What I also don't understand is this particular task is a 40 year model, yet BOINC thinks it will take longer than an 80 year model I recently finished. Shouldn't the 80 year models have roughly twice the FLOPS as the 40 year models?

I think the difference between the 40 year and 80 year models is the resolution. The 80 year HadCM3igeo models used the low resolution ocean of the older CM3 models. The HadCM3n models use a much higher resolution version of the ocean. This takes longer.

---

Also, fpops is set individually for each type of model, which can throw off the calcs that BOINC has done in the past.
The only way to 'train' BOINC, is to run several of the new type of model. Which, again, only works well for projects with very short completion times.

---

Backups: Here

For members whose duration correction factor has become inaccurate (you can check it on each computer's web page for CPDN) and is no longer close to 1, which is the ideal. The DCF can become inaccurate because of a single model whose run time is poorly estimated. But it takes ages to correct itself and return to normal. Each completed model will correct the DCF by 10%. So with our current long waits for work, long models and model defects causing crashes, we might have to wait a long time for the DCF to correct itself.

Here's how to correct it manually.

---

Cpdn news