Folding Forum

i appreciate all your good info. i am done researching this for now. FAH_22 uses the most memory x 2 for each gpu. bitdefender is next. radeon software is next. i am positive my issue is with fah. the past 24 hours has seen no ram increase. prior to that a long wu had accumulated to 7gb ram usage (total). it ran flat for several hours then ramped up. upon finishing it dropped to 5gb and hasn't gone over that since. several wu later it's running at 4.5gb. i've made no changes, continue to use browsers normally. i will continue to monitor wu until it happens again. i suspect some client is the root cause. work units vary in the way they copy information. i can watch that in real time. i will record any future work units that accumulate ram usage and post them.

cayenne187 wrote:i appreciate all your good info. i am done researching this for now. FAH_22 uses the most memory x 2 for each gpu. bitdefender is next. radeon software is next. i am positive my issue is with fah. the past 24 hours has seen no ram increase. prior to that a long wu had accumulated to 7gb ram usage (total). it ran flat for several hours then ramped up. upon finishing it dropped to 5gb and hasn't gone over that since. several wu later it's running at 4.5gb. i've made no changes, continue to use browsers normally. i will continue to monitor wu until it happens again. i suspect some client is the root cause. work units vary in the way they copy information. i can watch that in real time. i will record any future work units that accumulate ram usage and post them.

But even those values are insanely high.
This is mine after running for almost four days (columns are, in order: Process Name, PID, CPU, RAM)

how did you paste image. i can't do that.

cayenne187 wrote:how did you paste image. i can't do that.

I used imgbb.com, then fiddled with the "embed to forums" code that they provide post upload.

Mod Edit: Fixed Image Tags - PantherX

Mod Edit: Fixed Image Tags - PantherX

cayenne187 wrote:...i am positive my issue is with fah. the past 24 hours has seen no ram increase. prior to that a long wu had accumulated to 7gb ram usage (total). it ran flat for several hours then ramped up. upon finishing it dropped to 5gb and hasn't gone over that since. several wu later it's running at 4.5gb...

I am aware that some GPU Projects are big and can use ~2 GB of RAM per GPU. That would mean about 4 GB of RAM. Next time you see a spike of RAM, please do record the WU details for both of your GPUs. This information can be useful for future versions of FahCore_22.

Sometimes memory bugs happen, when you're running low on RAM, and the system is using SWAP space.
Just increasing the amount of RAM (in your case by an extra 4GB or so), could solve the issue.
There's a higher chance of data being written to swap, to get corrupted, than if it stayed in RAM.

I had a similar problem on Windows. Over time the RAM just filled up. It's unfortunately a Windows problem. For me the solution was to set a fixed amount of virtual memory, that solved the issue on both computers.
I'm back on Linux now, much less trouble.

this will probably be my final post on this. some wu accumulate ram usage. after a few hours ram usage goes up and up. mostly when wu completes it releases ram back. not always though. this may be a problem for some users. i have approached my 8gb ram limit when this occurs. i have added 8gb ram so it's no longer an issue. next step is getting fah to run as a user.

By chance, do you have the log file for when that happened or what Project those WUs belonged to?

11760 or 14563 were the latest projects that did it today. when i rebooted ram was near 3gb for fah. right now it's about 800mb which is normal. some work units do nothing. some work units accumulate ram but release it when done. some work units accumulate and don't release. rebooting returns to normal. my computer was hacked and my ssd destroyed 2 weeks ago. i thought it may have been a result of using drive for ram when i had 8gb. i have recovered from that and have 16gb ram now. the image i posted is a wu that released 2gb ram when done. some don't release and require reboot. thanks PantherX. if you want specific wu that do it i will try to record them but i am back to work now after lockdown so i can't monitor my pc much. what is important to me now is running fah as a user. logging on as a user is my primary firewall. my suggestions for fah is to standardize wu protocol so ram usage is stable. there should be some info for donors as to how much ram is used so we can upgrade if necessary.

Similar issue were some projects use 1.5 GB on Windows instead average of 500 MB.
viewtopic.php?f=38&t=35052

On a few occasions, I've had Core 22 fail to recover after encountering a NaN and attempting to restart from a checkpoint on a Radeon RX 460. GPU-Z reports load at 100% but memory clock is 300MHz instead of 1750MHz. Windows system memory usage for FahCore_22.exe remains constant. There's no progress being made so memory doesn't change as the work unit moves to the next phase in processing or writes a checkpoint. Rebooting will allow the work unit to continue or fail again and be reported back as bad. The most recent occurrence was on p14201, which is currently using 1.4GB of RAM.

cayenne187, please check the Sensors tab in GPU-Z next time if you don't see progress reported in the Log tab of FAHControl to see if the symptoms are the same.

i haven't had any failed wu to my knowledge other than loss of wifi or my system being hacked. i don't understand what sensors tab is. don't see it in radeon software. been folding for 24 hours now and total ram hasn't budged from 3.8gb. all depends on wu and the way they cache and upload segments. i am thinking too that this may be related to me having 2 gpus running. do you think many donors are running multiple gpus?