Hello Rosetta@home Team,
I'm experiencing persistent thermal issues with Rosetta@home tasks recently. Despite setting local preferences to limit usage (e.g. "Use at most 70% of CPUs and at most 37.5% of CPU time"), Rosetta@home tasks (Rosetta Beta ones specifically) consistently max out on CPU, eventually triggering thermal shutdown after a while. The behavior is specific to Rosetta@home, other projects like Einstein@home seem to respect the same limits and run without any overheating issues. This problem only started recently, the only change I made was updating my BOINC client to v8.2.4 (x64). Since my setup doesn’t have efficient cooling to run at full speed, throttling is crucial to keep things stable. As a result, I’ve had to suspend Rosetta tasks to stop my machine from crashing. Could you please look into this? Thanks in advance!

System specifications:
BOINC Version: 8.2.4
Host OS: Arch Linux x86_64
OS: GNOME 48 (Flatpak runtime) [6.15.6-arch1-1|libc 2.40]
CPU: Intel(R) Core(TM) i7-14700K (28) @ 5.60 GHz
GPU: Intel UHD Graphics 770 @ 1.60 GHz [Integrated]
Memory: 62.48 GiB

Preferences:

<global_preferences>
   <run_on_batteries>1</run_on_batteries>
   <run_if_user_active>1</run_if_user_active>
   <run_gpu_if_user_active>0</run_gpu_if_user_active>
   <idle_time_to_run>3.000000</idle_time_to_run>
   <suspend_if_no_recent_input>0.000000</suspend_if_no_recent_input>
   <suspend_cpu_usage>15.000000</suspend_cpu_usage>
   <start_hour>0.000000</start_hour>
   <end_hour>0.000000</end_hour>
   <net_start_hour>8.000000</net_start_hour>
   <net_end_hour>17.000000</net_end_hour>
   <leave_apps_in_memory>1</leave_apps_in_memory>
   <confirm_before_connecting>0</confirm_before_connecting>
   <hangup_if_dialed>0</hangup_if_dialed>
   <dont_verify_images>0</dont_verify_images>
   <work_buf_min_days>0.100000</work_buf_min_days>
   <work_buf_additional_days>3.000000</work_buf_additional_days>
   <max_ncpus_pct>50.000000</max_ncpus_pct>
   <niu_max_ncpus_pct>75.000000</niu_max_ncpus_pct>
   <niu_cpu_usage_limit>37.500000</niu_cpu_usage_limit>
   <niu_suspend_cpu_usage>0.000000</niu_suspend_cpu_usage>
   <cpu_scheduling_period_minutes>60.000000</cpu_scheduling_period_minutes>
   <disk_interval>60.000000</disk_interval>
   <disk_max_used_gb>0.000000</disk_max_used_gb>
   <disk_max_used_pct>90.000000</disk_max_used_pct>
   <disk_min_free_gb>1.000000</disk_min_free_gb>
   <vm_max_used_pct>75.000000</vm_max_used_pct>
   <ram_max_used_busy_pct>1.450000</ram_max_used_busy_pct>
   <ram_max_used_idle_pct>90.000000</ram_max_used_idle_pct>
   <max_bytes_sec_up>0.000000</max_bytes_sec_up>
   <max_bytes_sec_down>0.000000</max_bytes_sec_down>
   <cpu_usage_limit>100.000000</cpu_usage_limit>
   <daily_xfer_limit_mb>0.000000</daily_xfer_limit_mb>
   <daily_xfer_period_days>0</daily_xfer_period_days>
</global_preferences>

Hello Rosetta@home Team,
I'm experiencing persistent thermal issues with Rosetta@home tasks recently. Despite setting local preferences to limit usage (e.g. "Use at most 70% of CPUs and at most 37.5% of CPU time"), Rosetta@home tasks (Rosetta Beta ones specifically) consistently max out on CPU, eventually triggering thermal shutdown after a while. The behavior is specific to Rosetta@home, other projects like Einstein@home seem to respect the same limits and run without any overheating issues. This problem only started recently, the only change I made was updating my BOINC client to v8.2.4 (x64). Since my setup doesn’t have efficient cooling to run at full speed, throttling is crucial to keep things stable. As a result, I’ve had to suspend Rosetta tasks to stop my machine from crashing. Could you please look into this? Thanks in advance!

Preferences:

<global_preferences>
   <niu_cpu_usage_limit>37.500000</niu_cpu_usage_limit>
   <cpu_usage_limit>100.000000</cpu_usage_limit>
</global_preferences>

I'm finding this quite hard to read, but is this saying 37.5% CPU usage when computer is not In use, but 100% when computer is in use?

Can you double-check what your Options / Computing Preferences / Computing tab shows?

I hear what you say about other projects behaving themselves properly - naming Einstein - but this specific PC isn't connected to Einstein and the other projects it is currently connected to all have zero average usage
I'm hoping this is a simple oversight and an easy solution for you, because otherwise it's very strange

---

Thanks for getting back to me! Just wanted to clear up a couple of things:

I'm finding this quite hard to read, but is this saying 37.5% CPU usage when computer is not In use, but 100% when computer is in use?

Yes, that's exactly what I meant: I want to maximize usage when I'm using the computer (but only spare a little memory for BOINC WUs), and make use of unused cycle & RAM when the computer is not in use.
Here's a screenshot of my prefs for your reference:

I hear what you say about other projects behaving themselves properly - naming Einstein - but this specific PC isn't connected to Einstein and the other projects it is currently connected to all have zero average usage

I actually run other BOINC projects under a different account/CPID (not this one), so that’s why some projects might be missing or showing zero RACs here. I’d prefer not to talk much about this here.
—————————————————
Since English isn’t my native language, I used an AI tool to help refine my phrasing—but I double-checked everything before posting. Hope it didn’t cause any confusion!

Despite setting local preferences to limit usage (e.g. "Use at most 70% of CPUs and at most 37.5% of CPU time"), Rosetta@home tasks (Rosetta Beta ones specifically) consistently max out on CPU, eventually triggering thermal shutdown after a while.

The first group of Tasks might have run at full load, however the last 2 returned certainly haven't.
Almost 5 hours to do 1hr 20min of work.
I'd suggest fixing your cooling issue.

---

Grant
Darwin NT

I did a few experiments and here's what I've found:
CPU usage when running Rosetta@home only:


CPU usage when running Einstein@home:



Right now, I've set CPU affinity to force Rosetta@home processes onto what’s supposed to be E-cores (Efficiency cores on newer Intel CPUs). It’s working, things run a bit cooler, but not totally stable, and I’m still getting thermal trips once in a while.

This is not a dedicated number-crunching machine, I'm only contributing casually. But I’m going to clean the fan and check the thermal grease to see if it helps.

I hear what you say about other projects behaving themselves properly - naming Einstein - but this specific PC isn't connected to Einstein and the other projects it is currently connected to all have zero average usage

I actually run other BOINC projects under a different account/CPID (not this one), so that’s why some projects might be missing or showing zero RACs here. I’d prefer not to talk much about this here.
—————————————————
Since English isn’t my native language, I used an AI tool to help refine my phrasing—but I double-checked everything before posting. Hope it didn’t cause any confusion!

First, the AI translation is very good - that's not the problem.
Your explanation of using a different account/CPID for other projects explains why I'm seeing 0 RAC for other projects. That's fine - I'll bear it in mind that you do have other work going on

Thanks for getting back to me! Just wanted to clear up a couple of things:

I'm finding this quite hard to read, but is this saying 37.5% CPU usage when computer is not In use, but 100% when computer is in use?

Yes, that's exactly what I meant: I want to maximize usage when I'm using the computer (but only spare a little memory for BOINC WUs), and make use of unused cycle & RAM when the computer is not in use.
Here's a screenshot of my prefs for your reference:

This is a lot more helpful. and easier for me to understand and explains what's happening.
You're PC is running Boinc tasks at 100% because it's "in use", not "not in use", while your non-Boinc activity is less than 15% CPU.
That's your answer. It's doing what you've asked it to do.

Boinc only controls project tasks.
When you say you want to maximise usage when you are using the computer, I assume you mean maximise usage for whatever you're doing <outside> of Boinc.
But, by setting "at most 100% of CPU time", you are maximising Boinc's usage to 100%, not the availability of CPU time to you outside of Boinc.

Also, when you correctly check the field near the bottom that says "Leave non-GPU tasks in memory when suspended" it doesn't make any sense to "Use at most 1.45% of memory" when the computer is in use.

The computer I'm using now also has problems with cooling, but I address it in a different way. Here are the suggested changes I would make if I were in your situation.

When Computer is in use
'In use' means mouse or keyboard in last 0.00 minutes -> your computer can react In milliseconds, not minutes. Let it decide.
Uncheck Suspend when computer is In use -> no change
Check Suspend GPU when computer is in use -> no change because I don't know what other work you do. Rosetta has no GPU tasks anyway
Use at most 50% of the CPUs and at most 37.5% of CPU time -> this is what you intended in the first place and will solve all the problems you had
Check Suspend when non-Boinc usage is above 15% -> no change. This is a reasonable setting if you don't want Boinc interfering when you are using your PC for other non-Boinc things.
Use at most 50% of memory -> If you're running 14 of your 28 cores (50%) and allow 2Gb per task from your 64Gb of memory, this will be plenty for Boinc and plenty for non-Boinc, especially when you're leaving suspended tasks in memory anyway

All other settings can remain the same except the very last one.
Rosetta deadlines are 3 days, so your store of tasks must be less than 3 days and also allow for task runtime
Store at least 0.10 days and up to an additional 2.0 days of work

I hope this helps. Some changes will let you do more work. Other changes will stop your PC trying to do all the work at once and overheat

---

Despite setting local preferences to limit usage (e.g. "Use at most 70% of CPUs and at most 37.5% of CPU time"), Rosetta@home tasks (Rosetta Beta ones specifically) consistently max out on CPU, eventually triggering thermal shutdown after a while.

The first group of Tasks might have run at full load, however the last 2 returned certainly haven't.
Almost 5 hours to do 1hr 20min of work.
I'd suggest fixing your cooling issue.

I'm not certain, but I think this is caused entirely by the 3 minute delay after keyboard or mouse use is detected.
When I type or select a new window or scroll through social media I'm very stop/start. I'm pretty sure an hour of scrolling and reading would almost never meet this criteria while the PC is doing almost nothing.
I think people vastly underestimate how much of a killer this is.
This setting should always be zero or blank imo
My current settings are 100% and 68% in use and 100% and 90% not in use on my PC with weird cooling issues, yet I only have about 20 minutes of downtime over 12hr runtimes - only a 2-3% loss

---

Thanks for the suggestions! I actually did an experiment by tweaking the CPU time percentages way down to 1% (both in use & not in use), expecting to see CPU usage spike briefly and then idle, like working for 0.1 secs and resting for 9.9 secs. But it's still keeping at 100% usage! Not sure if it’s a bug or intended (I didn't pay much attention to the CPU usage stats before). For now, I’ll just wait and hope a future update address this.

This is not a dedicated number-crunching machine, I'm only contributing casually. But I’m going to clean the fan and check the thermal grease to see if it helps.

It's not about being a dedicated cruncher- it's about a system that should be able to run at 100% load for several hours without thermal throttling.
If it was a current model laptop (and not a desktop replacement unit or portable workstation), i would expect throttling as present laptops just aren't designed for extended full load operation, but a desktop system should be capable of 100% load 24/7 without throttling or stability issues.

---

Grant
Darwin NT

Thanks for the suggestions! I actually did an experiment by tweaking the CPU time percentages way down to 1% (both in use & not in use), expecting to see CPU usage spike briefly and then idle, like working for 0.1 secs and resting for 9.9 secs. But it's still keeping at 100% usage! Not sure if it’s a bug or intended (I didn't pay much attention to the CPU usage stats before). For now, I’ll just wait and hope a future update address this

Oh... OH! That's not good at all.
Even though your changes aren't exactly as I suggested, you should certainly have seen a sufficient change for things to work.

I can only make one different suggestion:
If your CPUs insist on running at 100%, perhaps you can try changing the number of cores that run.
So, try setting:
Computer in use
Use at most 25% of the CPUs and at most 100% of CPU time -> 7 cores/threads

Computer not in use
Use at most 35.7% of the CPUs and at most 100% of CPU time -> 10 cores/threads

Or perhaps 28.57% which is 1 task per performance core on your i7-14700K (8)
You can keep your memory setting at 25% for this reduced number of cores

If this works, great.
If it doesn't work, I wonder if it's better to ask the same questions in a Boinc forum to see if they can explain why Rosetta handles the settings differently from Einstein.
I'll be interested to read your results

---