I know this has been discussed before but we have new types of WUs (ie RNA etc) now and I was wondering if there is an optimum run time that the Devs would like us to use?.....Thanks, Rog.

It's the same recommendation for any WU type. It is the hours per day that helps the project, regardless of the runtime preference that you prefer. More crunching hours per day, regardless of how many hours per task.

I generally recommend that people select a runtime that completes a task per day. So if your machine is on 10hrs a day, you might select a runtime preferences in the 8 - 12 hours range. But for anyone on dial up modem, I'd go right to 24hr runtime preference because of the time savings downloading less tasks. Assuming you can still complete 24hrs of crunching within the 10 day deadline.

---

Rosetta Moderator: Mod.Sense

It's the same recommendation for any WU type. It is the hours per day that helps the project, regardless of the runtime preference that you prefer. More crunching hours per day, regardless of how many hours per task.

I generally recommend that people select a runtime that completes a task per day. So if your machine is on 10hrs a day, you might select a runtime preferences in the 8 - 12 hours range. But for anyone on dial up modem, I'd go right to 24hr runtime preference because of the time savings downloading less tasks. Assuming you can still complete 24hrs of crunching within the 10 day deadline.

does it make any difference for the project whether I crunch (let's say) 3 wus with 4 hour runtime, or just one wu with 12 hours runtime per day?

---

does it make any difference for the project whether I crunch (let's say) 3 wus with 4 hour runtime, or just one wu with 12 hours runtime per day?

Really the scientific result is the same. And the credit would be the same as well. The main difference is bandwidth, and number of results, and what is easy for you to manage with getting new work etc. Even if you have a full-time high bandwidth internet connection, there's really no reason to download 3 times as many tasks.

So go with the runtime that suits your situation. And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

---

Rosetta Moderator: Mod.Sense

And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

Only for three machines tho :( I'd find it really useful if i had more options (ideally one per machine) without having to use local over-ride files...

---

And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

Only for three machines tho :( I'd find it really useful if i had more options (ideally one per machine) without having to use local over-ride files...

I agree completely. BOINC already has a local options screen - it can't be too hard to add more options to that existing feature.

---

does it make any difference for the project whether I crunch (let's say) 3 wus with 4 hour runtime, or just one wu with 12 hours runtime per day?

Really the scientific result is the same. And the credit would be the same as well. The main difference is bandwidth, and number of results, and what is easy for you to manage with getting new work etc. Even if you have a full-time high bandwidth internet connection, there's really no reason to download 3 times as many tasks.

So go with the runtime that suits your situation. And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

How can the scientific result be the same?

* Suppose 100 WUs for one protein are sent out, everybody crunches just a single model. You'll get a chart of 100 models.

* However if everybody crunches 100 Models, you'll get 10000 models for that proetin, that would be far more precise!

Or do the different WUs containing the same protein depend on each other, so that the following WU will start where the previous ended? If it's that way, tailoriung the WUs for the uptime of the computers is a good idea to speed up things.

Thanks for the details in advance!!

---

does it make any difference for the project whether I crunch (let's say) 3 wus with 4 hour runtime, or just one wu with 12 hours runtime per day?

Really the scientific result is the same. And the credit would be the same as well. The main difference is bandwidth, and number of results, and what is easy for you to manage with getting new work etc. Even if you have a full-time high bandwidth internet connection, there's really no reason to download 3 times as many tasks.

So go with the runtime that suits your situation. And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

* Suppose 100 WUs for one protein are sent out, everybody crunches just a single model. You'll get a chart of 100 models.

* However if everybody crunches 100 Models, you'll get 10000 models for that proetin, that would be far more precise!

If you set your WU runtime to 2 hours, your computer will generate x models for 1 protein. If you set it to 1 hours, your computer will generate in the same time x/2 models for 2 proteins, and so on.. It doesn't matter which runtime you select, the computation power is always the same.

does it make any difference for the project whether I crunch (let's say) 3 wus with 4 hour runtime, or just one wu with 12 hours runtime per day?

Really the scientific result is the same. And the credit would be the same as well. The main difference is bandwidth, and number of results, and what is easy for you to manage with getting new work etc. Even if you have a full-time high bandwidth internet connection, there's really no reason to download 3 times as many tasks.

So go with the runtime that suits your situation. And keep in mind that you can designate a different runtime for different BOINC locations too. So you can tailor each machine if you have more then one.

* Suppose 100 WUs for one protein are sent out, everybody crunches just a single model. You'll get a chart of 100 models.

* However if everybody crunches 100 Models, you'll get 10000 models for that proetin, that would be far more precise!

If you set your WU runtime to 2 hours, your computer will generate x models for 1 protein. If you set it to 1 hours, your computer will generate in the same time x/2 models for 2 proteins, and so on.. It doesn't matter which runtime you select, the computation power is always the same.

I know, it was just to simplify; 100 models vs. 10000 models in precision.

---

Tom, you are correct, running through more models will give Rosetta a much clearer picture of the protein. More accurate prediction. What you have neglected to account for is just that in order for everyone to crunch 100 models instead, it will take 100 times longer to do all the work.

What you are touching on is that Rosetta's Project Team does not know EXACTLY how many models they will get crunched for a given task. But they do know about how many they would like to see. So they put out enough tasks to get what they guess will bring back what they want, and if they reach a point where the marjority have reported back and they are coming up short of the number of models they wanted to run, then they create a few more tasks for the same protein and queue them up to be sent out. It's very easy for them to create more tasks for the same protein, and method of study.

At the same time, they are trying to optimize the use of the TFLOPs that everyone is contributing, so they are constantly reasessing which tasks to create. There are many simulaneous projects underway within Bakerlab, and so if one wants to place more work on the queue, then another must wait. This is where more overall teraflops will help the project. They will have enough CPU power to pursue more of the ideas they are devising.

---

Rosetta Moderator: Mod.Sense