Hardware compatibility of CV SDK

Hardware compatibility of CV SDK

Hi,

The techncial spec said that 6th Generation Intel® Core™ processor is needed for Intel CV SDK. However, I want to know if it is also compatible on legacy platform, e.g. older chip similar to 4th Generation Intel Core processor?

Specifically, I want to evaluate CNN performance on 4th-Gen Core, e.g. running Caffe model on this kind of chip. Is such things possible?

thanks

 

4 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.

Hi, 

You may try to run it on CPU, but we don't guarantee the proper work and the same performance level as on 6th Generation, and we don't provide any support like fixing 4th Generation specific bugs. Intel(R) Processor Graphics offload via clDNN won't work. 

Best wishes, 

Anna

Hi Anna,

Thanks for the explanation. As we are evaluating some lower power processors, I have a few questions further more.

Question #1. Talk about "CPU plugin" mentioned in https://software.intel.com/en-us/dl-deployment-tool-installguide-prerequ..., as I understand it means Intel's CPU-optimized DNN inference engine depends on AVX2 instruction set, while as far as I know, Core or Xeon processors support AVX2 but Celeron/Pentium processors don't -- even if it was Skylake architecture Celeron like 3855U. 

That's to say, even if Celeron 3855U is 6th-Gen processor, the Intel inference engine cannot execute on it under CPU-mode. Am I correct?

Question #2. Talk about "GPU Plugin" mentioned in above link. As mentioned, and also you have explained, 6th-Gen Intel Processors with Skylake-architecture Intel HD Graphics can have clDNN engine working on them.

The question is, is the statement hold true on Skylake-based Celeron? For example, clDNN should be supposed to work on the 6th-Gen Celeron 3855U, right? 

Thanks

Charles

 

Hi, 

3855U is supported for both CPU and GPU acceleration.

Best wishes, 

Anna

Leave a Comment

Please sign in to add a comment. Not a member? Join today