LTT Official Folding Month 2020!!!

[chaozbandit]

1 minute ago, LubyMoo said:

I won't meet the 20 day requirement, that's why I ask.

If you resumed folding today, you still have over five weeks of folding time

[Baha]

4 minutes ago, LubyMoo said:

I won't meet the 20 day requirement, that's why I ask.

The folding month does not end at the end of October, it ends on November 22nd

[GOTSpectrum]

Well TeamLTT seems to be showing what high gear looks like, but will we see a over 3 billion day before the event is over?

 

[RollinLower]

26 minutes ago, GOTSpectrum said:

Well TeamLTT seems to be showing what high gear looks like, but will we see a over 3 billion day before the event is over?

should definitely be possible. would be awesome if we did!

[Collin_S]

6 hours ago, yaboistar said:

it's a real shame because part of the reason i got my 3900x was so i could throw 24 threads at this and hope it wouldn't interfere too much. alas, no. still, it's useful for video rendering etc.

my AMD system doesn't have any slowdown when i'm CPU folding until i'm 12/12 on threads there. I usually run 10/12 so there's overhead for FAH and watching youtube.

That's interesting. I have a 3900x and an RTX 2070 in my rig. I used to be folding at a lower thread count but I have raised it up recently to let the client pick its own at max (claims 23 threads but actually uses 22 according to task manager).  I did notice that when I raised the threads, my points went down initially, but that was at my own custom thread counts. Now it is on auto and full, and even though I don't have any OC on my CPU besides what might be happening automatically, I did turn OC from for the entire CPU to per CCX and that did seem to help with performance some.

 

Does Medium VS Full make a difference for GPU?

[Legxcy]

3 hours ago, Spotty said:

The little 1080Ti that could.

 

 

 

CHOOO CHOOOOO

Over 3M PPD estimated on this WU. More of these please.

Jesus, is the 1080Ti that much more powerful than the normal 1080? I've got my OC'd +144/+602 also and I'm only getting 1.5m PPD

[DirectingGnu2]

7 hours ago, yaboistar said:

it's a real shame because part of the reason i got my 3900x was so i could throw 24 threads at this and hope it wouldn't interfere too much. alas, no. still, it's useful for video rendering etc.

my AMD system doesn't have any slowdown when i'm CPU folding until i'm 12/12 on threads there. I usually run 10/12 so there's overhead for FAH and watching youtube.

I got a 3950x and CPU folding still bogs me down badly lmao.

[Spotty]

7 minutes ago, Legxcy said:

Jesus, is the 1080Ti that much more powerful than the normal 1080? I've got my OC'd +144/+602 also and I'm only getting 1.5m PPD

I think that was just a particularly good WU. I also had it OC'd and 120% power limit when I was on that WU.

Right now I'm running 70% power limit and estimated 2.1M PPD.

 

There is a pretty big gap between 1080 and 1080ti though.

https://folding.lar.systems/folding_data/gpu_ppd_overall

[DirectingGnu2]

I'm noticing that f@h is a lot more tolerant of core overclock on my Asus TUF 3090 (non oc version) over my EVGA 2080 super xc gaming. I could only run like +40 on my 2080 super, basically whatever MSI OC scanner got, while I've been running +110 on my 3090 with some system instability, but it crashes Windows before a WU fails, +80 for true stability.

[LubyMoo]

1 hour ago, Baha said:

The folding month does not end at the end of October, it ends on November 22nd

Oh wow my bad for not seeing that! Cool, thanks!

[Legxcy]

4 minutes ago, Spotty said:

I think that was just a particularly good WU. I also had it OC'd and 120% power limit when I was on that WU.

Right now I'm running 70% power limit and estimated 2.1M PPD.

 

There is a pretty big gap between 1080 and 1080ti though.

https://folding.lar.systems/folding_data/gpu_ppd_overall

yep I'm running 120% power limit, I guess for a 1080 1.5m PPD is quite good then. I didn't realise just how much of a gap there was between the non-Ti and Ti. My bad

[Legxcy]

2 minutes ago, yaboistar said:

there's ~1000 cuda cores between them, the 1080ti has more than 1/3 again as many cuda cores as the 1080, there's going to be a fairly big gap. plus we don't have many 1080's folding at the minute so new-WU stats are going to be lacking in the DB compared to 1080Ti's which we have no shortage of really

Yeah my bad, I didn't realise just how much of a gap there was between the non-Ti and Ti

[Baha]

7 hours ago, yaboistar said:

if you haven't done so already, assuming it's a NoVideo card, give it a memory OC and disable Force P2 using nvidia profile inspector. it'll help the PPD out a bit

I know this wasn't directed at me, but I wanted to thank you for the recommendation of disabling Force P2 for CUDA. I'd never heard of it, and the memory clocks on my card are tangibly higher now. Curious to see how much it changes my PPD at the end-of-day summary.

[Spotty]

Just now, Legxcy said:

yep I'm running 120% power limit, I guess for a 1080 1.5m PPD is quite good then. I didn't realise just how much of a gap there was between the non-Ti and Ti. My bad

This time last year I was doing 1.2-1.3M PPD with the 1080Ti. New COVID work units and CUDA support make a HUGE difference. I think the 1080 was around 800-900k back then.

[Legxcy]

6 minutes ago, Spotty said:

This time last year I was doing 1.2-1.3M PPD with the 1080Ti. New COVID work units and CUDA support make a HUGE difference. I think the 1080 was around 800-900k back then.

That's correct, I was getting around 1m from it before the CUDA update

[Collin_S]

8 hours ago, yaboistar said:

if you haven't done so already, assuming it's a NoVideo card, give it a memory OC and disable Force P2 using nvidia profile inspector. it'll help the PPD out a bit

Is this the right place to get that from?

https://nvidiaprofileinspector.com/download-nvidia-profile-inspector/ ?

[Cabanaj]

Well lesson re-learn for me this morning... got a failed unit on my RX 5700 XT in the night, and did not check this morning. That was a mistake. 
No production at all since then! 

Except for a weird "11  CPU units" in the middle of the night. 
 
I am kind of unlucky this month, happened a few times . Must check more often. Let this be a reminder to all. 
 

[Collin_S]

15 minutes ago, yaboistar said:

i'm sure there's more than one application that will do it, but this is the one i use

https://github.com/Orbmu2k/nvidiaProfileInspector/releases

Thanks. Didn't want to get something sketchy. So is the main thing I do with that is apply the setting for Folding@home?

 

Currently my 2070 is OC'd using the profile put together by MSI Afterburner's OC Scanner, so it is doing 1935MHz (with occasional short drops into the high 1800's). Power limit set to 120%, no memory OC (running at 6801) and with no overvolt. Fluctuating total PPD from anywhere around 2.2-2.3 Mil up to sometimes at 2.8 mil. Any ideas why?

 

CPU is folding also, letting client pick threads on full (claims to be using 23 threads but according to task manager has 22). For me it does seem like no CPU folding at all does reduce total PPD. Working on optimizing all of my settings.

[Baha]

1 hour ago, Collin_S said:

Thanks. Didn't want to get something sketchy. So is the main thing I do with that is apply the setting for Folding@home?

 

Currently my 2070 is OC'd using the profile put together by MSI Afterburner's OC Scanner, so it is doing 1935MHz (with occasional short drops into the high 1800's). Power limit set to 120%, no memory OC (running at 6801) and with no overvolt. Fluctuating total PPD from anywhere around 2.2-2.3 Mil up to sometimes at 2.8 mil. Any ideas why?

 

CPU is folding also, letting client pick threads on full (claims to be using 23 threads but according to task manager has 22). For me it does seem like no CPU folding at all does reduce total PPD. Working on optimizing all of my settings.

For the first question - yeah, just set the "Force P2 state for CUDA" setting to off for FAH, hit apply, pause your current work, and restart the folding at home client.

Your PPD will fluctuate throughout the day because different work units can be completed at different speeds. Totally normal!

CPU PPD changes are a little harder to track due to PPD estimate fluctuations, but personally I've found that removing one physical core per GPU present helps stop the CPU from impacting GPU folding speed very much. So if I had a system with 8 physical cores, 16 logical threads, and 2 GPUs, I would set FAH to 6 threads. But everyone seems to have a little bit of a different experience with it.

[Collin_S]

37 minutes ago, Baha said:

set the "Force P2 state for CUDA" setting to off for FAH

How do you do that? This is what I see, just selecting Folding@Home in the Profiles.

[d0ks]

5 minutes ago, Collin_S said:

How do you do that? This is what I see, just selecting Folding@Home in the Profiles.

Go to section 5 - Common, look for 3rd option, CUDA - Force P2 state select off from dropdown.

[LazyDev]

7 minutes ago, Collin_S said:

How do you do that? This is what I see, just selecting Folding@Home in the Profiles.

Scroll down, its under 5 - Common

[Collin_S]

Thanks. My GPU has 7 mins left on its current WU so I will let it finish that before restarting. Will I need to keep NVIDIA Profile Inspector open after getting FAH started with the right settings?

[Collin_S]

8 minutes ago, Collin_S said:

Thanks. My GPU has 7 mins left on its current WU so I will let it finish that before restarting. Will I need to keep NVIDIA Profile Inspector open after getting FAH started with the right settings?

WHOAH!!!! (I literally said that aloud). That simple change made my GPU TPF go from 1min 30 seconds down to 40 Seconds! That is a CRAZY increase. THANK YOU GUYS SO MUCH FOR THAT TIP

[d0ks]

Be sure to drop dem likes to the original tip poster