• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Mowing over high PPD folders from yesteryear

Agent_N said:
Hehe, I took your spot in the top 50!
Click to expand...
You probably did. I'll add a 4070ti to my server next month. I don't care if it heats up the server closet. Depending on the heat output I may put the one I plan on getting for my gaming rig to folding too.
 
firedfly said:
I've got 1.5 weeks before you pass me. I should be able to slow that down a bit during the competition in a few days.
Click to expand...
or maybe longer since I am taking 3 days here to focus on Primegrid before jumping back in for the contest. I think i might run it awhile after that though in between the other PG challenges, the Pent etc.
 
motqalden said:
Speaking of mowing...
Holdolin is absolutely crushing it.
Click to expand...
Well to be fair, this thing we do is the only "discretional expense" I have (I have no life at all outside of this) so I try to "go all in" as my kids put it hehe. Have to admit it's kinda nice not having to buy any heating fuel, as my fleet keeps the house warm in the winter and my basement warm in the summer.
 
Xilikon said:
Nice to see old faces here! I do remember I reached the top 10 (or it was the first position, been a while) with a stack of 20 work laptops waiting to be assigned and all folding with some computers at home. I remember breaking 20k ppd and it's so high. Today, A single 3090 can rack millions of ppd
Click to expand...
My ancient Athlon II 435 triple core can hit around 20k PPD!
 
Holdolin said:
Well to be fair, this thing we do is the only "discretional expense" I have (I have no life at all outside of this) so I try to "go all in" as my kids put it hehe. Have to admit it's kinda nice not having to buy any heating fuel, as my fleet keeps the house warm in the winter and my basement warm in the summer.
Click to expand...
I wish I had some extra money for "discretional spending" to buy moar GPUs for moar PPD!
 
Projects like this are killing PPD on the 4090. Time to upgrade to a 4070ti :ROFLMAO:

1677886831646.png
 
motqalden said:
thats bizzare...
Click to expand...

Ineed, and it's not the only project where the 4070ti is beating the 4090 by a large margin. At first I thought it was just me but it seems there are oddities in general.

I ruled out a driver issue, I rolled back to 527.56, same issue. The PPD drop does not coincide with a windows udpate install. That leads me to believe it is a fah_core or WU issue.

Another bizzare one, 18723.

1677892372492.png
 
Last edited:
There are definitely issues with the PPD on some work units. When I first got my Radeon 6750xt the work units I was pulling were mostly doing 2 million PPD but would occasionally get some which only did 1.2 million PPD which is seriously messed up. There's no way it should have been doing only a little more than twice the PPD of my RX570. Once I set the client to prefer Alzheimer's work units it started pulling in 3 million PPD.

I have no idea how they're doing PPD calculations anymore but something is definitely messed up with the calculation for some of the work units.
 
SmokeRngs said:
There are definitely issues with the PPD on some work units. When I first got my Radeon 6750xt the work units I was pulling were mostly doing 2 million PPD but would occasionally get some which only did 1.2 million PPD which is seriously messed up. There's no way it should have been doing only a little more than twice the PPD of my RX570. Once I set the client to prefer Alzheimer's work units it started pulling in 3 million PPD.

I have no idea how they're doing PPD calculations anymore but something is definitely messed up with the calculation for some of the work units.
Click to expand...

For sure some WU pay out more than others but I'd assume that a 4090 ( > 2x CUDA cores) would complete the same WU faster than a 4070 ti. In most cases it does. It seems the client needs to be updated to utilized the increased core counts / chip architecture.

Alz 18601 as an example.

1677897149290.png


I know task manager isn't the best tracker of GPU utilization, but this was interesting. The client switched from a 18723 WU to 18213. GPU utilization doubled. I'll switch th graph to CUDA when the nex low PPD WU hits.

1677897661475.png
 
Last edited:
None of my 4090's have ever dropped below 16M PPD and certianly never slower than my 4070Ti's. Must be (another) windoze issue lol
 
Holdolin said:
None of my 4090's have ever dropped below 16M PPD and certianly never slower than my 4070Ti's. Must be (another) windoze issue lol
Click to expand...
Could well be. Might create a Linux USB stick and see what happens.
 
Looking at my WU history, it looks like there was a base credit reduction from 200k to 120k on the 18723 WUs. looks like it happened about 2023-02-24. Sorry, my research is slow as it's getting late here or rather my day started much earlier than I wanted it to haha
 
Holdolin said:
Looking at my WU history, it looks like there was a base credit reduction from 200k to 120k on the 18723 WUs. looks like it happened about 2023-02-24. Sorry, my research is slow as it's getting late here or rather my day started much earlier than I wanted it to haha
Click to expand...
That makes sense on the ppd drop. Still wondering why the 4070 is faster.
 
There was an issue a few years back where smaller WU didn't like GPU's with lots of cores. The general theory was if a project was too small to keep all the cores busy it was bouncing around the cores not running as efficently as it could. Not sure if it was ever proven.
 
Nathan_P said:
There was an issue a few years back where smaller WU didn't like GPU's with lots of cores. The general theory was if a project was too small to keep all the cores busy it was bouncing around the cores not running as efficently as it could. Not sure if it was ever proven.
Click to expand...

Seems plausible. The 4070 ti is at the top on some WUs and the 4080/4090 are essentially identical. There is something wacky going on for sure.
 
If that happens with boinc you can sometimes overcome it by running multiple work units at once. Wonder if you could do something similar with fah using vm's
 
  1. I thought about going the VM route but I'm a bit lazy about that. I'll wait to see if the issue gets fixed.
  2. The 4070 ti and 4090 have similar boost clocks. My 4090 is running >3Ghz
 
Because of QRB, F@h rarely benefits from running multiple units. How does power consumption look during the "low PPD" events? I do agree with Holdolin's sentiment; after running F@h on Linux, it is really hard to go back to Windows, lol.
 
Power consumption seems "normal". What flavor of Linux are you using, I tried usuing the latest Ubuntu release (22.04.2) and the FAH control app doesn't respond. I have installed python but it seems the specific version that FAH control wants has been depricated.
 
russnuck said:
Power consumption seems "normal". What flavor of Linux are you using, I tried usuing the latest Ubuntu release (22.04.2) and the FAH control app doesn't respond. I have installed python but it seems the specific version that FAH control wants has been depricated.
Click to expand...

I use web control in mint for same reason
 
russnuck said:
Power consumption seems "normal". What flavor of Linux are you using, I tried usuing the latest Ubuntu release (22.04.2) and the FAH control app doesn't respond. I have installed python but it seems the specific version that FAH control wants has been depricated.
Click to expand...
A big problem with FAHControl on Linux is the old version of Python it uses. When I recently did an update FAHControl refused to start up on either my server or my main machine and it's because of Python. I found this quite annoying so I uninstalled it and installed a gtk3/python3 fork of FAHControl instead and so far haven't had any problems.

The problem with using old Python2 is not limited to FAHControl as there are quite a few other utilities and such which still require it and it has caused some issues since Python2 has basically been completely deprecated.

I'm running Manjaro on both my main machine and server. I've had great luck with it and have been running it as my primary OS for over two years now with few problems. It is a stable semi-rolling release distro so you'll typically see more updates than on something like Ubuntu since it updates more often. That said, I absolutely love the AUR for when I can't find something I need or want from the normal Manjaro repos. If you think of trying it out I will warn you against going hog wild with the AUR. There can be occasional problems/issues with stuff installed from the AUR especially when something like Python2 is deprecated. The issues usually aren't a big problem to fix but they do occur. The rule of thumb is to install any software from the normal Manjaro repos if available and only use the AUR for software which isn't in the normal repos.
 
It looks like my x570 Gaming Plus has weird issues with the 4090. I migh revist Linux after I am confident those issues are resolved or I buy a new mobo.
 
Just to toss it out there: have any of y'all tried the V8 beta client? I've given it some basic testing and seems to work ok. The monitor is web-based so there is no python crap to worry abou.
 
Took some troubleshooting and searching but it appears that the 4090 hates UEFI, at least on my MSI x570 Gaming Plus.
  1. To get into the BIOS I'd have to boot into Windows and then restart.
  2. The BIOS flash utility wouldn't display on screen.
  3. Sometimes I'd lose signal when switching KVM sources. Only restarting would fix the issue.
  4. This was 100% reproducable if I was in something other than Windows (memtest boot, BIOS, Linux on occasion)

I set the BIOS to CSM and it seems much happier now. I will get back to Linux boot when I'm sure the issues are cleared up.
 
You must log in or register to reply here.
Back
Top