jarroyo wrote:
I was speaking about the PA Semi architecture being a low latancy platform, much more efficient / balanced than the X86 architecture. This chip is acually being targetted for use in places like enteprise routers, a low latency high throughput application.
PA Semi is still PowerPC, correct? While PPC is better than x86, it is still not even in the top 5 for latency
It is true what you say about CPU's being bogged down by effects processing in software, but consider these factors:
1. CPU's are standardizing on a dual core design. the power will be there to spare.
2. The 128 bit matrix math units built into the chips are made to handle these kinds of transformations.
3. Its cheaper to do it in software, and with the video, and soon to be physics off loaded from the CPU; your processor will be sitting idler than in times past; at least for games.
Just my thoughts......
3.
1) And vendors are eager to put more onto the CPU, offsetting the gains.
2) Again, vendors are eager to put more onto these units, resulting in a net loss.
3) In software, sure, but software != CPU-bound. GPU's do their work in software, but off of the CPU, for example. You can retain the softwatre flexibility without bogging down the CPU, especially now with the advent of the new Torrenza format for co-processing off of the main CPU bus, giving you the best of both worlds.