IPP AVX vs SSE performance

[Oleg_F_]

Hi,

Does anyone knows how SSE absence in the CPU affects the performance of IPP library

I need to know what's the difference in performance (if any) can be between Intel processor with just AVX support versus Intel processor with both SSE and AVX support.

Can anyone help me with this question?

Thanks,

Oleg Fomenko

[Igor_A_Intel]

Hi Oleg,

do you have such CPU that supports AVX and DOESN'T support SSE? I don't know such one as all Intel CPUs are compatible from old to new - so all code developed for old CPUs will work on the new ones.

regards, Igor

[Oleg_F_]

Hi Igor,

Thanks for your quick reply

According to intel site, the following processors has only AVX extension (see instruction set extensions):

http://ark.intel.com/products/64590/Intel-Xeon-Processor-E5-2650-20M-Cache-2_00-GHz-8_00-GTs-Intel-QPI

Whereas this one has both sse and avx:

http://ark.intel.com/products/65719/Intel-Core-i7-3770-Processor-8M-Cache-up-to-3_90-GHz

Thanks,

Oleg

[Igor_A_Intel]

Ok, I see that information at ark site may be confusing... The first one is based on SandyBridge core and of course supports all "old" extensions - from FPU/MMX to SSE4.1/4.2 and AVX; the same is true for the second one - it is IvyBridge core that just a 'tick" to SandyBridge and has all the same extensions (the main difference to the 1st one - significantly improved memory subsystem).

regards, Igor