPE overhead dominates at around three or four SPEs.

单颗SPE的性能在1.5x p4以上。2颗spe再多以后性能没有线性上升是因为调度瓶颈，应该是可以通过程序优化缓解的。。

引用:

原帖由 火彩 于 2006-8-13 00:26 发表


麻烦你找出一个现在能够提出真正解决多核心任务分配与调度瓶颈优化程序的天才来吧，别站着说话不腰疼。
就算是扣肉这样的双核心，要软件真的能针对这种环境优化也是很困难的事情。

天才8知道，我知道IBM有无数的科学家正在CELL相关的研究上面努力工作。。