On the Famous models that keep crashing on people, is the info in the uploads that are produced along the way still of use to the project.

When you look at units you notice that a model crashes at the same point on all systems. If we know that the info sent in is still a help we will not feel as bad about so many crashing.

---

Keep on crunching Pizza@Home

Yes, all data is useful, although a full model is best, as the last zip contains data that allows it to be stitched to one from the next 'year series'.

If you use notepad to look at a trickle from a Famous model, you'll see that there's a lot more data than those from e.g. a slab model. (The file size is also larger; 5k instead of 1k.)

---

Backups: Here

Ray,
They tend to crash at the same place on similar hosts but not necessarily so among Windows/Linux/MAC. At this stage of the project, I suspect the Hiro gets as much, if not more, from crashes as from completions. The name of the game at this point is to discover which parameters and parameter combinations fall within and without the envelope. It's a necessary step in the process.

We were asked to run even more extreme parameters on the Beta site, expecting that all would crash. (A couple of outliers made it through, to my surprise.) Hiro said he was getting valuable data/information even from the ten-fold extremes. I give the scientists the benefit of any doubts.

Hang in there -- and watch out or those wabbits.

Edit: Beat me again, Les!

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

I have 2 questions about the Famous models.

First question. There are eight 200 year segments to each model. If one of those segment crashes, do they still send out all the unrun segments in that model? Without the crashed segment how do the Scientists string together all of the segments to get the entire model.

Second question. They seem to be sending out all time periods at once. They are not sending the segments out in chronological order and waiting for the results to come back before generating and sending out the next segment in that model. That means that there can’t be any carry-over from one segment to the next. Wouldn’t they get better results if they used results the of segment 1 as a starting condition for segment 2 and so on. I realize that this would take a lot longer to do it that way.

---

These are questions that only the researchers would know the answers to.

For myself, I just crunch the data, and leave the techy stuff to them.

These are questions that only the researchers would know the answers to.

For myself, I just crunch the data, and leave the techy stuff to them.

Is there any way to ask the researchers?

---

Perhaps the first thing to do is to read through the science posts on the beta test site: here.
Some of it only applies to the early test models, but it's all good stuff. :)

In particular, this post talks about running all sub-units simultaneously.

---

Backups: Here