No BOINC 16-17% of which 14-15% is FAH CPU control of the 2 GPU's

?
Sorry, but you are confusing me with the terminology you are using.
BOINC itself uses next to nothing. BOINC"s Projects, they make use of the CPU/GPU. FAH is a completely separate application, not related to BOINC in any way. But it makes use the CPU/GPU as well, hence the need to reduce the number of cores/threads BOINC can make use of so it doesn't have to try to compete with FAH.

FAH just kicked in hard and took the CPU up to 72% for a few seconds.
Note: FAH is set at Medium. If I set at low then nothing gets done. If I set it on high it takes up to much of the system.

And that could be the issue, if it occurs frequently enough, for long enough.
BOINC applications (usually) are set to low Priority. Any other application with a higher Priority level will take processing time away from them if there aren't any free cores/threads for it to use (same priority- share the core/thread processing time). So even though you have excluded a core from BOINC"s use, there are times where FAH home actually needs several cores and due to t's higher priority level it gets them.

Ok this is interesting...looking at Resource Monitor and it claims that I am using 110% of the CPU.
The highest values jump between WCG (CPU only) and Einstein (GPU only)

Don't use Resource Monitor for this, better just to use Task Manager or Process Explorer.

On my system Resource Monitor shows 116% CPU usage, it also shows 116% Maximum Frequency. Task Manager shows 100% CPU on the Processes tab.

I know FAH is separate, but I included it in the mix so you see what all is running.
I use the term "Boinc" loosely to describe the whole thing.
I reduced Boinc usage to 14 cores(90%). This should leave 1 core FAH and 1 core GPU control. I was 95% for 15 cores.

I still see 100% commit in processes (looking users tab 99.8 to a low of 96.1)
Performance shows the processor speed gained a little. Now running 3.98 to 4.02
Processes shows "Boinc" at 86% max
FAH seems steady at the moment. Now 7-8.2%
Vbox just showed up, which is LHC running, that is 4.8% max.

Probably better for the life of the chip in this configuration.

Greg_BE wrote

I know FAH is separate, but I included it in the mix so you see what all is running.
I use the term "Boinc" loosely to describe the whole thing.
I reduced Boinc usage to 14 cores(90%). This should leave 1 core FAH and 1 core GPU control. I was 95% for 15 cores.

I still see 100% commit in processes (looking users tab 99.8 to a low of 96.1)
Performance shows the processor speed gained a little. Now running 3.98 to 4.02
Processes shows "Boinc" at 86% max
FAH seems steady at the moment. Now 7-8.2%
Vbox just showed up, which is LHC running, that is 4.8% max.

Probably better for the life of the chip in this configuration.

The problem is Boinc, and most other computer applications, run at 100% for as many clock cycles as they are allocated, so yes Boinc runs at 100% for ie 9 clock cycles but then it backs off for ie 1 clock cycle and then runs at 100% again for ie 9 clock cycles and keeps repeating this process until it's either changed or stopped. Those numbers are based on a 10 count clock cycle and Boinc could be using a 100 count clock cycle or even a 1000 count clock cycle but you get the idea. The result is that in the overall average Boinc does 90%, based on your setting, but at times will still be running at 100%.

As for not using all 16 cpu cores at 100% 100% of the time yes I do that too to prolong the life of my parts, now if I had millions or billions of dollars, euros, whatevers to throw at the pc's then yes I would run them all at 100% and just replace the parts when they break, but my budget doesn't allow me to do that.

Greg_BE wrote

I know FAH is separate, but I included it in the mix so you see what all is running.
I use the term "Boinc" loosely to describe the whole thing.
I reduced Boinc usage to 14 cores(90%). This should leave 1 core FAH and 1 core GPU control. I was 95% for 15 cores.

I still see 100% commit in processes (looking users tab 99.8 to a low of 96.1)
Performance shows the processor speed gained a little. Now running 3.98 to 4.02
Processes shows "Boinc" at 86% max
FAH seems steady at the moment. Now 7-8.2%
Vbox just showed up, which is LHC running, that is 4.8% max.

Probably better for the life of the chip in this configuration.

The problem is Boinc, and most other computer applications, run at 100% for as many clock cycles as they are allocated, so yes Boinc runs at 100% for ie 9 clock cycles but then it backs off for ie 1 clock cycle and then runs at 100% again for ie 9 clock cycles and keeps repeating this process until it's either changed or stopped. Those numbers are based on a 10 count clock cycle and Boinc could be using a 100 count clock cycle or even a 1000 count clock cycle but you get the idea. The result is that in the overall average Boinc does 90%, based on your setting, but at times will still be running at 100%.

As for not using all 16 cpu cores at 100% 100% of the time yes I do that too to prolong the life of my parts, now if I had millions or billions of dollars, euros, whatevers to throw at the pc's then yes I would run them all at 100% and just replace the parts when they break, but my budget doesn't allow me to do that.

but my budget doesn't allow me to do that <--- neither does mine. There is a long story behind my latest setup. Short short verison, I made an error that toasted my mobo and then the repair shop missed a LED on the mobo saying the cpu was toasted. New cpu and repair labor costs almost equaled a new system. Have good cooling now (top rated for 2020 at least in the EU) this new cpu. So I need to protect my investment. I run only 16hrs a day. 14 cores for BOINC, 1 for GPU handling and 1 for whatever the system needs to use it for.

As for not using all 16 cpu cores at 100% 100% of the time yes I do that too to prolong the life of my parts, now if I had millions or billions of dollars, euros, whatevers to throw at the pc's then yes I would run them all at 100% and just replace the parts when they break, but my budget doesn't allow me to do that.

As long as the system is adequately cooled, running at 100% load 24/7/365 isn't an issue.
I've had systems running 100% 24/7/365 for over a decade. Power surges are what took them out, not crunching.

---

Grant
Darwin NT

As for not using all 16 cpu cores at 100% 100% of the time yes I do that too to prolong the life of my parts, now if I had millions or billions of dollars, euros, whatevers to throw at the pc's then yes I would run them all at 100% and just replace the parts when they break, but my budget doesn't allow me to do that.

As long as the system is adequately cooled, running at 100% load 24/7/365 isn't an issue.
I've had systems running 100% 24/7/365 for over a decade. Power surges are what took them out, not crunching.

Weird, because that was what was being said to me. I ran the system to hard. Stores can not or will not say that the CPU can hold up to that kind of strain since it is not a server cpu, etc.
I've got good cooling and I never ran the cpu over the max temp. It might have spiked there on and off, but never steady, but yet somehow I cooked the old one. BSOD and other instabilities were the norm.

I've got good cooling and I never ran the cpu over the max temp. It might have spiked there on and off, but never steady, but yet somehow I cooked the old one. BSOD and other instabilities were the norm.

Because it was running too hot; if it was running at the temperatures you say, then it did not have good cooling. Seriously- it was completely inadequate if the CPU was even remotely close to it's maximum rated temperature at any time.

The maximum is just that- the maximum. It should never ever even get close to that temperature when running under full load even when the ambient temperature is at it's highest.
My CPUs never get within 20°c of their maximum rated temperature (generally they get no higher than 70°c, other than the very occasional blip)- hence they last for years. If your CPU is at (or right near) it's maximum, then you should expect it to fail at some stage.
75°c is about as hot as you want any CPU or GPU to get for an extended period. If it's getting hotter than that, you need better cooling.

---

Grant
Darwin NT

I've got good cooling and I never ran the cpu over the max temp. It might have spiked there on and off, but never steady, but yet somehow I cooked the old one. BSOD and other instabilities were the norm.

Because it was running too hot; if it was running at the temperatures you say, then it did not have good cooling. Seriously- it was completely inadequate if the CPU was even remotely close to it's maximum rated temperature at any time.

The maximum is just that- the maximum. It should never ever even get close to that temperature when running under full load even when the ambient temperature is at it's highest.
My CPUs never get within 20°c of their maximum rated temperature (generally they get no higher than 70°c, other than the very occasional blip)- hence they last for years. If your CPU is at (or right near) it's maximum, then you should expect it to fail at some stage.
75°c is about as hot as you want any CPU or GPU to get for an extended period. If it's getting hotter than that, you need better cooling.

Multiple factors, manual OC and a confined space, even with lots of case fans.
Now it's out in the open so the heat is not confined and more importantly I am letting the BIOS and the CPU controller figure out how high of a OC it can do. I am not messing with that. I was told that was also part of the reason.

Multiple factors, manual OC and a confined space, even with lots of case fans.
Now it's out in the open so the heat is not confined and more importantly I am letting the BIOS and the CPU controller figure out how high of a OC it can do. I am not messing with that. I was told that was also part of the reason.

Yep, so the main reason would have been the lack of cooling.
At Seti there was a Nehalem CPU that was overclocked & overvolted way, way beyond it's rated values- and it ran reliably without producing compute errors for years. It only died when it's chiller unit stopped while the system was running- it was deceased within seconds.



The fact is that these days the CPUs do such a good job at overclocking themselves (boost/turbo boost etc), that it just isn't worth the effort to manually overclock them further IMHO. Unless you use a chiller unit or Liquid nitrogen to remove the heat produced, you're lucky to get more than a 5% gain over it's stock boost speeds even with a high end liquid cooling system. And to get that gain, you're looking at a 15% increase in power (at least) which just isn't worth it for me.
A good All In One (AIO) cooler, a large well ventilated case (or no case) and you'll get a high performance system that will last for ages- even when running at 100% 24/7/365.

---

Grant
Darwin NT

Multiple factors, manual OC and a confined space, even with lots of case fans.
Now it's out in the open so the heat is not confined and more importantly I am letting the BIOS and the CPU controller figure out how high of a OC it can do. I am not messing with that. I was told that was also part of the reason.

Yep, so the main reason would have been the lack of cooling.
At Seti there was a Nehalem CPU that was overclocked & overvolted way, way beyond it's rated values- and it ran reliably without producing compute errors for years. It only died when it's chiller unit stopped while the system was running- it was deceased within seconds.



The fact is that these days the CPUs do such a good job at overclocking themselves (boost/turbo boost etc), that it just isn't worth the effort to manually overclock them further IMHO. Unless you use a chiller unit or Liquid oxygen to remove the heat produced, you're lucky to get more than a 5% gain over it's stock boost speeds even with a high end liquid cooling system. And to get that gain, you're looking at a 15% increase in power (at least) which just isn't worth it for me.
A good All In One (AIO) cooler, a large well ventilated case (or no case) and you'll get a high performance system that will last for ages- even when running at 100% 24/7/365.

Once I got it out in the open (I had tried it in the confined space of a old computer desk) temps dropped off considerably. With todays electric prices in EU land going through the roof, I rarely leave my computer on for 24hrs like I will do tonight. Plus the SO wants a few hours without the fans, even though they are super quiet to me, but then I am "deaf" according to her.

I was suprised to see both of my machines, (4GHz i7's), were crunching just 4 Rosettas, the other virtual cores were idle. The tasks are a bit heavy on memory usage. I added a couple of projects, (SiDock and LHC),, to their portfolios for the time being.

* EDIT *

Later, I saw a couple of the Rosettas in the "Waiting for memory" state.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

I was suprised to see both of my machines, (4GHz i7's), were crunching just 4 Rosettas, the other virtual cores were idle. The tasks are a bit heavy on memory usage. I added a couple of projects, (SiDock and LHC),, to their portfolios for the time being.

* EDIT *

Later, I saw a couple of the Rosettas in the "Waiting for memory" state.

Which LHC project are you attached to? ATLAS is a memory hog and it prefers 4 cores.
Rosetta 4.2 uses on my machine between 800-1100MB per task.
Right now I am playing catch up on Rosetta so on a 16 core machine with 14 cores for BOINC I am running 13 tasks.

SiDock is not running right now so I can not see what its memory consumption is.

Your machines memory is lower than mine by 8 Gigs, you might want to increase it a little bit.
With 24 Gigs I can run LHC,Einstein,WCG,Rosetta,SiDock all at the same time depending on how BOINC splits them up. I also run FAH (non BOINC).

All this uses only 74% of 24 Gigs.

Grant,

After the reduction in cores to 14 out of 16 I checked on my 4.2 results.

This is impressive:

Run time 8 hours 8 min 2 sec
CPU time 7 hours 59 min 41 sec

Another one:

Run time 5 hours 49 min 18 sec
CPU time 5 hours 43 min 32 sec

#3

Run time 5 hours 45 min 30 sec
CPU time 5 hours 38 min 49 sec

And the rest follow the same pattern.

I've dug out the manual for my motherboards, (ASUS Sabertooth Z170 Mk 1), to check on the memory specs. The boards have four slots, so I am assuming, right now, that I can get another 16GB in each machine. It is fairly cheap nowdays.
I don't know why you say SiDock is not running, it is running fine for me...

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

I've dug out the manual for my motherboards, (ASUS Sabertooth Z170 Mk 1), to check on the memory specs. The boards have four slots, so I am assuming, right now, that I can get another 16GB in each machine. It is fairly cheap nowdays.
I don't know why you say SiDock is not running, it is running fine for me...

As in BOINC was not running it at the time.
That's average 65MB per task in memory.

What are you running on LHC? Sixtrack, ATLAS or ?

I've dug out the manual for my motherboards, (ASUS Sabertooth Z170 Mk 1), to check on the memory specs. The boards have four slots, so I am assuming, right now, that I can get another 16GB in each machine. It is fairly cheap nowdays.
I don't know why you say SiDock is not running, it is running fine for me...

As in BOINC was not running it at the time.
That's average 65MB per task in memory.

What are you running on LHC? Sixtrack, ATLAS or ?

I've dug out the manual for my motherboards, (ASUS Sabertooth Z170 Mk 1), to check on the memory specs. The boards have four slots, so I am assuming, right now, that I can get another 16GB in each machine. It is fairly cheap nowdays.
I don't know why you say SiDock is not running, it is running fine for me...

As in BOINC was not running it at the time.
That's average 65MB per task in memory.

What are you running on LHC? Sixtrack, ATLAS or ?

As for RAM, all websites say the same:
Max Unbuffered DDR4 SDRAM:
64GB

I could double that on my machine, but I don't see the need to at this point in time.
Right now with 4 WCG cancer and 2 einstein GPU and 4 sidock plus 5 rosetta I am only using 45% of my memory. That's a different story when ATLAS kicks in. I don't recall what they said that uses...8800 or something.

>>> What are you running on LHC? Sixtrack, ATLAS or ?

I've not had any work from LHC for a long time. It says it is pretty much Linux specific now.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

Sorry, won't let me edit.

>>> The project has been Linux only for a long while now.

From their website.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

LHC@home also has Windows apps:
https://lhcathome.cern.ch/lhcathome/apps.php

They certainly have had Windows versions of their tasks, I have a lot of credits from them. I've not had any work from them for a VERY long time, and in the message above is a direct quote from their home page. It is entirely possible it may come back as something I could crunch, but I'm not holding my breath when I see the comment I quoted above on their official page.

---

Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.

I've not had any work from them for a VERY long time
... I quoted above on their official page.

Could you post that issue at the LHC@home forum plus a link to that page?