As soon as the project has been released, I identified problems with MIP1 with various configurations Athlon II x4, Phenom II x6, i7-4770K, i7-6700K running Linux and Windows 7 Pro.
The "less worst" configuration is the i7-4770K machine running Windows 7 with 16GB RAM; however the MIP1 RAM consumption is over 8GB and the performance drops down on 15% percent and the RAM is not properly released after MIP1 WUs are completed, only a reboot does clean the situation.
The worst impacted configuration is the i7-7700K machine running Ubuntu Mate 16.03 with 16GB RAM: MIP1 uses about 3 to 4 GB RAM but the CPU performance (based on the measured CPU temperature) drops down significantly (48°C instead of 62°C) and the resulting performance is 50% less comparing to OET1.
The Phenom II x6 machines suffer a 35% performance decreasing, 25% for the Athlon II x4 machines, the four configuration running Ubuntu 14.04 or 16.04.
I reported this situation as soon as I did notice during Autumn 2017, but nothing happened.
Cheers,
Yves

Well, it's OK for use on those machines, as they don't use HT...but engine has some problems with HT CPUs, so those should have an issues!

Had to limit my MIP WUs to 2-3 at a time on 4c/8t machine on work!

Some points about Rosetta@Home (that uses Rosetta code).
At the beginning R@H was written in Fortran and, almost 10 years ago, they started the transition to C++, that is still in progress. Why is still in progress after all these years? For some reasons, but 2 are the most important:
- R@H (and Rosetta) can run a lot of very different and heterogenous simulations (ab initio, folding, docking, etc) so it's difficult to create a code optimized-for-all.
- Rosetta is manteined and developed by a lot of researchers from different organizations without a "centralized" and uniform way to develop the code.
During the last 2/3 years they have accelerated the adoption of C++ and create the android app, that runs only some kind of simulations and not all.

So, returning to MIP, are they running only one kind of simulation? Maybe they can consider to optimize the code (through extensions like SSEx or AVX) or create an Android app.
If Rosetta team will finish the C++ wave, maybe an OpenCl client will be possible.....dreaming

I do understand what you are saying. I 2nd that opinion about donating electricity power through computer power to science data.

To tell the truth, I've limited my HT machines (only 1, for now) to use only 1 WUs of MIP at any given time. That keeps me up crunching the MIP, but does not hit the "slow down" from HT same job on one core.

It is unfortunate to do such a thing, but it is how it is! When they change the code & update to later specs, we can release power of our CPUs more!

Hi boboviz,
that's a bad news, since such conditions would significantly limit the ability to optimize and to clean Rosetta's code.
It could also explain why we do not receive any answer from the MIP scientists since they are maybe helpless regarding the reported problems.
Cheers,
Yves