Informally, I'm curious how everyone here crunches for S@H/BOINC. In particular:

1. What % of CPUs do you run, and what % of CPU time do they run (i.e., the first settings in computing preferences).
2. Do you overclock your CPU, underclock it, or keep it at standard/boost speed?
3. Do you overclock your RAM?
4. Since we're talking over/underclocking, I might as well ask if you do the same for your GPUs, although to be honest I'm not as curious about that at the moment.

The reason I'm asking: I get the feeling that a lot of users here dial back the % of CPU time in their computing preferences. The reason for this (at least, my understanding, please correct me) is that it provides an optimal amount of crunching compared to the power consumed by the computer. At a certain point there is not much gained in what is crunched vs. power consumed.

However, I'm not sure I have a good understanding of how many people overclock or underclock their CPUs/GPUs. If you overclock your computer, but yet you also keep the number of CPUs and/or the run % below 100%, it seems counter intuitive to me. I would think you would want to push the software as hard as it can go first, and then the hardware. Perhaps this analogy isn't perfect, but I wouldn't you want to use all the lanes on the highway before increasing the speed limit?

As for myself, I run BOINC 100% of CPU time on my computers, and they are not overclocked. On my Ryzen, I actually have the boost off, so I'm only moving at 3.5 GHz. The main reason I have the system at the standard speed is to keep the noise level quiet. I also run my RAM at 3200 (its design speed), but I am looking to overclock it.

First and foremost I am curious how everyone has their computers set. Secondly I'm curious why you have it set that way.

---

Seti@home classic: 1,456 results, 1.613 years CPU time

I run 100%/100% of cpu on all my hosts. I then limit the number of cpu tasks running with a max_concurrent statement. I overclock all my cpus and gpus. I overclock my memory. On Ryzen, you get the most improvement from overclocking the memory the most you can get out of it because memory clocks are directly tied to cpu clocks. The faster the memory, the faster the cpu. I don't worry about noise or power consumption.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

+1

---

The Ryzen iGPU series (2200G/2400G) have a different gpu to cpu architecture than the cpu -> discrete gpu.

This has been an issue in everything the reviewers and various Seti people have tested using a variety of different AMD APU's etc.

1) Because the memory bus is shared between the cpu and the gpu, if the cpu is turned up all the way, the gpu will not run at full tilt.
2) The faster the ram is, the faster the gpu is.
3) The commonest way to allow the gpu to run at full speed is to disable 1 or more cpu cores.

Since you have a 4 core, no SMT/Hyper threading cpu, I predict if you run at 75% of the available cpu cores and turn up the gpu frequency in the motherboard bios, your gpu processing time will drop.

I also think the gpu app commandlines can help some.

For what it is worth. My currently down Amd 2400G is reporting long wall clock times with something like 2 minutes of cpu time for the iGPU tasks.
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8671627

Tom

---

A proud member of the OFA (Old Farts Association).

1) Because the memory bus is shared between the cpu and the gpu, if the cpu is turned up all the way, the gpu will not run at full tilt.

That is a good point about the memory being shared. I might experiment with that to see what happens.

---

Seti@home classic: 1,456 results, 1.613 years CPU time

1 - 100%
2 - 12 hours per day
3 - No
4 - Whatever they came out of the factory with

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

I run all cores & all threads, no manual overlocking of anything. Water cooled CPU, lots of fans in the case & just let the CPU & GPUs own boost functions do their thing.
I'm not concerned so much about power consumption as such, it's just that spending several hours (days, weeks?) to get a stable system that uses 15% more power for 5% more output just doesn't appeal to me.

---

Grant
Darwin NT

+1

And produce errors. https://setiathome.berkeley.edu/workunit.php?wuid=3479669561

+1

And produce errors. https://setiathome.berkeley.edu/workunit.php?wuid=3479669561

inconclusive aren't errors..........

State: All (5314) · In progress (200) · Validation pending (2463) · Validation inconclusive (54) · Valid (2597) · Invalid (0) · Error (0)

---

+1

And produce errors. https://setiathome.berkeley.edu/workunit.php?wuid=3479669561

inconclusive aren't errors..........

State: All (5314) · In progress (200) · Validation pending (2463) · Validation inconclusive (54) · Valid (2597) · Invalid (0) · Error (0)

Semantics.
Tasks that are Invalid, or probably will be declared Invalid, are Errors in my book.

The tasks that are errors like the one I have, probably isn't, it's probably a ghost of some sort, it never reached my computer and was timed out after 5 minutes.

Tasks that are Invalid, or probably will be declared Invalid, are Errors in my book.

Most Inconclusives will end up validating, it's only if you get stuck with a dodgy host they end up as an Invalid. Even so, the less Inconclusives there are, the better.

---

Grant
Darwin NT

Tasks that are Invalid, or probably will be declared Invalid, are Errors in my book.

Most Inconclusives will end up validating, it's only if you get stuck with a dodgy host they end up as an Invalid. Even so, the less Inconclusives there are, the better.

Almost certainly on my part, I cannot say the same about wingmen who overclock.

Tasks that are Invalid, or probably will be declared Invalid, are Errors in my book.

Most Inconclusives will end up validating, it's only if you get stuck with a dodgy host they end up as an Invalid. Even so, the less Inconclusives there are, the better.

Almost certainly on my part, I cannot say the same about wingmen who overclock.

Given that your inconclusive rate is 2.02% compared to my 2.07%, you should look at your own computer following your own advice.

---

Tasks that are Invalid, or probably will be declared Invalid, are Errors in my book.

Most Inconclusives will end up validating, it's only if you get stuck with a dodgy host they end up as an Invalid. Even so, the less Inconclusives there are, the better.

Almost certainly on my part, I cannot say the same about wingmen who overclock.

Given that your inconclusive rate is 2.02% compared to my 2.07%, you should look at your own computer following your own advice.

I can say quite honestly that my results will almost certainly be validated. Of the 6 I have at the moment 4 are computers using intel gpu on Darwin, one is an ATI graphics task, these are all known to be fragile and frequently produce odd results.

Yours is the sixth.

1) Because the memory bus is shared between the cpu and the gpu, if the cpu is turned up all the way, the gpu will not run at full tilt.

That is a good point about the memory being shared. I might experiment with that to see what happens.

Since yesterday morning, I have set the 2200G to <ncpus>3</ncpus>. Since then, performance of the GPU has increased significantly! I was getting average times of 4,200 - 5,300 seconds (70 - 88 minutes), now I am in the range of 1,900 to 2,200 seconds (32 - 37 minutes)! I have not added a command line yet, nor have I touched the RAM OC.

Although that is an incredible amount of time saved, it currently appears that my RAC is decreasing slowly. If the trend continues, I would think I would be better off running four tasks concurrently (3 CPU + 1 GPU) than three tasks (2 CPU + 1 GPU). I'll monitor this for a few more days and see what happens.

---

Seti@home classic: 1,456 results, 1.613 years CPU time

1) Because the memory bus is shared between the cpu and the gpu, if the cpu is turned up all the way, the gpu will not run at full tilt.

That is a good point about the memory being shared. I might experiment with that to see what happens.

Since yesterday morning, I have set the 2200G to <ncpus>3</ncpus>. Since then, performance of the GPU has increased significantly! I was getting average times of 4,200 - 5,300 seconds (70 - 88 minutes), now I am in the range of 1,900 to 2,200 seconds (32 - 37 minutes)! I have not added a command line yet, nor have I touched the RAM OC.

Although that is an incredible amount of time saved, it currently appears that my RAC is decreasing slowly. If the trend continues, I would think I would be better off running four tasks concurrently (3 CPU + 1 GPU) than three tasks (2 CPU + 1 GPU). I'll monitor this for a few more days and see what happens.

RAC has been dropping for everyone since Arecibo Tasks ceased awhile back... My RAC has dropped about 1K this past week.

Give things a chance to stabilize. You will be better off running 2 CPU and 1 GPU in the long run. I run CPU in BOINC Manager at 50% CPU and 50% CPU Time. On my 4c/8t i7 7700K this reserves 4 Threads for System and BOINC GPU Usage. I also, currently, choose NOT to Crunch CPU Tasks due to electricity costs being HIGH in California. Hackintosh-Andromeda, (i7 7700K System), on iMac 18,3 Profile is currently at 42.5K RAC on its own ONLY Crunching on a MacVidCards' GTX-1070 8GB Card and a secondary EVGA GTX-1050 2GB Card. My dad's i7 7700 System Crunching on the Integrated HD-630 GPU brings my Total RAC into the 43K Range.

Let things be for awhile and see what happens... Give it about a month to stabilize and see what happens with your RAC then.

ymmv


TL

---

TimeLord04
Have TARDIS, will travel...
Come along K-9!
Join Calm Chaos

Since yesterday morning, I have set the 2200G to <ncpus>3</ncpus>. Since then, performance of the GPU has increased significantly!

I've always just made use of app_config.xml to reserve a CPU core for GPU support. If the GPU runs out of work, the core then starts processing CPU WUs again. When more GPU work comes along, it just goes back to supporting the GPU. Add or remove a GPU, it automatically adds or removes a supporting CPU core.

<app_config>
 <app>
  <name>setiathome_v8</name>
  <gpu_versions>
  <gpu_usage>1.0</gpu_usage>
  <cpu_usage>1.0</cpu_usage>
  </gpu_versions>
 </app>
 <app>
  <name>astropulse_v7</name>
  <gpu_versions>
  <gpu_usage>0.5</gpu_usage>
  <cpu_usage>1.0</cpu_usage>
  </gpu_versions>
 </app>
</app_config>

---

Grant
Darwin NT

Since yesterday morning, I have set the 2200G to <ncpus>3</ncpus>. Since then, performance of the GPU has increased significantly! I was getting average times of 4,200 - 5,300 seconds (70 - 88 minutes), now I am in the range of 1,900 to 2,200 seconds (32 - 37 minutes)! I have not added a command line yet, nor have I touched the RAM OC.

Although that is an incredible amount of time saved, it currently appears that my RAC is decreasing slowly. If the trend continues, I would think I would be better off running four tasks concurrently (3 CPU + 1 GPU) than three tasks (2 CPU + 1 GPU). I'll monitor this for a few more days and see what happens.

It looks like you have dropped a core in the bios. I was suggesting you run 4 cores but use the local option of the Boinc Manager to drop it to 75% of the cpu. This provides that "extra" cpu core whenever it is needed without you having to re-boot the system.

I wonder if your relative position on that machine on the Leaderboard will change. Since everyone's RAC is going down, your position shouldn't vary, heck it could go up :)
https://setiathome.berkeley.edu/top_hosts.php?sort_by=expavg_credit&offset=6080

Tom

---

A proud member of the OFA (Old Farts Association).

Looks like I have to catch up this morning!

RAC has been dropping for everyone since Arecibo Tasks ceased awhile back... My RAC has dropped about 1K this past week.

Yeah, I knew that could be affecting things. I haven't been paying close enough attention so I didn't remember if we still had Arecibo, didn't, or if we were sideways ;) I also switched a laptop to Ubuntu 18.04 and I'm curious how the RAC has changed for that, but this isn't helping things. Oh well.

I've always just made use of app_config.xml to reserve a CPU core for GPU support. If the GPU runs out of work, the core then starts processing CPU WUs again. When more GPU work comes along, it just goes back to supporting the GPU. Add or remove a GPU, it automatically adds or removes a supporting CPU core.

Yup, I have already had that in app_config since the beginning, so we're good there!

It looks like you have dropped a core in the bios. I was suggesting you run 4 cores but use the local option of the Boinc Manager to drop it to 75% of the cpu. This provides that "extra" cpu core whenever it is needed without you having to re-boot the system.

No, I did not drop a core in the bios, I just restricted Boinc to 3 cores. Windows still has the ability to use that spare core. If I want to switch back to 4 cores all I have to do is edit app_config and restart boinc.

---

Seti@home classic: 1,456 results, 1.613 years CPU time

It looks like you have dropped a core in the bios. I was suggesting you run 4 cores but use the local option of the Boinc Manager to drop it to 75% of the cpu. This provides that "extra" cpu core whenever it is needed without you having to re-boot the system.

No, I did not drop a core in the bios, I just restricted Boinc to 3 cores. Windows still has the ability to use that spare core. If I want to switch back to 4 cores all I have to do is edit app_config and restart boinc.

The website says you have 3 cores. Since when you drop to 75% of your available cores in BOINC, the Seti website still reports 4 cores, I suspect that something odd is going on.

Tom

---

A proud member of the OFA (Old Farts Association).