#folding-and-boinc
1 messages Β· Page 84 of 1
I'd see if your PPD goes up if you pause the CPU
For your v100's I'd say you need 2-3 CPU threads each
Also according to smi each card is only pulling 200W out of 400
By default f@h only gives 1 thread
I feel this may be hampering your cards
If it were me I would even go as far as to assign each CPU slot a CCX die on the epyc.
And leave 1 CCX "empty" just to feed the GPU's
@topaz flame was that paused for a long time? Because in the picture you sent your CPU PPD is awful.
I have a i7-4790k with 5 threads that puts out that kind of PPD
Reduced the cores to 200 for the 256 available leaving more than enough for the 8 cards
will watch
started getting errors because of no jobs that size, reduced it to 16 cores and will slot it up later π (now that I know whta the config looks like)
If it were me I would even go as far as to assign each CPU slot a CCX die on the epyc.
And leave 1 CCX "empty" just to feed the GPU's
@hushed bobcat
Hard mode:
Manage thermals and assign 90% utilization on every other core so thereβs a less used core in between.
Since he's in a rack I don't think thermals are a problem π
made a post on F@H Support forums just to make sure its not a software issue π
We got a few CPU in the mix now
sweet, im quite curious if that makes a difference in your PPD
does SMI show more power usage?
@bronze minnow nope
Official answer from F@H peeps https://foldingforum.org/viewtopic.php?f=80&t=36079&sid=81ac5215c02ad04cffb09c2f00a375d0
They are going to make some adjustments to get things squared away π
well damn
4 Billion points per day
that's double the production of the entire LTT Team
60k WU per day
i'm doing good if I do 60
me confused what im looking at can i get some details ?
We get points for completing Work Units while folding
Generally we express power in how many points we can accrue in one day
PPD = Points Per Day
@spice badger yp
no u
ill check in a bit
im using a new fah_core22 bin
and using a given project id
specifically to benchmark them
Developer of fah is interested in the new hardware π
maybe because its a bench WU?
if thats the final number they did some serious optimizing somewhere lol
What kind of setup is kicking out 335,918 WU?!?!?!? DAMN!
The 4th's ranked anonymous.
@vital radish all the poeple that signup for LTT team but did not setup a folding name for themself.

@rich turtle those numbers don't make any sense ! 693W is like only like what 4 or 5 GPU's !!

My 4 GPU run at around 550W
π
A 2070 super and 1070 pushed 650 watts
Im pulling around 1450w~
A 2070 super and 1070 pushed 650 watts
@stuck escarp something not right there. 2070 super should be 215W and a 1070 should be 150W.
my 1080 Ti, P106-90, and 8700k rig pulls about 450W
but the 1080 Ti and 8700k are power limited to reduce heat/noise
I've got 2x 2060 and 2x 2060 super all set to 80% power limit. The 2060 draw an average 135W and the supers 140W and I can't see any real difference in PPD than running them at full power.
yea same here
i ran my 1080 Ti at 120% at first
i think it was pulling 250-275W
went down to 75%
now its pulling 180-200W
PPD is maybe 100k less
its hard to tell because different WU's have different PPD
If the PPD is less it's not enough to worry about. The cards run a little cooler and and my power bill is a little lower so it's all good.
i agree completely
Hey everybody don't forget LTT Folding Month 2020 Event Starts October, 12th with lots of prizes to be won don't forget to signup.
Hello, im just starting out Folding@Home with my laptop, i have a question about the cpu speed
Why does the cpu speed clocks down when its at 50% utilization? Is that something i must worry about?
It could be throttling due to temperature
Very few laptops can handle a sustained heavy load
@tired crow Looks like they didn't quite get it fixed. Woke up to less point than I should have again.
@rich turtle I'm sure they will turn up sooner or later again
Oh yeah, for sure. It's nothing new, just happened right at a point where I deployed new hardware for it sucks from a gathering data standpoint.
Things always go wrong when you least want them. That's a hard and fast rule.
Indeed. Like every year when folding month starts and the Standford stats server decides to go on vacation lol.
Feel sorry for me I'm one of the people that will have to check the scores every night.
An unenviable task. You are appreciated for it though! (at least over here lol)
You know I wonder about people at times. There have been 142 people signup and about 15 people have managed to messup filling in 2 names.
Bwahaha. This is why spec signs me up.
I can't spell my name. π
@tired crow Thats only ten percent... people are improving
lol
I still like the one that filled in @GOTspectrum for his forum name. I'm still not sure that's not you trying to enter twice
Is this good for a PNY 2070 Super Mini? Reduced the power draw by 30W and the voltage from 1.05 to 0.975
Max frequency as reported by Afterburner is 1965Mhz even tho the manufacturer states 1770 but I believe this is due to GPU Boost 3.0
@tired crow We usually end up with about 100 people who just write either 223518 or LTT Folding Month for their name lmfao
@rich depot so your saying we're do good at the moment ?
hey I got a plug in for the event in the WAN show but it was in youtube chat
Nice, I'm at work today so couldn't stay up for WAN show
I need to get setup so I can send superchat
Hello, im just starting out Folding@Home with my laptop, i have a question about the cpu speed
Why does the cpu speed clocks down when its at 50% utilization? Is that something i must worry about?
@wooden sapphire it might be clocking down if it has a certain boost time
@rare blaze sounds like your cpu is burning
given that it's laptop that's more than likely.
laptops arent usualy best devices to run folding@home
Does it have removable battery @rare blaze ?
my dud, that isnt even my quote
derp
its @wooden sapphires
@wooden sapphire if your laptop has removable battery - take it off while running folding
the battery will die turbo fast in that kind of use.
helps with your heat issues too when it won't try to charge the battery constantly.
and runs on AC power straight up
@wooden sapphire it might be clocking down if it has a certain boost time
@rare blaze Min speed is 1.8GHZ
It clocks down to 1.68GHz
@wooden sapphire if your laptop has removable battery - take it off while running folding
@unborn hollow It doeant have removable feature, i just uninstalled it yesterday since
i've started folding now
Is an estimated of 17755 points per day on cpu and 477295 points per day on GPU acceptable?
IMO as long as you can finish any given WU within a day you're doing well
good good ^_^
i've started folding now
Is an estimated of 17755 points per day on cpu and 477295 points per day on GPU acceptable?
@proven ruin It depends on your hardware. For GPUs PPD check - https://folding.lar.systems/folding_data/gpu_ppd_overall
The upcoming releases should have some good improvements for the FAHCore
Fahcore_22 specifically for CUDA accel
I've been noticing that my 1080Ti is hovering around 85% usage
@hushed bobcat there are some weird WU's around at the moment, I've been getting some that are in the low 70% usage.
Does a 5400 RPM HDD and a 10 mbps connection significantly effectnfolding?
I got my hands on some other rigs
@proven ruin It depends on your hardware. For GPUs PPD check - https://folding.lar.systems/folding_data/gpu_ppd_overall
@tired crow True! uhmmm i7 7700 and a GTX 1060 6GB that's lost a % of perfomance due overclocking
Well to 1060 is doing a bit above average so that's good. I have no idea what a 7700 should be doing.
Eyyy thank you for the info ^~^
@stark bough It shouldn't really affect it too much, but you may need to adjust the Next Unit Percentage if you are seeing the slots sitting idle for too long while downloading a new WU
Anyone recently try folding on an eGPU?
Well if the connection has enough bandwidth I don't see why it would pose a problem
Is this channel working ? It keep vanishing from my list.
Is this channel muted?
I'm not sure what is going on but something is screwed up with this channel at the moment
Seems fine to me 
^
ive never noticed an issue
The issue I'm having is that it keeps vanishing from the list
is it possible to use 2 diff pcs for same acc simultaneously
I believe so
i am on a 820m and i3 4th gen ande i have another one which which i3 2nd gen
what is this?
@tranquil stirrup you can use as many PC's as you want, all you need to do is put the same user info in the client on setup. (Username, Passkey, team Number)
just got myself an i3-8100
wonder how well it will serve as a november spacehater though
during ryzen gen 2 they were pretty comparible
Is there a list somewhere of half slot cards or 1 slot that 'might' be useful in folding if sold at a decent used price?
what is folding?
what is folding?
@median badge the simulation of protein folds
Donβt we have an LTT video on the topic yet? Feel like we should link it
im using a rtx 1050 and the WU has an estimate time of 30 hours
is something wrong
@median badge https://youtu.be/KU4qOebhkfs
@wintry anvil sometimes it takes a while for the estimate to read correct, also the gtx 1050 is a pretty low power card.
yeah
keep in mind that WU's have different difficulties
3070
3080
3090
2060
2060 S
2070
2070 S
2080
2080 S
2080 Ti
Titan RTX
3080 is almost triple the Single Precision Floating Point performance vs the 2080
10 vs 29.8 TFLOPS
Well damn
Now I can have a killer PPD AND feel guilty about not playing more games on my video card
i dont know how to read that
the numbers are completely diffrent
one is gflops and one is tflops
hold on
okay it makes sense now
blame Wikipedia lool
I think we are going to see a lot of underutilised cards when folding first rolls out to then, there simply aren't many WUs large enough to use the full compute power. Oh how I wish we were able to run multiple WUs concurrently like BOINC allows
Wouldn't it be hard to evenly split up the power?
@hushed bobcat boinc projets manage it, and why would you have to do it evenly, you'd get more work done overall
I'd be concerned with a big difference
like 99% on 1 with 1% on the other
i'm not sure GPU architecture can multi-task properly like a multi-core CPU can
unless the BIONC projects your talking about are using GPU (the ones i've done are CPU only)
@hushed bobcat They are BOINC GPU tasks, I wouldn't be comparing them otherwise
anyone tried folding at home on 3090 yet?
I'd say most people that currently have a 3090 are under NDA
i want to see the 3080's PPD
OCC says 3.5M-4.5M but its still early
I want to see PPD/W
does folding degrade your hardware?
Heat and voltage kills hardware
Bearings wear out
But itβs mostly the dust
Keep it clean
Just wanted to say hello to all you fine folx! I joined the Discord server today, and joined the LTT folding team yesterday, and am sitting at ~1.6M points already. Not too bad for one GPU folding, I guess?
what gpu you got
ASUS GTX 1080Ti Turbo
I wonder if anyone has any advice?
I have a pair of EVGA GTX 980Ti SC+ cards sat in a drawer. Is there any benefit to me trying to underclock/undervolt them so I can use them for folding, too, or are they simply too old/slow to be bothering with?
(I can't justify 500W additional power draw... the 1080Ti is sucking down quite a bit of juice as it is π)
I'm also not bothering to fold with my current 24/7 CPUs, because for the extra power they'd draw the tiny PPD they'd generate doesn't make sense.
@dusky maple undervolting/power limiting any GPU will usually get you a better PPD/Watt
I understand it seeming like a waste to fold on CPU's, but keep in mind some WU still need to be done on a CPU
so it all helps
granted I would recommend using a modern CPU set to base clock
Of course. I stopped folding on my dual Opteron server a while back as it was drawing an additional ~150W when folding and generating a whopping 10K PPD...
my Ryzen 5 3600 does about 100k PPD and draws around 65-85W
I know it says CPU: 8 but really its using 10-11 threads
Indeed. Massive difference there! My own Ryzen 5 3600 isn't going to be folding, though, as general performance of the PC (this one, in fact) suffers too much.
only folding while idle should work
Once this CPU gets moved over to my Plex server, it'll be folding. For now, it's a desktop with multiple uses, some of which seem to be unhappy sharing with F@H.
looks like a 980 Ti should net you around 800-900k PPD each
VS 1080 Ti
it actually makes more sense to sell your 980 Ti's and buy 1 more powerful card thats a little newer
For example, selling your 980 Ti's and buying a 2060 Super (arguably the best PPD/Watt card currently on the market)
2060 Super
Interestingly, that chart appears to be based upon TDP ratings, and not actual electrical draw.
yes
Here's my current power draw, whilst folding on my 1080Ti:
That's total system power consumption... and the machine is doing a lot more than just folding.
actually i've found that pretty close
my 1080 Ti is an AORUS Extreme
it's MAX power draw is 375W
That's insane!
WAY higher than the 250W TDP on a founders 1080 Ti
but even when folding it doesn't pull more than 260W
and I power limited it to 80%
so now it draws around 200 W
also HWInfo64 should give you a card power draw
I haven't played with undervolting/power limiting at all on my 1080Ti... and my R5-3600 is happily sipping juice. π
Current info from HWMonitor
i saved that profile and have it set to boot with windows
if i want more power i just select another profile with that slider set to 150%
If the card I have was watercooled, I'd happily be boosting it. It's not, and that's simply because I didn't realise ASUS didn't use the reference design for this card before buying it.
Temps, noise and power draw have all dropped nicely. Thank you! Such a simple adjustment π
I've played with the "45W Eco" mode in the BIOS. That seemed to work very well.
(I'm not currently in eco mode)
Oh? I don't have Ryzen Master installed, in case that matters.
Ahh. Yeah... that old chestnut!
my main gaming PC is an i7-8700k
i have it OC'd to 5.0 GHz with and AVX offset of -2
Shiny
All that will do is prevent it boosting further, I suppose?
basically it disables Turbo Boost completely
3.6 GHz is the base clock for an 8700k
but you can do it from within Windows at any time
That's a pretty handy tip.
Thank you π
As you seem to know your stuff when it comes to hardware tuning... (and apologies if this is the wrong channel to ask)
Do you have any tips on how to enable the all-core turbo that the people buying the Chinese motherboards from AliExpress seem to all be doing? I have an ASUS Z10PE-D16 WS board in my old primary rig (that I haven't turned on in about four-five years), which currently has a pair of E5-2620v3 CPUs in it. I bought the board, RAM, CPUs, case (Obsidian 900D), etc. all brand new retail back in 2014... and barely used it because the triple SLI 980Ti setup I was running kicked out so much heat it was unbearable to sit in the room for more than an hour at a time.
Now that I've got a 1080Ti, I'd be interested in seeing just how hard I can push those little six core CPUs π
you know i have an E5-1620 V3 in a server, never could get it to boost on the Supermicro Mobo
also i have never really heard of this magic AliExpress mobo
is that the mobo Linus did a video on?
The Z10PE-D16 WS is one of a very select handful of boards for 2011-3 with overclocking intentionally in mind (ASUS' website mentions "up to 10% overclock", which seems mean when boards like that have such awesome VRM cooling)
Maybe? I've seen them modded on Craft Computing and a few other channels.
ohh dont do that
they are talking about Base clock overclocking
that's just a recipe for pain
BCLK is one thing, sure... but the mods people do to the BIOS images they install on the Chinese boards seem to be able to magically get all of the CPU's cores to run at full boost... not just a single core, but all cores.
sorry i have no knowledge on this
No worries π
i use xeon's for servers, i dont even think about OC'ing them
My first Xeon build at home was a pair of E5645 chips on an EVGA SR-2... so maybe I'm just nuts π
I still have the SR-2 board. I should probably sell it some time. I paid Β£500 to import it used from the US back in 2011-ish.
Nice! I have an X5675 sitting in an ASUS Rampage II Gene... that I've never used.
It would be
Back in the day, I tried folding for a bit on the SR-2 with two E5645s... which, when the CPUs were running at 3GHz+, wasn't too bad. The power draw was crazy for the PPD, though.
Sure, but at what cost?
you could hear those 9000 RPM fans across the house though
Is that the GamersNexus mod mat?
it's a modmat
didn't want to ruin the tablecloth
well the TDP of the 2660 is 95W
so probably 180-200W for 150k PPD
Sometimes I've used the box the board came in to do test builds... Bad idea for static, but never had an issue.
you see its slightly above the mod mat?
its sitting on the little boxes spark plugs come in.....
haha!
i ended up putting Noctua low noise adapters on the CPU' Fans
worked great and only went up like 8C
Nice!
Hmm... Just noticed my SSD is getting rather toasty. It's 58C currently. Should I be worried? It's a 1TB Samsung 970 Evo.
that is rather warm
That's what I thought!
I just used the ASUS AI Suite to tweak my fan profiles to force more air through the case.
Hopefully that'll help the SSD. It's already dropped 3C.
It's been fun chatting with you, but now I need to get some sleep. I've been awake for the last 22.5 hours...
π
Hopefully we will chat again soon π
Is this mostly a problem on my side or are some WUs just unlucky?
Sometimes good GPUs will fail random WUs. I'd be worried if it became regular.
2 in a row so far... π
"Why is my GPU so quiet? Oh"
well in case anyone was wondering what a 3090 can do
Hope it gets added to the LAR database
just give it time

@gaunt wasp i just noticed you had 11 attempts to start on that WU
if it's still not working could you post the log for that slot?
^^
Impressive
Whats PPD/W?
probably worse than upcoming 3070
n/m saw it. Page loading slowly for me
yeesh
any way to filter by power cost?
Been doing it manually. wow. the 2060 KO looks like the best value so far for points per Kwh
3090 seems likes it $0.31 so far but I'm trying to look for sub $350, and then sub $200 cards with best PPD un new/used markets.
@hushed bobcat I think the issue was the fact that I bumped the power strip switch & computer flicked off for a sec, causing a server/client desync issue for the houseβs IP. Iβve two computers folding & both were affected.
All good now, but Iβm giving the Mac Pro a bit of a break, as itβs been folding almost non-stop since April & while I donβt mind the fan noise (rear fan has no rubber mounts), itβs very old and I donβt wanna overwork the 13 year old fan motors.
a day off here & there depending on work unit time from here on out. Meanwhile I can task my Lenovo to fold starting tomorrow. π
Is the guy running LAR on this discord?
3090 seems likes it $0.31 so far but I'm trying to look for sub $350, and then sub $200 cards with best PPD un new/used markets.
@stark bough The best bang for the buck at the moment used would be the RTX 2060 super.
what are people upgrading from 2060S going for? probably 3070?
wondering if prices will drop after 3070 launch
@tired crow
Thanks for doing the research. now time to find out how much power is going to cost this winter during run
It's going to be a while before there is enough data on 3000 cards to figure out an upgrade path. It could turn out some 2000 cards are better.
In some circles i run in, some 1070/ti users are looking to go 3080
For now I'm going to stick with my 2x 2060 and 2x 2060 supers and wait to see what numbers the 3000's are doing.
If I had the money I'd grab 2 3080s and relace all my GPUs with them too
But with only working about 12 weeks this whole year so far not looking like I'll be getting very much
i'm still trying to replace 660 ti's
i think the cost per 1 mil points is like $7 lol
i want to replace everything with 2060 supers, but now i want to wait and see what the 3060 looks like
at the very least it should reduce the price of 2060S's
at least i hope
so 3090 so far best in folding?
4.5-5mil reported so far
Aaaa my 5700XT got 15mil so far π
oh yea if it matters, if my 3090 is the only one on there, its an Asus TUF 3090 running at default settings
non oc version
also im noticing some fluctuations so wait a bit to get more data
3090 looks best in folding, for now, but cost to deploy. Hooo boy
Imagine 3090s during age of mining
If anybody here has a rtx 3080 or 3090 and is getting crahes or an unstable exprince while running f@h plz notify me
i have no issues, however I have the TUF version which is apparently not having any crashing issues like the others
What's the power draw on the card while folding?
thats because Asus didn't cheapout on caps and TUF is one of cheapest cards at least here
does anyone know if you have to leave the LAR systems web client open all the time for it to send GPU PPD data?
Yes tab has to be open
π¦
"It can only log data when the extension is open as it's web based."
I just have it in it's own window open all the time in no sleep mode.
i should be able to let it sleep right?
No have it in NO sleep mode
No sleep in the tab so the extension continues collecting data
unfortunately one of my systems is Linux
i dont even know how to open the web client lol
Im not sure if he has the firefox extension live yet, but I made sure the chrome for linux one worked
Also don't forget to Disable Animations
yea i did that as soon as i found it
Anyone happen to know why F@H seems to like to completely drop WU's on logout/restart despite a) being nowhere near finished, b) F@H running as a non-user systemd service?
with my 3090 i generally leave power limit at 100%, it seems to use all avaliable power generally
i was thinking more of an actual power draw in watts
for example my 1080 Ti at 100% draws around 225W
TDP is 250W
yea
then the cost per 1 mil points is probably accurate
thats impressive
its 2x more efficient than my 1080 Ti then
granted my 1080 Ti is 2 generations behind now
o i think one big reason why it can maintain tdp is because of the boosting
like i remember on my 2080 super it was using like 80% of tdp on some workloads
while with 3090 it compensates by boosting much higher to compensate
some workloads it gets to lkke 1745 mhz, sometimes it goes above 1950
all at around tdp
ok rn its sitting at a very consistent 340w
yea 339-343 is what hwinfo is telling me
RIP AMD
Yup
Man, now I really want grab a 3070 if I can
Between this and Topaz, it's seriously tempting
This jump in points is nuts....
rocky launch aside, these cards to bring the ooomph
I meant the Cuda thing
im looking at that chart now
my 660 Ti's may match my P106-90's now....
well....
not counting the boost to the P106-90
1080 Ti nearly 30% boost
does anyone know when will this be implemented?
is it already implemented?
it just says update your graphics drivers
i just updated today to 456.38
also i dont know what "core22 0.0.13" is referring to
we've been on core22 for some time now
looks like 456.55 was released today
ill give this a few days to settle i guess
It's already implemented. Running my GTX680 right now, pumping 280 000ppd
Wish there was an easy to use benchmark to compare opencl to cuda performance
Yeah I was thinking βwait we werenβt optimized for CUDA already?β
wellp time to update some drivers
@thorny inlet do you need 456.55?
I installed 456.38, works fine too
whats update precedure again? is it just GFE?
I updated to 456.55 today and it's working just fine
Anyone have benchmarks yet before and after?\
it does shave off some time but it depends on the WU
out of conteXt PPD lol
lol
I should take all 7 days to make it
yup
I'm running 2x 2060 and 2x 2060 super and my PPD is up about 40%
But it's still early we will have to see how it all settles down
it says avg for that WU but i know my 660 Ti usually only does 80-100k PPD normally
its a good thing we can scrape infro from such a large pool
huh. think my app might actually report folding stats.
this is interesting
stats on a GTX 970
my 970 now
whats interesting is this is on a Linux system thats been up since April
with old drivers
linux driver 440.1
1080 Ti just broke 2 mil PPD
Now if I can get this working on mobile
Yeah, seems like im getting an extra 20-30K compared to what im used to
gold contributor here I cime.
watch them find a way to turn intel quicksync into a booster.
I got a bunch of 8th gen i3s sitting around
lol
8th gen i3 would do great in a NAS
not quite this level but still.
4790 is going into a NAS. i3 is a dedicated plex server
anyone watch last WAN show? someone else mentioned 12vDC as a possible low power build for LTT
You mean the new intel 12v standard?
Car battery straight to the motherboard π
lol
overevving increasses PPD at cost of MPG.
Mechanically driven PPD
PPD/MPG tables when
Actually if we wanted
I have a backup generator that I know the fuel burn rate
We could calculate points per gallon
...
I want to do that now
thats one way to fold
ehh yolo
@hushed bobcat No do this
My backup generator is 12.5 kW continuous, according to the service manual best fuel efficiency is at 80% load or about 10 kW
the 2060 Super draws 175W
Grab your 6-pack of LTT Exclusive Vanilla Cold Brew 40% off TODAY and enter for the chance to get caffeinated for FREE β use code βLINUSβ at checkout! https://t.co/KDeUVXULHa
279
57 of those is 9975 Watts
Calculate the PPD Per Madrinas.
shitt
Show off your inner techie in style with the LTT vacuum-insulated water bottle. Linus Tech Tips vacuum insulated water bottles areΒ intended to keep hot drinks hot for up to 8 hours, and cold drinks cold for up to 16 hours. Holds 21 fluid ounces Important safety notes:This prod...
LTT Water bottle is 21 oz.
how many LTTstore.com water bottles to get 1 mill points
PPD / Gallons = 6,234,375 Points Per Gallon of Diesel
over 9000
avg diesel price is $2.40 a gallon
LTT store should send a free water bottle full of diesel for folding month to every folder
π
i'm brain farting on calculating the cost per point
i took 6,234,375 / 2.40 and got 2,597,656.25
so does that mean you can get 2.6 mil points for $0.01?
that doesnt sound right
wait it would be 240
not 2.40
25,977 points per $0.01 of fuel spent
if my math is correct 1 mil points would cost $0.38
to much math
if you look at daily cost at 0.17 per kWH 57 of these would be $40.47 per day
my gen going 24h would be $46.08
thats actually much closer than i thought it would be
my math is probably wrong then lol
just let the ltt group test it out for you
we need some real world testing
someone send me 57 RTX 2060 Supers
and enough 8 GPU server chassis to put them in
π€£
ill provide the diesel
just put on mining rig frames
well the problem with those is you can usually only get 3 x8 slots
a threadripper CPU or enthusiast Intel CPU could get you 4 slots
but i was going for density
do you not know how to setup a mining rig?
are you talking about mining boards with lots of USB on the board?
and you use USB to PCiE adapters?
pcie risers
βοΈ
you can always attempt bifurcation.
from my testing you need a minimum of x4 PCIE 3.0
in theory you can also use eGPU
even then you will probably take a PPD hit
true
but i dont think you can daisy chain eGPU's
even if you could you are limited by thunderbolt bandwidth
pcie 4.0 x4 in theory can handle the bandwidth bottlenecks in loopback mode.
not over TB3
ahh yes
is there a thunderbolt 4 on the way?
i think so
usb 4.0 was greenlighted for pcie 3.0 x4 bandwidth, not pcie 4.0
what about those remote PCIE things linus used in his 8 Editors 1 workstation build?
think it was 6 editors using 8k
i suppose if you had a server board with 3 PCIE x16 slots
and all of them supported x4x4x4x4 Bifurcation
last one was 7 gamers 1 cpu
oh they were dual xeon
let me look at some diagrams of Supermicro Threadripper boards
maybe 1 has 4 slots
the nightmare was in usb connectivity iirc.
Check out Displate's metal posters at https://lmg.gg/displateltt
Sign up for Private Internet Access VPN at https://lmg.gg/pialinus1
We finally finish and test the 6 8K editors, 1 CPU build....
Buy Tyan Motherboards:
On their site: https://lmg.gg/8KVcq
Buy EKWB Watercoolin...
Here was the bifurcation issue
no way
if im reading that right every slot is full electrical
so the x8's have 8 lanes each
and every x16 has 16 lanes each
it looks like the PCIE 8x slots to share lanes to the CPU
but it seems like they have the proper lanes reserved
i'd have to really dig into the manual to see if every one supports Bifurcation
128 lanes baby
but if they did,
thats 24 graphics cards
each with PCIE 4.0 4x slot
you would need a 16c/32t EPYC at a minimum just to feed that many GPU's
that many graphics cards would draw 4,200 Watts
would probably need 300W for the CPU and motherboard
so 4,500-4,600 watts
you would need a 240V circuit for that
yeah boi use all those lanes
you would need a 240V circuit for that
@hushed bobcat
Just run your second PSU to another room or unplug the electric stove π
https://www.youtube.com/watch?v=FoW4vXnkhJw
powering that thing ia going to require a new breaker panel
Cut together scene from Jurassic Park showing only the breaker reset sequence, because that's what I see in my head when I reset my home's breaker panel.
Do I risk a WU failing if it was assigned when it had 2, then I paused it and gave it 6 simply cause I want the WU to be complete before I go to bed?
CPU
bah. 2 it remains then. 9 hours here we go.
i updated the drivers on my Linux folding machine
from 440.1 to 450.66
it DID NOT like it
RIP. rollback possible?
can you do that on Linux?
also my remote management isn't working now either
i think its time to put Windows on it again lol
DO you remote manage over internet or just LAN?
IF you can get linux setup it works great
been meaning to figure it out.
but when it breaks you basically have to start over
because i dont know how to fix it properly
i use VNC to manage it over LAN @stark bough
Linux box has X11Server and my main PC has RealVNC
So everything to one computer, and then VNC to that?
?
when i say manage i mean manage the desktop
so for updating/ rebooting/ checking temps
if you're talking about FAH then i just have the clients connected to the Advanced client on my main PC
sorry, yeah thats what i meant
how do you check your main PC when you're out. remote desktop software too?
well i don't check my main PC when im out
but if i wanted to i would remote into my network via my Private VPN
Yeah ive been meaning to set something up like that.
i have it set up so that when i log into my VPN it acts like im on my local LAN
or a site to site VPN.
Grandma doesn't pay electricity and usually has the heat on high in november, so...
nevermind lol
i think it found it's happy place
the CUDA improvements soo far have been very impressive
reminder that the official folding month starts on the 12th.
Read the details and everything here: https://linustechtips.com/main/topic/1240962-ltt-official-folding-month-2020/?do=getNewComment
LTT Folding Month 2020 Event Starts October, 12th (00:00 BST) Ends November 22nd (23:59 GMT) Sign up for the event here: Sign ups will close October, 9th, 00:00 BST https://docs.google.com/forms/d/1b3C_vMnxgzF1_3aSoY1fQXHIE-usPuxU72f4RlzaCWk This event will be eligible for F@H...
Oof. Reminds me. I need to repaste
My 1070 is now matching my v64 on P13426 at 1m ppd, TPF within a few seconds welp
Probs a stupid question, but BOINC is used for cluster servers, right?
I was getting 'headaches' from when my GPU was folding. trying ti find out why
coil whine?
oh dear. that's not an easy thing to fix
normally I run at 80% but looks like a software update reset my presets.
Much better after I dropped it to 80%
Even maintaining my clocks.
I only go +23 on GPU, no Mem boost.
Anyone had to tune their cards with this new update?
Nvidia or AMD?
I noticed a new adreniline edition came out but I rarely update my drivers unless I'm having a problem
i had to restart my pc after installing the latest nvidia driver just so i could put the OC back on my gpu
My nvidia control panel is broken after the driver update
thats happened to me before
Have you been able to fix it?
I tried a few things from a quick Google search but no dice
Im afraid that I'll have to DDU and install the drivers again
I had a few weird things happening after driver update but a reboot fixed everything
@tired crow wanted to say thank you for doing the data handling for the event, so thank you
I mean I think it's you that's doing something at least
It all helps
And that reminds me anyone here who has signed up for folding month please check your entry on the signup sheet and make sure you signed up correctly. We have a lot of entries the need corrections.
Yeah. im not on there.
Sodadog
Anyone else reset the OC on their MSI afterburner with latest update?
You can't enter the event folding for PCMR you have to fold for LTT team #223518
lol
Yes you are I think Spec signed you up before he signed himself up
Used you as a test subject
i had to restart my pc after installing the latest nvidia driver just so i could put the OC back on my gpu
@stark bough
ok im signed up for folding month
Ok. See myself now. For some reason CTRL+F didnβt see me in the page
@tired crow I didn't sign mobius up, that was denfi who I signed up.
@stark bough Just hit the sign up form, I'd rather end up with two entries than not have you on there man
@rich depot Thanks bud. Iβm on there. 139.
Couldnβt find it earlier with search
Weird, it didnt show up with a search for me either... We'll blame google!
His name is SodaDog
I just got confused
And he needs to claim his Silver folding badge
Ahh, I see, hes one of those people who have wildly different names across the internet
@stark bough Me too tbh
oh wow
im getting no WU's available for one of my folding boxes
haven't seen that in awhile
I just got back into it now that temps are dropping, and i was gonna say the same. CPU is idling
yea 8 thread CPU slot idling for me
@rich depot Bbbjoo
Folding, I don't get you, 6 months ago I got this all set up, and my GTS 450 was no longer pulling tasks, get annoyed, gnash my teeth, then hot season happened, that's all done, power my old folding rig thinking i'd find another use for it, bam, it's pulling tasks again like nothing changed
lol
wait is this ppd boost from the cuda update
@lapis gulch Prob
@fickle needle Welcome back into the fold!
i'm here for a good time, not a long time
Just passed 100 million points!
https://folding.extremeoverclocking.com/user_summary.php?s=&u=835318
F@H stats user summary for Mister_Rogers. Help Folding at Home fight Coronavirus, further medical research, and prevent diseases with distributed computing!
nice
Woah. I'm rank 467 in the team now. I used to be like rank 110 or something. RIP me I guess
dam
if you start running another 1660 you'll get the same PPD as my 2070
it sounds like it
how much power does the card pull
i might get a job so i can get a few of them cards
i think around 200ish watts. i have it under a heavy oc
+100 on the core and +800 on the mem
that should give you a better idea of what my gpu is doing
hwinfo doesn't refresh very quickly
my card is water cooled to so it helps, but i can get the card up to 2.1 ghz on the core if i want to but it start to become unstable
i ran it with the stock air cooler for about a year and it did decent
thats the funny part i did run it under load without the pump running and it stayed coolish
but i do watch the temps tho
https://us.msi.com/Graphics-card/GeForce-RTX-2070-GAMING-Z-8G thats the card i have
thats enough to run the card
hmm
one 8 pin and a 6 pin
i have an rmi
its used but its been working well
the psu is the hardest part to pick out
yup
but i did build my first pc when bitcoin was very popular so my gpu was very expensive tho it was a 1060 6g. so a psu is a bit better in that terms
i mine every now and then
the worst card i have for folding is an amd r9 270x
Oof I got one of those
Was actually thinking of seeing if I can bring it back to life
hmm
this is interesting
@brittle coral you were the one looking at PCIE bus bandwith usage for graphics cards right?
yea back a while ago
where can I find up to date list of hardware PPD? I am trying to find how much PPD those new nvidia cards (RTX 3080/3090) get
am interested to see how much does the doubled cuda core amount boost the PPD
so the doubled cuda amount does something. Thanks
@hushed bobcat not feeling great today and reading comprehension is at 0, did y'all have a question about PCIe bandwidth?
Not a question
I'm folding on some p106-90's
I noticed the bus usage is 80%+
Yep
who made it
GPUz confirmed it
is it some china shiz?
Zotac
lmao zotac probably cheaped the fuck out because those cards were never meant for folding
and crypto is fine on a 1x
get data into card, card crunches hashes, hashes spit back out. very low bandwidth and ram usage
zotac being the fucking cheapskates they are probably lowered the spec
i will remember this zotac
I looked up the specs for the card and it says PCIE 3.0
Let me get a screenshot
if it were only scaling down to like gen2 or something i would think maybe it's something fucky with your hardware elsewhere cause you';re using an old rig, but you know better than to make that mistake
here is the card
every idea i can think of for why that might be doesn't make sense specifically because its fucking 1st gen PCIe its reverting to, and nobody's used that for ages
i mean for mining they do get away with plugging them in via USB
that sounds about right
tbh
i wonder if bus useage stats in afterburner are not entirely useful or if it's not reporting correctly and actually is at gen3
i know there's a point where a gen3 4x will bottleneck a 2080 super
maybe a 2070
i havent tested
they badly bottleneck a 2080ti
i added a P106-90 to my gaming system with the 1080 Ti
even at x8 the 1080 Ti barely needs 20% bus usage
the system that the 2 P016's and 1 660 Ti are in has been giving me fits all morning
freeze ups
BSOD
BSOD on boot....
getting stuck in the BIOS
the only word i can think of for this setup
"Monstrosity"
at that it is
huh? 3 GPU and only 750K PPD?
Yep
that looks oddly low
ooh. fair enough
Before the CUDA update the 660 Ti would only do 80-100k PPD
cuda update?
how do I know if I have that?
my RTX 2070 does 1,8M-2M PPD
Just need 7.6.13 client and the latest nVidia drivers
and where I can find my client version?
My 1080Ti was around 1-1.5 mil, now it's 1.8-2.4 mil PPD
Should have an "About" section in the advanced client
Or look at system information in the advanced client
date apr 27 2020
Time to set power target even lower
Or take a card out of the picture
Do I have to keep f@h client tab open for this addon to track my PPD and workers?
The lar system dark client?
aye
Set power plan to max or equivalent
Yes
Hmm, https://stats.foldingathome.org/os looks weird.
Used to be like 4.4 ExaFLOPS a week ago
now it's less than 10% of that, something in their statistics isn't adding up
Do you guys use software to underclock your folding gpus? I pay for my electricity so I kinda experimented a little and I found that limiting my gpu to 60% power, it still does 90% of the work, it's a good compromise between power and efficiency.
Also, a lot less noisy.
I have used MSI AB and set the power target between 60 to 80%.
Yep I used MSI AB too
But lately I've been kinda lazy and I just let the card blast the packages at maxed power limit with oc
Mine is in the bedroom soo.. the lower noise level kinda helps me sleep π
Ah, looks like they had a mismatch in their gpu counts and estimates.
Itβs always fun to watch the number of CPUs and GPUs running Folding@home! Earlier this year was particularly exciting as we saw explosive growth in Folding@home in response to ourβ¦
Propably purged the old stats and started again
i fold at 80% power limit
Same. undervolt?
I can hear coil whine at 100% max folding
strangely doesnt bother me when gaming though.
games don't really tax the gpu to the same extent
which is why i was suprised to hear it when i first started really folding in depth
is the 3080 on folding.lar.systems around here?
All my GPU's are at 80%
i think my laptop is folding under my old folding name even tho i have it on my new name
bro this new ryzen folded like 6x the points that my old dual core did in a single day lmaoooo
and its way more power efficient. plus it can heat up my room in the winter
also i do wonder something. my rtx 2060 seems to only be 30 percent utilized on full folding power with the web client? why would that be?
task manager says it -w-
shit imma hit a billion points this week sometime
im thinking of stealing my 1060 out of the nas to replace the 270x
too hot too fold right now :/
So uhh, how about that CUDA support aye
Not bad for a 1080Ti
IT'S WORKING! π
@vernal lodge task manager isn't very good at accurate monitoring, especially the GPU.
if you want to see actual usage you need to change one of the charts to either CUDA or Compute_0
oh
Mine does 2.25m at 80% fairly regularly
Is folding cheaper than running the space heater I wonder
^^
im thinking of stealing my 1060 out of the nas to replace the 270x
@timber grove Why did you put a 1060 in a NAS?
is it woth it getting folding 1060s for fah?
i haven't pulled it out yet
o boi. My RTX2070 pulls like 200-220W when folding, and it wants more it seems. Throttled by power limit
@shut halo that is way too high a 2070 should only pull a max of 175w
not when its on an oc
that's tru
Where are they rated at 225 ?#
top is around 175 I think
But which spec sheet ? spec I've seen say 175
but the card can pull over 200 watts
thats about what my card is doing
my card is a Asus Strix 2070.
super is 215
I could get higher efficiency if I turned the OC down
1.209M PPD looks oddly low in that sheet
with current project that is
If I let this run a whole day I would end up around 1,8M PPD probably
there it says 1,6M PPD for 2070. That's more accurate
The data is real time and at the moment is changing fast becasue of cuda that is why there is a note on the sheet
Once everything has seteled down I'll update the sheet
AiB's can go WAY over factory TDP on cards
this is factory
my 1080 Ti
375W
the only way i've been able to hit 375 is with a synthetic benchmark or Furmark
when folding i have it limited to 80%
or about 200W
that gives me 1.8-2.2m PPD
250 is the spec for a 1080Ti
I have an Xbox one laying around in my room is it possible to get it to fold?
