Linux kernel patch for Bulldozer cache aliasing issues benchmarked

[BlueBlazer]

Linux kernel patch for Bulldozer cache aliasing issues tested >>> AMD Bulldozer Kernel Patch at OpenBenchmarking.org

 

[blckgrffn]

So... is that just no change or am I missing something?

 

[gevorg]

The magic will come with B3 stepping.

 

[Vesku]

Guess the Linux kernel guys were right that it seems like it effects few enough possible applications that it should not be a default global patch but an optional available patch.

 

[remyat]

I'm not sure how repeatable are those benchs but to me, all I'm seeing says "margin of error"

 

[PreferLinux]

I'm not sure how repeatable are those benchs but to me, all I'm seeing says "margin of error"

Did you notice that they give the Standard Error? i.e. each benchmark was run multiple times, and the mean and SE calculated – the mean being the length of the bar, and the SE being shown below the name and also with error bars (which are too small to see). So not "margin of error".

 

[TakeNoPrisoners]

The magic will come with B3 stepping.

Magic as in 0.1 GHz higher clocks.

 

[Idontcare]

Please, for the love of pepperoni pizza, can we move on to the acceptance phase of this grief cycle now?

Surely this helps us move past the negotiation/bargaining phase.

 

[remyat]

Did you notice that they give the Standard Error? i.e. each benchmark was run multiple times, and the mean and SE calculated – the mean being the length of the bar, and the SE being shown below the name and also with error bars (which are too small to see). So not "margin of error".

I'm sorry I didn't check the link, just the image posted. The differences are minimal anyway.

I'dont know how many patches or steppings does BD need to look just a bit better. Hope AMD reduce power consumption at least, performance/watt will skyrocket.