LTT Official Folding Month V (2022)

[podkall]

@GOTSpectrum Yo, so I was casually folding when I decided to stop, when I do that I usually wait for a full percentage to go, so if its 12,56% I wait when it surpases 13%, after I stop I also wait for like a whole few minutes in case the program needs some time to save the progress,

 

and usually that works, but sometimes I close it with 13% and open it with 0%, also I didn't run out of deadline it was actually a new WU I have 5 days to complete the work unit.

 

Any idea why it went from 13 to 0%?

[justpoet]

16 minutes ago, podkall said:

@GOTSpectrum Yo, so I was casually folding when I decided to stop, when I do that I usually wait for a full percentage to go, so if its 12,56% I wait when it surpases 13%, after I stop I also wait for like a whole few minutes in case the program needs some time to save the progress,

 

and usually that works, but sometimes I close it with 13% and open it with 0%, also I didn't run out of deadline it was actually a new WU I have 5 days to complete the work unit.

 

Any idea why it went from 13 to 0%?

Sometimes if you have multiple slots, or it thinks configurations have changed, it will reset.  As for the % rule, it is more important to see when the checkpoints are, which are configurable.

[gravitywave]

On 10/19/2022 at 4:09 AM, Schnoz said:

Woohoo! Great to have you on board!

 

Thanks!

[gravitywave]

23 hours ago, Gorgon said:

For larger cards (2070s & up) you want at least PCIe3x4 or they will start decreasing in performance. Generally using PCIex1 mining risers you will see some degradation in performance. The more performant the card the higher the impact. Windows tends to be more chatty on the PCIe bus than Linux so is more impacted.

I've never really tested this, but whenever I'm folding I never see the bus interface load (via GPU-Z) at more than 1-2% on any of my cards. Is there something more at play here?

[Slave2school]

I just upgraded to AM5 but I'm too poor to get a new gpu right now, still giving it my all with 14 cores going full blast 1 left for netflix and 1 left to feed the 3080 ti. Anyone doing this on am5 should tune the PBO curve carefully. Mine was "stable" at -30 across the board until I started folding. Minus 25 now is good. I'm heating my basement with this shizz.

[podkall]

1 hour ago, justpoet said:

Sometimes if you have multiple slots, or it thinks configurations have changed, it will reset.  As for the % rule, it is more important to see when the checkpoints are, which are configurable.

Also, I've clicked the protein viewer and it jumped from 0,something% to 20,something%, so maybe it just doesn't show the accurate progress..

[podkall]

1 minute ago, Slave2school said:

I just upgraded to AM5 but I'm too poor to get a new gpu right now, still giving it my all with 14 cores going full blast 1 left for netflix and 1 left to feed the 3080 ti. Anyone doing this on am5 should tune the PBO curve carefully. Mine was "stable" at -30 across the board until I started folding. Minus 25 now is good. I'm heating my basement with this shizz.

+1 heater

[IkeaGnome]

4 hours ago, dogwitch said:

installs dual 2080 Tis with malicious intent .    hahaah. that made me luagh.

what about me? i need a crazy line due to hurricane.

Water cools the whole side of the country for more performance. 

[BiotechBen]

8 hours ago, Shlouski said:

Well, it took just over a week.

 

I'm impressed, it actually makes my phanteks evolv x case look small.

 

Hopefully the A770 will arrive next week, along with my bifurcated riser.  

 

 

 

*Halo: Reach music starts playing*

[BiotechBen]

3050ti is raking in 1.1-1.2M PPD now, +125K from 5800H, 600K from 480, 100K from 2600x.

2M PPD here I come.

[dogwitch]

2 hours ago, IkeaGnome said:

Water cools the whole side of the country for more performance. 

how did you know...

[Shlouski]

7 hours ago, Alex Atkin UK said:

Irony is it would probably improve the chipset cooling on my X570 compared to the 3080, as for some bizarre reason they thought putting it right under the GPU was a good idea.  That said, not honestly sure if the graphics card airflow is a good thing given its potentially forcing the chipset fan to move faster.

 

Was worse when I had the 2080 in there though as being a blower the plastic case almost entirely blocked it.

 

The board in my picture is also a X570, the fan doesn't spin at all anymore after setting the fan to silent mode, fortunately the temps seem fine. Like you say graphics cards cover the chipset, a combination of the gpu and case fans blowing over it seem to do enough. I hate these little fans as they often end up one of two ways, not working or making annoying noises, mine is the latter.

[Shlouski]

6 hours ago, Gorgon said:

Nice - I'd suggest leaving that system's cause preference set to Alzheimer's for a bit to see what it can do with WUs with large atom counts. Unless, of course, you want to get one of those small WUs that will make it look like a 2070

 

I'd also be interested to see at what frequency peak efficiency can be found and if it falls even further off the cliff at low frequencies than Ampere does.

Thank you, I have set it to Alzheimer's for now.

 

Later today I will likely use your method to profile the gpu to get some accurate numbers, but I have been playing with the power slider and observing the stabilized changes to ppd reported in FAHControl.

 

These are my numbers:

 

PL 100% = 370W - 23 MIL PPD    Core Clock 2775mhz

PL 90%   = 350W - 22 MIL PPD    Core Clock 2750 - 2775mhz         (wattage fluctuation high)

PL 80%   = 350W - 22 MIL PPD    Core Clock 2700 - 2750mhz         (wattage fluctuation high)

PL 70%   = 320W - 21 MIL PPD    Core Clock 2500 - 2700mhz

PL 60%   = 270W - 21 MIL PPD    Core Clock 2400 - 2500mhz

PL 50%   = 220W - 19 MIL PPD    Core Clock 2100 - 2200mhz

PL 40%   = 180W - 13 MIL PPD    Core Clock 1300 - 1500mhz

PL 30%   = 130w  - 9   MIL PPD    Core Clock   800 - 900mhz

[cbigfoot]

9 hours ago, TVwazhere said:

  Reveal hidden contents

March 27, 2020. No photo/meme.

 

April 6, 2020

 

April 11, 2020 BOINC request, no image

 

April 15, 2020 Boinc Bronze, no image

 

May 6, 2020

 

May 17 2020,

 

 

I cant find the plat request interestingly enough

I think platinum I posted smart @$$ meme on the event thread and you issued it

[Gorgon]

58 minutes ago, Shlouski said:

Thank you, I have set it to Alzheimer's for now.

 

Later today I will likely use your method to profile the gpu to get some accurate numbers, but I have been playing with the power slider and observing the stabilized changes to ppd reported in FAHControl.

 

These are my numbers:

 

PL 100% = 370W - 23 MIL PPD    Core Clock 2775mhz

PL 90%   = 350W - 22 MIL PPD    Core Clock 2750 - 2775mhz         (wattage fluctuation high)

PL 80%   = 350W - 22 MIL PPD    Core Clock 2700 - 2750mhz         (wattage fluctuation high)

PL 70%   = 320W - 21 MIL PPD    Core Clock 2500 - 2700mhz

PL 60%   = 270W - 21 MIL PPD    Core Clock 2400 - 2500mhz

PL 50%   = 220W - 19 MIL PPD    Core Clock 2100 - 2200mhz

PL 40%   = 180W - 13 MIL PPD    Core Clock 1300 - 1500mhz

PL 30%   = 130w  - 9   MIL PPD    Core Clock   800 - 900mhz

Looks like the sweet spot (at least for 18601) is around 2.0-2.2GHz on Ada - Nice

[Gorgon]

7 hours ago, gravitywave said:

I've never really tested this, but whenever I'm folding I never see the bus interface load (via GPU-Z) at more than 1-2% on any of my cards. Is there something more at play here?

PCI Bus Utilization is notoriously difficult to measure as it's clocked so fast and getting stupidly fast these days with Gen4 and 5 at play. About the only way to figure out if your bus bottlenecked is by measuring yields of the same WU in different slot configurations. There's a 18 page thread over at Folding Forum where people did just that. In my BiFrost build in my sig I also did some testing when using m.2 to PCIe3x4 in x16 risers to see what impact, if any, was there going down to PCIe3x4 from x8 and x16. TLDR is Linux - No Problem - Windows about a 20% decrease in PPD for higher end GPUs. Granted this was before Windows improvements with Core22 ...

[dogwitch]

21 minutes ago, Gorgon said:

PCI Bus Utilization is notoriously difficult to measure as it's clocked so fast and getting stupidly fast these days with Gen4 and 5 at play. About the only way to figure out if your bus bottlenecked is by measuring yields of the same WU in different slot configurations. There's a 18 page thread over at Folding Forum where people did just that. In my BiFrost build in my sig I also did some testing when using m.2 to PCIe3x4 in x16 risers to see what impact, if any, was there going down to PCIe3x4 from x8 and x16. TLDR is Linux - No Problem - Windows about a 20% decrease in PPD for higher end GPUs. Granted this was before Windows improvements with Core22 ...

very true. tr intresting  with lanes and such for mobos

[GOTSpectrum]

On 10/3/2022 at 11:54 PM, GOTSpectrum said:

20x 100USD LTTStore Giftcards

So we have some juicy prizes from @jakkuh_t

[Alex Atkin UK]

2 hours ago, Shlouski said:

Thank you, I have set it to Alzheimer's for now.

 

Later today I will likely use your method to profile the gpu to get some accurate numbers, but I have been playing with the power slider and observing the stabilized changes to ppd reported in FAHControl.

 

These are my numbers:

 

PL 100% = 370W - 23 MIL PPD    Core Clock 2775mhz

PL 90%   = 350W - 22 MIL PPD    Core Clock 2750 - 2775mhz         (wattage fluctuation high)

PL 80%   = 350W - 22 MIL PPD    Core Clock 2700 - 2750mhz         (wattage fluctuation high)

PL 70%   = 320W - 21 MIL PPD    Core Clock 2500 - 2700mhz

PL 60%   = 270W - 21 MIL PPD    Core Clock 2400 - 2500mhz

PL 50%   = 220W - 19 MIL PPD    Core Clock 2100 - 2200mhz

PL 40%   = 180W - 13 MIL PPD    Core Clock 1300 - 1500mhz

PL 30%   = 130w  - 9   MIL PPD    Core Clock   800 - 900mhz

They really weren't kidding, that efficiency at PL 50% is insane.

[Captainmarino]

17 hours ago, Shlouski said:

Well, it took just over a week.

 

I'm impressed, it actually makes my phanteks evolv x case look small.

 

Hopefully the A770 will arrive next week, along with my bifurcated riser.  

 

 

 

Holy COWWWWWW! Would you look at that behemoth?! 

[CT854]

So, I've been realizing that my RTX 3080's been only doing ~2.2M PPD on Alzheimer's, so I decided to investigate.

 

Turns out using my PS5 hooked up to a PCIe capture card was enough to decimate my folding production. Guess I've gotta hook up my PS5 directly to a display if I want to play MLB The Show

[rkv_2401]

8 hours ago, Gorgon said:

PCI Bus Utilization is notoriously difficult to measure as it's clocked so fast and getting stupidly fast these days with Gen4 and 5 at play. About the only way to figure out if your bus bottlenecked is by measuring yields of the same WU in different slot configurations. There's a 18 page thread over at Folding Forum where people did just that. In my BiFrost build in my sig I also did some testing when using m.2 to PCIe3x4 in x16 risers to see what impact, if any, was there going down to PCIe3x4 from x8 and x16. TLDR is Linux - No Problem - Windows about a 20% decrease in PPD for higher end GPUs. Granted this was before Windows improvements with Core22 ...

When are we starting Gorgon's FoldingTechTips?

[cbigfoot]

2 hours ago, rkv_2401 said:

When are we starting Gorgon's FoldingTechTips?

I thought he already did

 

also haven’t seen my pal @Zberg sign up yet. You up for a challenge. 

[Alex Atkin UK]

OK, I'm not THAT desperate. https://www.amazon.co.uk/dp/B0BHD9TS9Q

or even this much https://www.amazon.co.uk/dp/B09YCLG5PB

[rkv_2401]

21 hours ago, GOTSpectrum said:

Can you guys post some SUs to try to get people to join up to the event 

I wrote a nice one, I hope someone sees it lol