To start off, I've been crunching wu's for the past 2 years on all Boinc projects. Yes Boinc & Oracle are updated to the latest version; I'm running windows 10 pro, amd ryzen 3900xt, 12/24 core, 32g ram, 1t ssd, with 850 gigs available (free) Plus several other computers running with similar problems.

I've been crunching numbers for Rosetta for the past 2 years. Recently in the past 90 days I wanted to focus my computing power to Rosetta to increase my numbers from 280k to over 500k, its been 90 days and I'm stuck at 420k :/

-Insufficient work units
-Nonsense errors coming back from the server saying I don't have enough disk space (on all computers), when in fact I have over 800 gigs available and free for Boinc to use
-Long run times for WU's with very little credits earned in return; compared to ALL other projects
-WU's are very system taxing, especially on ram & hard drive data transfer

>I wanted to suggest an increase in credits earned, I believe and I'm sure others would agree the credit rate is undervalued. An increase of 1/3 to 1/2 is more than reasonable.

>FYI, Under project preferences, it would be nice to select which project you would prefer to run. FYI the "Rosetta 4.20" project is coming up with allot of computation errors within 10-22sec and allot of wasted time; more than a 95% error rate.

I do appreciate the hard work and effort put into this project, a little fine tuning is all it needs... Thank you!

-Nonsense errors coming back from the server saying I don't have enough disk space (on all computers), when in fact I have over 800 gigs available and free for Boinc to use

I had a talk on Github and it was determined that if you leave use no more than ___ GB empty or 0 you will get the error.
So uncheck all the other boxes and just set an amount say 90-100% of your total disk capacity in that box and the errors will go away. It's a bug in the BOINC code. To complicated to explain quickly.

-Long run times for WU's with very little credits earned in return; compared to ALL other projects
RAH is known for this, nothing new.

-WU's are very system taxing, especially on ram & hard drive data transfer
That's python for you. Someone over in Number Crunching has a cache writing program they use to slow down the writes. I have a new SSD for BOINC and I checked the life of the drive and it still has a long way to go before it will have problems from to frequent of writes.

>I wanted to suggest an increase in credits earned, I believe and I'm sure others would agree the credit rate is undervalued. An increase of 1/3 to 1/2 is more than reasonable.
Never going to happen. They talked about this some years ago and did nothing.

>FYI, Under project preferences, it would be nice to select which project you would prefer to run.

Most of us long term crunchers would like this as well, but again, won't happen.
RAH has and will be you get what you get when they want you to have it.


FYI the "Rosetta 4.20" project is coming up with allot of computation errors within 10-22sec and allot of wasted time; more than a 95% error rate.

That is also a known issue, there is a researcher that just throws stuff out on the server without testing it.
There is nothing to do but let it go through the system and take the hits. We already had this with one group of tasks. Perhaps that is what you are referring to.

BTW, no one from the team bothers to come here and read things anymore.
We are flying solo now.

FYI the "Rosetta 4.20" project is coming up with a lot of computation errors within 10-22sec and a lot of wasted time; more than a 95% error rate

A quick and obvious note.
If a task comes up with an error after 10-22secs (which it often does, you're right) it doesn't result in a lot of wasted time. It results in 10-22secs of wasted time, which is as good as no time at all.
Wasted bandwidth, yes. Wasted processing time, as good as none.
It literally takes longer to type about it than the time it wasted.

We discovered that these tasks run perfectly on Linux systems, but fail almost immediately on Windows systems.
Because the project can't tell who's going to download the task, and all would be valid tasks if run on a Linux system, it's more efficient for tasks to download and fail on Windows systems than to remove them at their end.
Sad but true.

---

I wonder if they could write a small program that checks hosts OS before sending a task? But then again that's to complicated for them.

I wonder if they could write a small program that checks hosts OS before sending a task? But then again that's to complicated for them.

BOINC does that already, so if it's known in advance which tasks will fail on Windows, they could simply split the rosetta application in "Rosetta" (all supported OS) and "Rosetta for Unix" (Linux/MacOS/Android).

---

.

I wonder if they could write a small program that checks hosts OS before sending a task? But then again that's to complicated for them.

BOINC does that already, so if it's known in advance which tasks will fail on Windows, they could simply split the rosetta application in "Rosetta" (all supported OS) and "Rosetta for Unix" (Linux/MacOS/Android).

To complicated for these people. They can't even do the simple stuff.
But cool to know.