System stability and efficiency
22 Dec 2020 21:58:19 UTC
Topic 224280
(moderation:
Over the past 3 days I've read hundreds (thousands?) of posts on the "Subject". Very helpful! I still have a number of questions, though.
I'll start with the simplest and most basic.
1. (Simplest) What does RAC stand for? (Googling it was not enlightening)
2. (Basic) Fan and temperature control. I have tried AI Suite 3 (Since my MB and GPU's are ASUS), MSI Afterburner and GPU-Z. All 3 of the above are running at present (plus some BIOS setup that I haven't gone back into). I know that this is hardly desirable!!!
AI Suite 3: I tried a former version about a year and a half ago and swore I'd never touch it again. This latest version seems much better, but still does not answer all my needs (desires?). I appear to get pretty good CPU & case fan control and temperature info, but haven't got real control of my GPU fans and temps. (I've also got a water pump and a small M.2 fan with decent data and control). The fan speed seems fixed. It's possible that AI Suite 3 can't override a setting in Afterburner. I've read a lot of negatives about AI Suite 3, but (as I said) the latest version seems much improved.
MSI Afterburner: In the past this has been my "go to" system. My main problem is that I'd like to see a lot more averaging on some of the outputs. (graphs). Through former posts, I've learned the importance of setting the % Computer time to 100% (Thanks Gary!!). That really stabilized the temp cycling. (for good reason). My GPU usage is bouncing all over the place!
GPU-Z: Just for info purposes right now.
So - I'm trying to elicit your thoughts on how to control fans, temps, workloads. I've managed to get 2 GW work units/GPU running, but no WCG cpu work is getting loaded. Most of you will probably laugh when I mention the temps (GPU and CPU) at which I get nervous, let's just say they're much, much lower than temps I've seen mentioned in the posts. I'm revising my expectations, but it will take a while!
Has anyone tried AI Suite 3 recently?
That enough for now. Again, thanks for all your past posts. Great info!
Merry Xmas and a better New Year - stay safe
Doug
MontanaDoug wrote:1.
)
RAC is an acronym for Recent Average Credit. A simple English explanation right from the official web site reads.
"...is computed as an exponentially weighted average with a half-life of one week. In other words, if an entity is not granted any credit for one week, its RAC will decline by 50%."
For really gory details you can read directly, paying particular attention to section 4.1
BOINC credit page at Berkeley
This gets pretty personal. My own preference is to use MSIAfterburner as the control interface. While the graphs in it are somewhat helpful, when I want averages over a particular interval (up to many days), I use HWiNFO (there is a button to click to restart all averaging). They both run most of the time on my machines. I have not noticed adverse interactions.
MontanaDoug wrote:Over the
)
The problem with case fans is there really isn't alot of 3rd party software that can control them because there are sooooo many cases makers and models. The same goes with water cooled stuff, it's not that it isn't mainstream it's just that there isn't alot of people screaming for it to be controlled by a 3rd party.
As far as why your WCG tasks won't run you need to check your resource share settings and you may need to just wait it out until they do run, that can takes weeks if not months to sort things out.
I skipped past the point that
)
I skipped past the point that the OP was interested in case and CPU fan control as well as his interest in GPU control and monitoring.
Long ago I had bad experiences using Speedfan for those purposes--not just not controlling the fans well, but actually conflicting with other software on the system. That was over five years ago. Meanwhile, even though the motherboards I buy are not high end at all, they offer considerable control over case fans and CPU fans, and that is where I do that function.
Even though I now have had two generations of AMD graphics cards, and am shopping for a big Navi, I don't use the AMD GPU fan and clock control software, preferring MSIAfterburner for those functions.
archae86 wrote: Even though
)
That's the only one I use too as it works on both AMD/ATI and Nvidia gpu's. Now different cards have different controls, with the older cards have less control than the newer ones but that works for me
Archae86 and Mikey, Thank
)
Archae86 and Mikey,
Thank you for your prompt and informative replies.
Items:
1. WCG tasks woke up last night and now seem reasonably in line with my settings.
2. I'll try using Afterburner and HWiNFO solely, uninstalling AI Suite 3. (Will stop BOINC while I'm playing with this - don't want to melt the little snow we have this year in Montana).
A couple of other tidbits (especially since you're working with 100+ times the number of GPU's that I am)
a. I've had a good deal of "luck?" , heat soaking defective GPU's. (ie. sticking them in the oven for about 10 min - I forget the exact times & temps to be used - easily Googled). Just be sure to remove the fan half and the soft plastic plugs protecting the inputs (ruined one with melted plugs!) The technique "heals" microfractures in the card.
b. I'm working on a liquid methane cooling system (-161.6°C). Will let you know when it's perfected - great for superconductivity.
c. I'm not always serious.
Again, thanks for the feedback and please stay safe!
Doug
MontanaDoug wrote: Archae86
)
LOL!!!
If it were me I would find a way to put a fan in a window and blow that cold Montana air into the room where the pc's are IF they are in their own room, obviously if they are in a family area that's not a good idea.
mikey wrote: If it were me I
)
I have a dedicated "man cave". Wife lives across the street. Being Montana, we jerk our front doors open at exactly 5pm daily and are allowed 1 shot (not shotgun) at each other. Fortunately, neither of us is very steady but passerby's risk everything.
Thus the PC is in a dedicated room (converted Master bedroom). I have toyed with the idea of installing a closed loop system to fan force outside air through the PC then exhaust it to the outside. Would be very useful in the summer when it regularly gets in the mid to low 50's at night. I do have a whole house fan that can cool things off rapidly in the evening - no need for A/C in this part of MT (but some people here do waste energy with one)
I've been working on my system balance. I was successful in changing to 2 jobs/GPU but my recent change to 3 jobs/GPU doesn't seem to have taken (I'll be patient a bit longer to see if "preferences" get accepted.
There are a few people in the threads that have imposed minor reduced voltages to their GPU's. I've never played with overclocking or other kinds of tweaking. Your thoughts? (People have commented that their temps decrease without hurting their throughput).
Annoyance!! In BOINC
)
Annoyance!!
In BOINC Manger>Options>Computing preferences, It states that it is using web based preferences from WGC. I've tried many different ways of changing that to Einstein - NO LUCK!
Suggestions?
This is part of my system stability project. Progress is being accomplished (at a snail's pace)
Some more questions: 1.
)
Some more questions:
1. My recent change to try 3 jobs/GPU seems to have done something. I had been running consistently 2/GPU, now I'm running 3 on device 0, and 1 on device 1. Am I limited in GPU memory?? CPU memory??? device 0 is my 8GB unit, device 1 is 4GB.
2. "Running (0.9 CPUs + 0.5 AMD/ATI GPUs(device #))". Can you refer me to an explanation of this, like - why .9 CPUs and is that toggled in the preferences somewhere? I also thought that the 0.5 GPUs would change when I went to 3/GPU. but no????
Fun on Xmas eve!
Cheers, Doug
BOINC applies the very last
)
BOINC applies the very last web preferences updated at ANY project and propagates/applies that to ALL projects.
WCG in your case seems to be the last preference changed.
The propagation is usually slow and inconsistent depending on when projects are updated either automatically or manually in the Manager.
Try and change one of your Einstein venue parameter settings and Update/Save on that page. Then hit update in the Manager and hopefully that is picked up by the Manager and shown in the Options>>Computing Preferences screen.
If that doesn't pick up the change, try exiting BOINC and restarting to have BOINC round-robin contact each of your projects at startup.
Or try and change to Local preferences in the Manager temporarily and then shifting back to Web based preferences to have the Manager pick up your Einstein preferences.