Folding@Home bigadv EOL 2nd Anniversary Challenge on 12:00 UTC January 23rd: The race is over.

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
I hoped that 'client-type normal' (as opposed to advanced or beta) would not receive FahCore_a8 work. But alas it does. I may have to give up on CPU folding.¹

Here is another forum thread on FahCore_a8:
https://foldingforum.org/viewtopic.php?f=72&t=36097
It's user feedback, and they all speak well of FahCore_a8.

I don't have a foldingforum account and am not going to create one, although this a8 issue irks me a lot. This is such a huge regression; it's strange that nobody has it reported yet. (Maybe because most people are like me and can't be bothered to create a foldingforum account.)

————————
¹) Don't get me wrong; I am talking solely about CPU folding in really large CPU slots. My opinion stands that CPU folding is complete and utter nonsense, unless you have a 56-threaded computer or better. I would never fold on CPU slots smaller than this, including in contests. The little credit they give for CPU folding means that they don't care for your CPU work, and that they want you to use your CPUs elsewhere where it actually can do something good.
 
Last edited:

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,219
3,798
75
That has been my experience as well, outside of that I do not have any machines that are 64 threads or higher. I was excited to get an A8 task thinking the PPD would be much better due to AVX2/FMA3 and it being a newer core, but the PPD was significantly lower than A7 tasks on both a Ryzen 3950x and Haswell Xeon.
Funny, my Haswell Xeon is faster on a8. I dedicate real cores to my GPUs and don't use hyperthreading.

Also, my Haswell Xeon is about the same speed as my 750 Ti, on the a7 core. So I don't consider CPU folding that inferior.
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
my Haswell Xeon is about the same speed as my 750 Ti, on the a7 core. So I don't consider CPU folding that inferior.
Dual 1080Ti @ 250 W TBP, fed by i7-7700K @ 3.6 GHz, nothing running besides FahCore_22:
5.6 M PPD / 570 W = 9.8 kPPD/W​

Dual 7452 @ 180 W PPT (~2.8 GHz) running FahCore_a7:
2.8 M PPD / 370 W = 7.6 kPPD/W​

So this 7nm CPU folder is quite close to the 16nm GPU folder in efficiency, but only because it has got a lot of threads = far bigger quick return bonus than most other CPU folders. (Watts are "at the wall", i.e. whole system power. PPD are with the random workunits which were running at the time of this post, but reasonably representative.)
 
  • Wow
Reactions: crashtech

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
Screenshots attached of current WU's that are running on my machines. Both machines are CPU only, no GPU, running based on FAH auto configuring the slot so it's using all threads. I understand that each project gives different PPD, etc. but I seem to pretty consistently get less PPD on the A8 core.

Here is a 3950x (32 thread) working on an A8 task for estimated ~365k PPD:

1-27-21-a8.PNG

Here is a 3900x (24 thread) working on an A7 task for estimated ~471k (or 467, depending where you look) PPD:

1-27-21-a7.PNG

Edit - and now an A7 task I just got on the 3950x getting ~579k(or 576k depending where you look) PPD

1-27-21-a7-3950.PNG
 
Last edited:
  • Like
Reactions: Assimilator1

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
External temperatures have been wobbling around the freezing point here all week.

Weather forecast claims that there is going to be a night with -13 °C (9 °F) ........ after this race is over. :-|

Very low temperatures would be good because then I could put several more hundred Watts through my computers. Temperatures near freezing point are inconvenient, because cooling capacity is limited then, and at the same time I am not inclined to open the windows for better cooling while I am not at home. (For example, if the cable modem connection acts up, it'll get cool in here, although there is central heating on minimum setting to prevent the worst. The risk would be lower if F@H had work buffers like BOINC.)
 

Skillz

Senior member
Feb 14, 2014
911
929
136
Luckily I live on the 2nd story so I keep my windows open all the time when it's cool outside. Even have a box fan in one window helping the cooler air circulate inside. Granted, it doesn't get THAT cold here, but still in the upper 30s (F) and lower 40s (F) this past week.

With all my GPUs running (and CPUs) my apartment usually hovers around the upper 60s (F) and lower 70s (F). If it gets into the 20s (F) then I have to turn the fan off because it'll get in the 50s (F) in my apartment real quick. lol
 

Skillz

Senior member
Feb 14, 2014
911
929
136
Hey, so why were your WUs failing part way through?

I was using an old client build plus something to do with opencl. They were failing before starting though.

apt install ocl-icd-opencl-dev

This command I ran and it seemed to fix it. Though the client logs still complain about something with opencl or something, but they've been running and completing tasks so I'm not touching it anymore. Soon as this contest is over I'll go back to mining so I'm not too concerned with it. Long as they finish and gimme the points.
 

motqalden

Member
Jun 13, 2009
82
111
106
During the summer when i was working away from home i had my window box fans setup on wifi switches with timers that would come on at certain times of the day to give the cooling a little boost. I would also periodically teamviewer in to check temps and manually turn on more fans if needed :p
 

Skillz

Senior member
Feb 14, 2014
911
929
136
I have mine on a smart plug also (the fan) and I use my blink cameras to check the temps.

I stopped using TeamViewer because they kept locking my stuff up. Went to AnyDesk a few months ago and haven't had any issues at all.
 
Last edited:
  • Like
Reactions: motqalden

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,434
136
I was using an old client build plus something to do with opencl. They were failing before starting though.

apt install ocl-icd-opencl-dev

This command I ran and it seemed to fix it. Though the client logs still complain about something with opencl or something, but they've been running and completing tasks so I'm not touching it anymore. Soon as this contest is over I'll go back to mining so I'm not too concerned with it. Long as they finish and gimme the points.
Yes, I posted this somewhere in the DC forum a while back....
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,718
136
I too had windows open the entire time in Q2–Q4 of 2020. I often also needed to have a pedestal fan shoving hot air out, in any of the seasons, including several times in December (which a few of us ran in pentathlon mode). I decided for myself that I'll return to a somewhat more normal lifestyle this year.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,434
136
I had my 2060 super hung for some reason. I lost about 1.5 million ppd in 12 hours. Back to 40 million ppd/.

A reboot fixed it, odd.... (linux box, EPYC 7742 cpu)
 

Assimilator1

Elite Member
Nov 4, 1999
24,120
507
126
CPU changes don't take effect until the next WU starts as far as I understand.
I've left the 2nd rig on 12T overnight (I meant -1 setting earlier as well as now, not fixed 12T), the CPU usage is still ~83%, so their is no point in setting 10T as it reserves a core for GPU crunching. Which is what I thought happened with recent versions of F@H.
My main rig is still set to 10T threads and current total ppd is 497k, so the higher ppd earlier must have just been down to score variations between WUs.