viming_aint_easy

Oh jeez I'm not been here for a few days. You guys are awesome though. We gotta keep it up and we will smash Curecoin and take the real number one spot! 

Niiiiiiiice 

I believe that this is a consequence of what GPU you have.  Nvidia cards use polling (which works best with dedicated cores), while AMD cards use interrupts (which doesn't care).  I'd link to some folding forum posts, but it looks like their SSL cert expired today. 😅

You need to keep in mind:   PPD is just fake internet points for us geeks to have a little fun trying to dial in our gear to get some more of them.   

They provide no value to the community of the scientists using the F@H project to process their data.
 
This means F@Hs primary goals are keeping the client / work servers working well enough to get the data out to us, us to crunch it in a reliable way, and get that data back to them in a time window which is days... not the hours we all strive to get them back to them in to max out PPD.
 
You don't even really get the PPD unless you get a F@H API key... which has no bearing on the science... you can do work just fine without one, but the numbers will look low, just something fun they added so Den-Fi you could get more points and show off / have fun.
 
So with this in mind... the software does not access or keep track of the voltages, over/under clocks... particular GPU models and editions by manufacturer etc. 

F@H would not want to spend their resources or time to integrate and expand the scope of the platform to do any of this... cause it could only introduce problems or barriers to installation (admin access, driver / os patch changes etc.) more so then they already have to deal with the current base service, and all is off the mission of research, remember for most users it's really intended to be on your computer and just never be seen using spare time on your hardware.
 
This is 100% the right attitude for them to have as it's about the science first to them which I respect.
 
So you get the log analyzers, services like I'm making, forums full of people with suggestions to help squeeze a little more PPD out and have some fun with the gamification of doing some charity processing for a good cause.   
 
But in the end the only thing that matters is that the work units get completed before they expire, and that your system is running stable 100% of the time so you're not wasting F@H researcher time creating bad / incomplete work units from overclocks etc.
 
This is why a person with a room full of 3090s is as important as the people folding on their laptops, each WU counts and the points don't matter. 
(unless your Den-Fi)
 

Yes, if you want to kill it to say restart It due to a stuck slot I usually
FAHClient --send-pause ps -ef | grep fahclient kill -KILL <child PID> kill -KILL <parent PID> Kind of the root equivalent of hitting it in the head with a hammer.

Ubuntu 20.04 LTS isn’t quite ready for prime tIme yet WRT folding. Mostly due to the age of the client and partially due to the usual fsckery Canonical does.
 
Oh, and the init script really needs a rewrite.

Meeting went well! Woo! 
 
You guys have been busy on the thread jeez. 

Currently under Windows or Linux you can run a 2080S on only 4 PCIe3 lanes with no discernible decrease in PPD and using PCIe bi-furcation and m.2 to PCIe3 x16 adapters you can get 6 x4 lanes and hence 6 GPUs running off a x570 board using a low end 4c/8t processor and 12-16GB system memory.
 
When you calculate the per-slot (GPU) cost of HEDT systems they're just not worth if for folding. They might make sense if you want to do BOINC concurrently but likely the per-thread cost of a 3900x or 3950 would still be lower than ThreadRipper.
 
My longest running system is a 12 year-old iTX motherboard with 4GB DDR3-1666 and a PCIe2 x16 slot and an Athlon II x64 at 2.8GHz. It will easily keep a 2070Super folding at full bore. As long as you keep your GPUs and CPUs running below 80C and don't abuse your VRMs mid-range Desktop systems will happily run BOINC and F@H for years with only the occasional dusting and less-frequently a re-pasting required.

@TheDailyProcrastinator As others have said CPU WUs don't yield a lot of points and these days you likely want to leave a couple extra threads free in addition to the 1 per NVidia GPU and 1 for all AMD GPUs recommended in the past. But some recent WUs use a lot of CPU when starting up so this may be why some folks have noticed a slight decrease in PPD when CPU folding or it may just be a Windows thing (I use Linux for all my dedicated folding systems).
 
But though I started building my systems with low thread count CPUs and the minimum amount of RAM now that I'm contributing to BOINC I've replaced the CPUs with higher thread counts and increased the RAM for BOINC.
 
x299/x399 systems just aren't worth it for multi GPU setups. A lower-tier x570 board that supports bifurcation, a PCIe3 x16 to 4 PCIe3 x4 in x16 slots, combined with 16-32GB RAM (for BOINC or 8-16GB for just F@H) and a R5 or R7 processor will happily feed 6 mid range NVidia GPUs in a mining frame (see my sig and this thread) and cost less per GPU than an x299/399 system.
 
The downside of having many GPUs per system is when, not if, but when, the WU servers start screwing up and you have to reboot to get a single slot folding again you roll ALL the GPUs back to the last Checkpoint and it gets messy.
 
Though I like AMD GPUs, and bought a couple of RX5700XTs, they're just not as efficient at folding as NVidia cards even without the recent boost in NVidia cards due to an improved CUDA core.
 
Theoretically a 5700XT should be able to match a 2070 - 2070S in PPD but in practice they yield about the same as a 2600 used to (1.1 - 1.3MPPD). This appears to come down to AMD Driver issues and given their track record getting their gaming drivers working properly I don't expect to see them putting the effort into the compute portion of the GPU drivers required for Navi or Big Navi anytime soon.

You didn't know about the daily updates? Nice!

Well I guess you all waiting with bated breath for day 2 results so it's time for me to get to work. The points geathering script is about  to run.

And I need a box full of 3090s, but here we are 

Big thank you for helping @leadeater diagnosing the issue, @marknd59 and @cbigfoot helping with validation, 
 
Can I just say, you are all very naughty folders, invalid folding names and duplicates breaking our sweet sweet script! 

Same thing applies to any GPU. I don’t think there’s a way to easily power limit on Navi.
 
I imagine what you’d want to do is a negative clock offset and/or an under volt both of which will have the same effect of reducing the core clock and hence the power.
 
In practice, once you get an idea of what performance (PPD) the card does at stock and what the clocks and power draws are at that point, you reduce the clocks by 20-25% and see what your performance at that power is. PPD/W.

Just an FYI, I will be unavailable for much of today due to personal circumstances, I have responded to all outstanding messages and I will be popping in from time to time if possible. If you need anything in the meantime @marknd59 has full capability to check entries and the like. 
 
Thank you for your understanding as always!
 
 

I'm calling this one "scrappy" 'cause it was made from parts lying around
Asus H170m-Pro
i5 6500
1 x Gskill 4GB DDR4-2400
Samsung 128GB SSD
Corsair 750W PS
Veddha Minercase 6
EVGA GTX 1070Ti, 1060 6GB
GigaByte 1060 6GB
Ubuntu 18,04.4 LTS
 
1070Ti in PCIex3 x16 Slot, EVGA 1060 in m.2 slot (x4)
Gigabyte 1060 in Chipset PCIe3 x4 slot
 
3.4MPPD on current WUs at Stock Power Limits and Clocks
Load average is about 3.25 and it's only using 512kB Swap.
 
Who says you can't fold on a potato!
 
Me thinks I need to go buy some RAM as the first set of Large WUs that come along are going to bring this to it's knees.

*sad radeon noises*

afaik the atom count is tailored towards making the WU's workable on most systems. that's why with large atom count moonshot WU's we saw pretty dramatic PPD increase on high power cards.
this is me recalling past discussion threads tough, don't quote me on this.

Oh, don't you worry.  This puppy will keep chugging along even if I were to get 2 3080's.  I have a nice spot on my local NAS (repurposed from my previous desktop) that would fit beautifully.

You are right it should be about the science and this is good for the science but I am far more invested in the science than most. If I could I'd send you some of my extra points to keep you folding with your RX5700.

So, after some arts and crafts time, please can we all welcome the New and "Improved" Foldatron AKA The Janky Dust Magnet. (Now with 100% more hash browns)
 

 
Also can we get a prayer for this poor RTX 2060 with a dead (And now removed) Fan.

 

Headless is a work in progress... needs to have that dashboard portal to see how your machines are performing over the course of a WU to see if something is leeching GPU / PPD during the folds.

Like the current development branch of the web client does

 

Oh, hell yeah, you are correct, I forgot to remove Beta from my main rig. That is totally my bad not noticing and saying that in my post.