Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

A lot of that is a demonstration in why microbenchmarks are usually worthless for determining real world performance.

Xenos has a unified shader architecture while RSX does not. That means that during microbenchmarks, it can dedicate all of its shaders to that task, and put up relatively impressive numbers. In reality, that means that those shaders can't be used for anything else, meaning your real world performance will look nothing like the microbenchmark. AA performance is almost entirely due to the on-package EDRAM. While that was a smart design decision, it's not indicative of some huge overall power advantage.

Look, Microsoft made a number of very smart design choices with the 360 that have allowed developers to squeeze a lot out of the hardware, there's no doubting that. I think it's also fair to say that Xenos is a bit faster than the RSX -- as was the case with that generation of discrete GPUs from ATI and Nvidia -- but throwing around things like twice as powerful and "crushes" is getting into hyperbole territory if you're looking at actual performance from the hardware.



Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: