Opteron sse3 !

[SolAirQuebec]

I look the result of cpu-z-137 and it give me sse3 on my opteron 285, any reason why the sse3 for p4/pd will not run?
Thanks

[Vyper]

The reason is that the IPP compiler specifically uses a command intel_fast_memcpy that doesn't exist on the AMD platform, there is a way to patch executables which are Intel specific to run on a AMD platform with somewhat a speedimprovement but that's nothing that Simon and others cannot implement because that violates the licensing deal with Intel.. So..

Thats the case.. If someone else have a better explanation here at this forum feel free to correct me but that's what i've figured out as far..

Kind Regards Vyper

[KarVi]

It's not correct that the Intel "only" versions use a special version af memcpy, thats only supported on Intel systems.

What it does is check the processor ID string for the string GenuineIntel. If anything else is reported (like AuthenticAMD), the executable fails to execute.

There's more about this here: http://www.swallowtail.org/naughty-intel.html

The patch that some peoble use to get the executable to run on AMD's, only remove the check for Intel cpu's, and the patch will run on _any_ system that supports SSE3. No commands are replaced, that concern memcpy.

Another thing to take into consideration, is that the Intel "only" SSE2 version, is faster than the SSE3 version on AMD's, and often even faster than the generic version, that is supposed to be the fastest for AMD processors.

[BenHer]

Both KarVi and Vyper are correct,

The code checks for intel processor specifically, and then if intel processor is dected it allows intel_fast_memcpy to execute. The check also allows several other intel_ named functions to execute only on intel CPUs.

Mind you, these functions are all capable of running on AMD, and have no instructions in them that would not work on AMD...and they all would run faster than the alternative function that ends up running on AMD processors.  They are named intel_<some name> only because they are part of Intel's compiler library.

[SolAirQuebec]

To resume , I can install the SSE3 for P4/PD + the patch and I will run faster...  I am right?

Thanks for your help.

[Vyper]

You should get an increase in performance yes (in theory) but how it is in practical way it could differ, it kinda seem to vary from AMD system to system..

GL..

Kind Reg. Vyper

[Hal2710]

To resume , I can install the SSE3 for P4/PD + the patch and I will run faster...  I am right?

Thanks for your help.

Hi,

* cough *


Any clues to the url of this patch ??


Hal2710


(AMD Operton170@2904mhz 24/7, with a LOT of spare cycles !!!)

[Simon]

http://www.swallowtail.org/naughty-intel.html

HTH,
Simon.

[citroja]

Could this be a sign that Intel is hurting? and that they are/were scared by AMD?

I know that they have just released their Quad core chips but they have to be worried about the new AMD multi core chips that are scheduled for Q3 and Q4 of 2007.

Anyways...at least there is a patch that can be applied to correct the issue.

-citroja

[Simon]

Hi citroja,

this situation stems from way back when, actually (P3/Athlon XP/P4). It's not anything that was done recently.

HTH,
Simon.

[SolAirQuebec]

Anyone have the final version (for opteron), I realy didn't know how to apply this patch.

Thanks everyone for the all the input (and a +++ to chicken).