LTT Official Folding Month VI

[justpoet]

Had units throttled up a bit ago, but it got warm.  I'll be throttling up the CPUs again probably tomorrow, and of course through the event.

 

Gonna give the beta a try on something too, to see how that goes.

[Alex Atkin UK]

8 hours ago, justpoet said:

You could do it like most heating thermostats do, which is to call off the heat (set WUs to Finishing) a degree early, but keep circulating so the heat continues to work out of the boiler/furnace until the correct temp (the Finishing time of working, and the cooler ramping down), or a degree above.  Then when it drops to a degree bellow, start WUs.

Yeah its complicated, because mum has Diabetes so I can't predict what temperature is comfortable for her at any given moment anyway.  Also we have the bathroom fan on constantly for air quality reasons and to dry the washing.

 

So despite electricity being more expensive than gas, were probably making quite efficient use out of Folding overall.

[Antistatic12]

Happy to help out again.
I probably won't be able to participate as heavily like the previous years, but I'll give it a good crack.

[Dutch_Master]

12 hours ago, justpoet said:

they just don't turn out current GPU numbers, even if the CPU is current

So, what I hear you saying is that if one wants to make headway on the numberboard, CPU-crunching is basically dead.

 

There goes that 3rd full EPYC system I was planning to buy

[RollinLower]

4 hours ago, Dutch_Master said:

So, what I hear you saying is that if one wants to make headway on the numberboard, CPU-crunching is basically dead.

 

There goes that 3rd full EPYC system I was planning to buy

i think @leadeaterusually manages to get some decent scores just CPU crunching. but that is on quite the number of chips.

[leadeater]

6 hours ago, Dutch_Master said:

So, what I hear you saying is that if one wants to make headway on the numberboard, CPU-crunching is basically dead.

 

There goes that 3rd full EPYC system I was planning to buy

CPU on anything around 12 cores and modern does quite well. Zen 3 is quite a big leap per core and Intel has always been quite good just more limited on total core counts until more recently.

 

A good 12 core CPU does 1m PPD which is the same as a RTX 3050 or Tesla T4.

 

So if you start pushing around 100 cores, about an RTX 3080 Ti, then you'll be doing very well, just horribly on the efficiency front.

[robertobsc]

@GOTSpectrum Do you usually reach out to the scientists when we have these events? I have the feeling we may run out of WU's depending on how much folding power we increase.

[leadeater]

24 minutes ago, robertobsc said:

@GOTSpectrum Do you usually reach out to the scientists when we have these events? I have the feeling we may run out of WU's depending on how much folding power we increase.

F@H is a huge global project, our smaller monthly events don't actually have any impact in regards to running out of WUs. The only time that was a real issue was middle of the pandemic when we had something around 9000 participants in the LTT folding month and many many more overall.

 

Edit:

8144 to be exact, I kept the event data 

 

Over 22 days the event participants made 40,930,185,882 points. Last month the entire LTT team, all 119k of them, did 30,992,890,674.

[justpoet]

19 minutes ago, robertobsc said:

I have the feeling we may run out of WU's depending on how much folding power we increase.

Worth noting here is that while you may read about running out of stuff in BOINC Pentathlon that LTT also participates in, as @leadeater said, this isn't really an issue with F@H. F@H isn't as fine grained project specific, and everything fails over to what is available after preferences, unlike in BOINC.

[justpoet]

@LAR_Systems, do you have a beta extension I could run to help you collect data to compare V8 vs V7 clients?

[dogwitch]

main pc. corrupt mbr....

had to replace main drive and re install windows.

also tracking down a softcrash bug to atm.

[Alex Atkin UK]

On 10/6/2023 at 2:01 PM, Dutch_Master said:

So, what I hear you saying is that if one wants to make headway on the numberboard, CPU-crunching is basically dead.

 

There goes that 3rd full EPYC system I was planning to buy

I'm only doing CPU for the heat and contribution as the PPW is awful:

 

Bearing in mind the CPU is pulling 145W and the GPU 170W on these WUs.

[Alex Atkin UK]

On 10/5/2023 at 5:10 PM, podkall said:

I've got a question and good news.

 

First the question, why is my F@H lagging even when not working or when CPU is limited to less than all cores.

 

(2nd smaller question is that when selecting a thread count is it actually selecting the physical core count? because the F@H uses 7,5 theads when I select 4 "threads" in the configure)

 

And good news, seems like the thermal issues on my GPU are partially caused by the fan curve (partially by huge gap between average and hotspot temps),

88% fan speed lets me use 75% power limit, instead of previous 55% and the hotspot temp stays under 80°C.

 

100% more efficient than 30 series? so 4070 pulls double 3070 at same wattage?

I was getting 1M more PPD from my 4070 Ti at 170W than my 3080 which I didn't check but was probably pulling 320W.

[robertobsc]

18 hours ago, leadeater said:

F@H is a huge global project, our smaller monthly events don't actually have any impact in regards to running out of WUs. The only time that was a real issue was middle of the pandemic when we had something around 9000 participants in the LTT folding month and many many more overall.

 

Edit:

8144 to be exact, I kept the event data 

 

Over 22 days the event participants made 40,930,185,882 points. Last month the entire LTT team, all 119k of them, did 30,992,890,674.

Amazing stats! I do remember f@h running out of WU’s during the covid time. One of the few good things of that time was seeing how f@h could scale. 

[Livin]

Missed this for many years but looking forward to doing it again :).  Hardware has changed a ton since the last time I participated so interested in seeing how the new stuff does.

 

Out of curiosity, is it worth throwing both CPU and GPU at folding or is CPU going to be burning more power than it's worth for the results?

[TylerD321]

2 minutes ago, Livin said:

Missed this for many years but looking forward to doing it again :).  Hardware has changed a ton since the last time I participated so interested in seeing how the new stuff does.

 

Out of curiosity, is it worth throwing both CPU and GPU at folding or is CPU going to be burning more power than it's worth for the results?

 

10 hours ago, Alex Atkin UK said:

I'm only doing CPU for the heat and contribution as the PPW is awful:

 

Bearing in mind the CPU is pulling 145W and the GPU 170W on these WUs.

Cpu probably isn’t worth it but your best source I know of is @Alex Atkin UK

[BiotechBen]

37 minutes ago, Livin said:

Out of curiosity, is it worth throwing both CPU and GPU at folding or is CPU going to be burning more power than it's worth for the results?

Depends if you are using it to heat a room. If you are: heat good, if you aren't using it like a space heater: not amazing returns on CPU, but it also heavily depends what CPU architecture you have, newer is significantly better. (I've got low tdp Skylake and it's atrocious Pts/W)

[BondiBlue]

1 hour ago, Livin said:

Out of curiosity, is it worth throwing both CPU and GPU at folding or is CPU going to be burning more power than it's worth for the results?

Out of all the machines I fold on I only do CPU folding on two, and that's only becauase they're going to be on 24/7 otherwise. Both of those machines are ancient, so they're terrible in terms of PPD, but I don't really mind. I basically never bother with CPU folding on anything else. 

[Livin]

6 hours ago, BiotechBen said:

Depends if you are using it to heat a room. If you are: heat good, if you aren't using it like a space heater: not amazing returns on CPU, but it also heavily depends what CPU architecture you have, newer is significantly better. (I've got low tdp Skylake and it's atrocious Pts/W)

Not looking for any more heat, so I'll probably leave it off then.  It's a 13700k, so it can take a chunk of extra power to run.

[TylerD321]

Anyone have any tips for maximizing PPD per machine for GPU? Any tuning to be done in bios w/ fan curves? Drivers? MSI Afterburner? I assume best practice is having all other programs turned off but I am sure there are things I am missing.

[BondiBlue]

4 minutes ago, TylerD321 said:

Anyone have any tips for maximizing PPD per machine for GPU? Any tuning to be done in bios w/ fan curves? Drivers? MSI Afterburner? I assume best practice is having all other programs turned off but I am sure there are things I am missing.

What sort of GPU are you using? On most of my cards I tend to just max out the power and thermal limits with Afterburner, and as long as temps are good I'll leave them like that. For the cards I'm running I don't do any modifications to the clocks, but I know some people adjust the clock speeds to achieve certain goals, including efficiency. 

 

With the increased power limit I do like to monitor things with GPU-Z to make sure I'm not hitting any temperature limits, including hot spot temps. One of my cards can't maintain correct hot spot temps with the stock fan curve, so I manually raise the fan speed for that card. The others are usually fine. 

[TylerD321]

7 minutes ago, BondiBlue said:

What sort of GPU are you using? On most of my cards I tend to just max out the power and thermal limits with Afterburner, and as long as temps are good I'll leave them like that. For the cards I'm running I don't do any modifications to the clocks, but I know some people adjust the clock speeds to achieve certain goals, including efficiency. 

 

With the increased power limit I do like to monitor things with GPU-Z to make sure I'm not hitting any temperature limits, including hot spot temps. One of my cards can't maintain correct hot spot temps with the stock fan curve, so I manually raise the fan speed for that card. The others are usually fine. 

Currently have a 3070 and a 3050 I am running. Here in the next couple of weeks I will be swapping out the 3050 for a 3080 and getting a 3070 ti as well!

 

Edit: It is worth noting that I have hearing aids and normally wear headphones when I play games anyway so sound really isn't an issue. You can see my specs on the main machine I am running FAH in my signature - I have the fans minimum speed on bios set to 25% and then they go up from there - I can't hear them, much louder and I can though. I haven't adjusted MSI afterburner but according to HWINFO my GPU usually sits around 70 with hot spot at 80, the thermal throttle temp is 83 it says.

Edited October 8, 2023 by TylerD321
More info

[NetoriusNick]

@JamesTheFolf your folding username does not match the one you provided (JamesTheFolf) can you please check and link the Extremeoverclocking page   https://folding.extremeoverclocking.com/search.php

Or confirm you have never folded ever under that name?

[GOTSpectrum]

Just now, Schnoz said:

How much CPU does F@H need, and is it multithreaded? My dual 2080 Tis are happily pulling 5-6M PPD each on my 2699A V4-equipped system, but the thread that FAHControl runs on is almost always pegged at 100%. Is each GPU able to use a separate CPU thread for control?

Generally on higher end GPUs you have one CORE, so two threads available per GPU plus one CORE for the OS 

[podkall]

1 hour ago, GOTSpectrum said:

Generally on higher end GPUs you have one CORE, so two threads available per GPU plus one CORE for the OS 

CPU? I thought it assigned each individual cores