AMD Bulldozer

ok, so here is a question I have about AMD Bulldozer in general. I have heard (Tekzilla) and read (I can’t remember where) that performance with Bulldozer was poor, overall. AMD came out defending Bulldozer saying that windows 7 was not yet optimized for their new design. Recently, I have heard, and read, that tests with recent builds of Win 8 would seem to hold this to be true as performance on the same hardware under Win 8 was better then under Win 7. I was just wanting to know what AMD did with this new design of theirs that would cause this?

AMD integrated the sub units of the processor more tightly in Bulldozer. 1 Floating point unit per core and 2 Integer units… trouble is windows wasn’t optimized for those changes. Where Hyperthreading Fakes 2 cores essentially to keep the sub units of 1 core busy… Bulldozer has 1 core with multiple subunits as it trys to improve single threaded performance by adding more hardware to get through more of a single thread faster when tasks pile up in the pipeline on a single thread.

I might have some of that a bit off but that was the gist of it I think.

Thanks :slight_smile: I would be curious if any of the devs here have an opinion about how AMD has done things.

http://support.amd.com/us/Processor_TechDocs/47414_15h_sw_opt_guide.pdf - “Software Optimization Guide for AMD Family 15h Processors”

http://www.xbitlabs.com/news/cpu/display/20120116163742_AMD_Describes_Piledriver_Architecture_Peculiarities_to_Software_Developers.html