Hi all! First time I've used Rosetta@home in a while in light of COVID. I had a technical question. My computer has 4 CPUs and an AMD GPU (which my understanding would suggest that the GPU isn't particularly used by Rosetta@Home).

To optimize Rosetta@Home, would it be better for me to dedicate 1 or 2 of the CPUs at 100% CPU Time or to dedicate all of my CPUs at 50% CPU Time? Is there even a difference?

Thanks!

Yes there is a difference. Using 4 CPUs 50% of the time means 4 tasks running, consuming memory, contending for L2 cache, etc.

Better to run 2 CPUs at 100% of the time.

---

Rosetta Moderator: Mod.Sense

Hi all! First time I've used Rosetta@home in a while in light of COVID. I had a technical question. My computer has 4 CPUs and an AMD GPU (which my understanding would suggest that the GPU isn't particularly used by Rosetta@Home).

To optimize Rosetta@Home, would it be better for me to dedicate 1 or 2 of the CPUs at 100% CPU Time or to dedicate all of my CPUs at 50% CPU Time? Is there even a difference?

Thanks!

In my opinion, if your CPU has HT, set it 50% number of threads when working on the PC. If you have a dedicated PC running nothing but Rosetta, then set it to 100% number of threads.

I always leave the CPU time setting at 100%. Much rather have half of my threads running @ 100% than having the whole CPU switching on and off every so many seconds. Kinda makes me think there'd be way more thermal stress that way (heating and cooling so often between 100% usage and 0% usage).

---

I agree with keeping cpu time 100% and acting on cpu numbers if you have to reduce the workload. Have the processes run at 100%, just have less of them. So if you really need to half the workload on your 4 core cpu, slap in a 50% so you will run only 2 WUs concurrently.

the GPU isn't particularly used by Rosetta@Home

Rosetta@home is a CPU only project indeed.

---

I've got my 4 cores set to run 99% of the time - does that change the dynamics?

---

I'm not happy what happens when you set the CPU to anything but 100% because the program shuts all of the threads down together and brings them back up together, which makes the temperature of the whole CPU package continually bounce which also confuses my fan control program.

It has to be a worst case for the longevity of your motherboard and cpu to have the machine continually cycle from 0 to 100% just slow enough for the cpu package temp to bounce for months on end.

I'm not happy what happens when you set the CPU to anything but 100% because the program shuts all of the threads down together and brings them back up together, which makes the temperature of the whole CPU package continually bounce which also confuses my fan control program.

It has to be a worst case for the longevity of your motherboard and cpu to have the machine continually cycle from 0 to 100% just slow enough for the cpu package temp to bounce for months on end.

My understanding is that the switching onoff of the CPUs is it does it multiple times per second. I'd be very surprised if the fan control programme responded on the same scale. Maybe it does, but I'd be surprised.

And speaking as someone who's just had a motherboard failure on my main desktop while running at 100% everything 247 for a few years, the demands on the system are pretty excessive at the best of times. But at least I know I've got full value out of what's failed <sigh>

---

My understanding is that the switching onoff of the CPUs is it does it multiple times per second. I'd be very surprised if the fan control programme responded on the same scale. Maybe it does, but I'd be surprised.

And speaking as someone who's just had a motherboard failure on my main desktop while running at 100% everything 247 for a few years, the demands on the system are pretty excessive at the best of times. But at least I know I've got full value out of what's failed <sigh>

On all of the systems I’ve run it drops all of the cores for a second or two every 15 to 60 seconds depending on the percentage used so it isn’t as subtle as you expect.

My understanding is that the switching onoff of the CPUs is it does it multiple times per second. I'd be very surprised if the fan control programme responded on the same scale. Maybe it does, but I'd be surprised.

And speaking as someone who's just had a motherboard failure on my main desktop while running at 100% everything 247 for a few years, the demands on the system are pretty excessive at the best of times. But at least I know I've got full value out of what's failed <sigh>

On all of the systems I’ve run it drops all of the cores for a second or two every 15 to 60 seconds depending on the percentage used so it isn’t as subtle as you expect.

Perhaps I'm getting distracted by one of those sidebar apps I used to run under Vista (a fair while ago) that showed CPU utilisation going 0-100-0-100 a few hundred time per minute. I don't know for sure it was accurate

I'm back running btw, with only a few hours to go before my deadlines. More failures coming...

---

I'm back running btw, with only a few hours to go before my deadlines. More failures coming...

20 tasks aborted by the server for not starting before their deadline. Oh the irony...
At least some more came down

---

But why lol. Just keep them at 100%.

---

I am running at 100% time and 100% to 75% cores with HT enabled, while I underclock the CPU to reduce temperature and noise.

According to my measurement, I always apply the "partial use of all cores" philosophy. It can be easily tested a thermally-constrained system.

If I set it to only use 1-2 cores, it will dial up the fans to 100% RPM and throttle like mad, but if I run over all cores and modulate the computing cores idle time in sync by hand in 50% of the time, the fans are spinning at a quiet setting (BOINC's CPU usage modulation doesn't seem to be that efficient).

This is related to turbo boost, power increasing with the square of voltage, allowing HT (almost 30-50% gain for negligible additional power), and package power saving optimizations.