Hi
I'm curious about the Validation Process of Rosetta@home.

Say the current WU I'm crunching, 170081942, only has one task 186214175 which is running on my laptop.

How can Rosetta@home validate the result I will return if there're no other computers doing the same work?

Cheers

Rather then have two machines doing the same work, Rosetta prefers to see twice as many models produced. The output files contain the models you have produced. They are not easily falsified.

---

Rosetta Moderator: Mod.Sense

Rather then have two machines doing the same work, Rosetta prefers to see twice as many models produced. The output files contain the models you have produced. They are not easily falsified.

Hi
I didn't mean falsifying, I mean what if there're mistakes in the results, how to identify the mistakes?

Cheers

Rather then have two machines doing the same work, Rosetta prefers to see twice as many models produced. The output files contain the models you have produced. They are not easily falsified.

what if you have a heavily Overclocked PC?

---

That's a thing i was wondering about too. I hope someone has an answer.

---

I do not know the specific answer. But I can tell you that if your machine slips during the caclulations, there are numerous places where it would cause the model to fail. If the specific data that was corrupted was more subtile then that, it is at least safe to say that the project's results are unharmed. It is compiling results from thousands of models to make a final assessment. And any that make the top of the list are closely studied, and possibly even reconfirmed on machines at U of W.

It is a holiday week in the U.S. so I wouldn't expect an immediate reply. But give this a bump in 2 weeks if you still haven't seen one.

---

Rosetta Moderator: Mod.Sense

Hi,

I would still like to get an answer, my pc is highly overclocked and i want to be sure that my pc doesnt provides the project with false calculation, all the crunching time would be senseless if it would be the case.

So can i be sure that if the results are validated and i get credits for the work units that the results are without any calculation mistakes which can appear on a overclocked computer.

---

Hi,

I would still like to get an answer, my pc is highly overclocked and i want to be sure that my pc doesnt provides the project with false calculation, all the crunching time would be senseless if it would be the case.

So can i be sure that if the results are validated and i get credits for the work units that the results are without any calculation mistakes which can appear on a overclocked computer.

Any computer can potentially make mistakes, but if it's stress-tested and stable (i.e. prime95 etc) then it should be contributing useful work. I'm fairly sure that the validation of a structure isn't a calculation-intensive process - it's determining the low energy to get to that structure that is calculation-intensive, so validation of the models produced is easily done by the bakerlab servers...

---

With my over clocked computers, I would usually get calculation errors. You will see these in the task list and your computer will abort the work unit.

When I initially attached to the project, my over clocked systems would generate several calculation errors and then start running work units correctly. I have no idea why the first WUs would fail but after an hour or so, everything would work fine. If you start to get calculation errors after hours of running error free, you likely have pushed the system a little too far.

I have a Q6600 G0 stepping running at about 3.4GHz on air. It is not the fastest Q6600 but I love the RAC on that system.

Good Luck and thank you for crunching for R@H.

Paul

---

Thx!

Paul

With my over clocked computers, I would usually get calculation errors. You will see these in the task list and your computer will abort the work unit.

When I initially attached to the project, my over clocked systems would generate several calculation errors and then start running work units correctly. I have no idea why the first WUs would fail but after an hour or so, everything would work fine. If you start to get calculation errors after hours of running error free, you likely have pushed the system a little too far.

I have a Q6600 G0 stepping running at about 3.4GHz on air. It is not the fastest Q6600 but I love the RAC on that system.

Good Luck and thank you for crunching for R@H.

Paul

Damn.
Intel did a great job with their new processors. I built a system for a cousin with a AMD Phenom X4. Even when the Q6600 isn't the best of the best from intel, it still beats it in benchmarks (Max. RAC takes time to determine). My AMD Athlon X2 6400+ is almost tied with yours in benchmarks, except that mine has two cores instead of 4.

Regarding calculation errors, the Phenom tends to throw a few calc. errors every so often for some reason. Maybe because the Quad tech isn't as perfect as dual or single core (?). Try lowering the CPU target time to avoid the errors cut your CPU time (an error could pop up at the last hour in an 8 hour CPU target time)

Felipe

---