I'm cycling my new AMD64 X2 4800 through rosetta and while collecting data I saw that it would do between 17 and 34 models in the default time setting, then I saw:

1qgtA_BOINC_ABINITIO_SAVE_ALL_OUT__1240_4418_0

which completed 254 models in the default time.

Anyone else seeing this? Credit/hour is the same as the other (roughly).

I'm cycling my new AMD64 X2 4800 through rosetta and while collecting data I saw that it would do between 17 and 34 models in the default time setting, then I saw:

1qgtA_BOINC_ABINITIO_SAVE_ALL_OUT__1240_4418_0

which completed 254 models in the default time.

Anyone else seeing this? Credit/hour is the same as the other (roughly).

I've seen one result, for ResultID #38197791, '1qgtA_BOINC_ABINITIO_SAVE_ALL_OUT__1240_4183_0', that generated 111 decoys in around two hours on an Intel Celeron M 1.40 GHz. Again, typical credit/hr, and most prior WUs at two hours on this machine are around four to 14 decoys.

The few newly-released 'PSH___________' WUs that I have seen are running above average in the number of decoys, too: for example, ResultID #38248140, 'PSH_0003_looprlx_129_138_0104_1242_7_0', at 60 decoys in around two hours for that same machine.

Remeber abinitio work units will in most cases generate more models (specially if they contain the SAVE ALL OUT instructions) they are basically your entry level work units where the models that may be submitted for further refinement are generated .

The previous norm for decoys per CPU hour (4 to 14) I have seen includes a wide population including LARS, some CASP7 refinement, and a number of Ab Initio runs. The runs for that specific protein generate a sigificantly higher number of decoys/hr--along with the newly-arrived HIV cure-related "PSH..." WUs. Thanks for your input.

Over on Ralph I am getting about 100 models/hour per core on my AMD64 X2 4800+ (not O/C'd). Some had only 57 steps for each model.

EDIT: Those are all PSH______ WU's.

FYI, answer to question about 'PSH_____' models.