I plan to add per GPU info:
* temp
* fan
* broken shares

broken are the shares with bad compute. Info about network-related rejects are irrelevant for a GPU, it's a internet/ping/pool problem.
And the yellow report would include the overral shares found/broken/pool-reject.

Config:
* worksize is the exact same
* intensity in JCE is the multihash parameter in the config, whose value is the same as in Claymore or Stak
* alpha and beta are JCE specific parameters

I strongly advise you fine-tune from scratch, a good SRB config is not always good in JCE, our code is completely separate.
Use the 0.32a, it starts very fast you can tune easily. And you can disable CPU with param --no-cpu

test version : ok thanks a lot, i start to understand what part is broken (even if it doesn't explain why it is broken) and on Ellesmere only.