Folding Community Board

[GOTSpectrum]

Guys, I'm right in thinking to have the option of getting beta work units you just use the client-type advanced flag right?

[GOTSpectrum]

1 hour ago, yaboistar said:

Spec... you might want to check your mind into specsavers mate

 

client-type advanced = "Advanced" work units

client-type beta = Beta work units

Shush... It's beer festival week in my town

 

Thanks for that though lol 

[Gorgon]

On 1/29/2020 at 3:18 PM, yaboistar said:

i suspect you got lucky on your cards as my partner's 1070ti (well, yeah, technically mine) won't go above 1800 at a low power limit

No, not that unusual.  My EVGA 1070 Ti XC can run for weeks at 2010 MHz. It gets a little hot (mid 70s) when pushed that hard. It was running a little hotter after around a year of 24x7 but repasting it helped.

[GOTSpectrum]

I just got a core 22 WU on the standard branch, project 11737. Core 22 is officially being rolled out!!!

 

I'm seeing a nice boost for my 1080ti too!

 

Check out the link below for the announcement over on the folding forum. 

 

https://foldingforum.org/viewtopic.php?f=24&t=32144

[GOTSpectrum]

I do have to say though, it makes my 1080ti a toasty boi!

 

 

70C on a hybrid card!

[GOTSpectrum]

46 minutes ago, yaboistar said:

spec, on an unrelated note, what ppd numbers are you seeing for your 2060?

1.3-1.45m PPD on core 22 work.

 

I'm hopefully going to grab a KO soon to compare

[GOTSpectrum]

For anyone wondering I have seen about a 25-30% increase in PPD from my GTX 980 FE with core 22

[GOTSpectrum]

Ignore this

[Metallus97]

52 minutes ago, GOTSpectrum said:

For anyone wondering I have seen about a 25-30% increase in PPD from my GTX 980 FE with core 22

Thats sort of in line with what i have on my VII. Maybe 30-35 but depends on the Atome count and stuff. How is power draw? Stayed about the same for me with the VII at stock

[GOTSpectrum]

2 minutes ago, Metallus97 said:

Thats sort of in line with what i have on my VII. Maybe 30-35 but depends on the Atome count and stuff. How is power draw? Stayed about the same for me with the VII at stock

on Maxwell around the same, on Pascal and Turing though ,I am seeing higher power, but not much more. 

[Gorgon]

On 2/1/2020 at 8:59 AM, GOTSpectrum said:

1.3-1.45m PPD on core 22 work.

 

I'm hopefully going to grab a KO soon to compare

The KO has no observable advantage over a regular 2060. It’s fixed 100% Power Limit is a disadvantage as is it’s anaemic 1660Ti cooler.

[Favebook]

36 minutes ago, Gorgon said:

The KO has no observable advantage over a regular 2060. It’s fixed 100% Power Limit is a disadvantage as is it’s anaemic 1660Ti cooler.

Saw a review on Jayztwocents's youtube? 

[TVwazhere]

Uh oh

[TVwazhere]

2 minutes ago, yaboistar said:

FAH stats are updating, i suspect Jason is doing some more work.

[GOTSpectrum]

1 hour ago, TVwazhere said:

You have to give it to the guy, he's always working on something!

[Metallus97]

2 hours ago, TVwazhere said:

Uh oh

Same... maybe another server hickup?`

[Favebook]

1 hour ago, yaboistar said:

@Favebook

 

run

 

  Reveal hidden contents

DOO-DOO-DOO-DOO-DOO-DOO-DOO-DOO-DOO-DOO-DOO

 

OY... At least I will hit 500M before you surpass me. 

However, I'll have to put my 1080 Ti in folding machine so you do not surpass me... EVER  

[TVwazhere]

Well, the core 22 rollout has bumped me from 1.5MPPD average to looking more like 1.6M. I checked ITNOS this morning, F@H client reported a 1.7M PPD estimate, but that's probably high assuming the core22 values are all the same and completable in the relative same time. Both cards were indeed bumping into the 60% power limit, though the top card (FTW2) was now also getting close to the 65ºC cap as well. Not sure if this is because of the new WU's or just being the card with less access to air in general...

 

Notably, both cards were in the 1800 and 1700 mhz ranges (the 1800 being the top FTW2 card) despite both having +150mhz core offsets and +300mhz memory offsets. I know I've seen them both at 1900ish ranges in the past, but that was probably before core22. I tried moving them up to 65% which did increase the core speeds, but also increased the power draw on my Kill-a-Watt meter from 319 to 340, so I'm gonna leave it at 60% for now.

It's pretty obvious when Core 22's rolled out, wouldnt you say?  

[GOTSpectrum]

3 minutes ago, TVwazhere said:

Well, the core 22 rollout has bumped me from 1.5MPPD average to looking more like 1.6M. I checked ITNOS this morning, F@H client reported a 1.7M PPD estimate, but that's probably high assuming the core22 values are all the same and completable in the relative same time. Both cards were indeed bumping into the 60% power limit, though the top card (FTW2) was now also getting close to the 65ºC cap as well. Not sure if this is because of the new WU's or just being the card with less access to air in general...

 

Notably, both cards were in the 1800 and 1700 mhz ranges (the 1800 being the top FTW2 card) despite both having +150mhz core offsets and +300mhz memory offsets. I know I've seen them both at 1900ish ranges in the past, but that was probably before core22. I tried moving them up to 65% which did increase the core speeds, but also increased the power draw on my Kill-a-Watt meter from 319 to 340, so I'm gonna leave it at 60% for now.

It's pretty obvious when Core 22's rolled out, wouldnt you say?  

Honestly the mem-OCs make pretty much no difference, if you kill them you free up some watts for them sexy silicon cores (in theory) But either way no point pushing the memory hard for folding. Just wearing the mem chips and mem controller down over time. 

[TVwazhere]

7 minutes ago, GOTSpectrum said:

Honestly the mem-OCs make pretty much no difference, if you kill them you free up some watts for them sexy silicon cores (in theory) But either way no point pushing the memory hard for folding. Just wearing the mem chips and mem controller down over time. 

I can bump them back to stock to see if that does anything, but my thought was the memory OC can support a more stable clock OC. I've seen that in other cases, but admittedly I'm not sure entirely how that relationship works, or if it matters at this level of OC

[GOTSpectrum]

Just now, TVwazhere said:

I can bump them back to stock to see if that does anything, but my thought was the memory OC can support a more stable clock OC. I've seen that in other cases, but admittedly I'm not sure entirely how that relationship works, or if it matters at this level of OC

From my experience of running 2000-2100 mhz on 10 series cards with no memory OC I don't think it makes all that much difference 

[GOTSpectrum]

1 minute ago, yaboistar said:

i run my cards at stock memory and bumped core. memory OC does not matter much to FAH by all the seems of it.

One interesting thing I have noticed, running single channel on system memory seems to affect folding though, not exactly sure why. Especially the nre core 22 seems to be affected

[GOTSpectrum]

Just now, yaboistar said:

core22 uses about 1gb of system memory which iirc is more than c21 so if memory usage is up and it's actively using that memory, it would make sense that access speed and latency would help

That make sense then, I've not exactly had time to look into it yet as I've been busy with work and the like. Moving house finally some time this month is going to be a ball-ache to be sure. 

[GOTSpectrum]

1 minute ago, yaboistar said:

honestly i really want to know why my 2080Ti is racing through C22 units like it's on a five day meth binge and it's || <-this close from an aneurysm

My understanding is that it uses the CUDA cores more efficiently, so that especially on high atom count units, more of the die is in use. Though would also explain the higher power draw too. I wonder if we need to do some more research into PCIe bandwidth. I feel like core 22 may be more pissy with it. 

[GOTSpectrum]

Anyone who's folding on a 2070, got numbers for C22? 

 

@Gorgon @Den-Fi I think you two have some 2070 cards.