#folding-and-boinc
1 messages ยท Page 69 of 1
yikes
thats pretty common for me lol
also those spikes @onyx canopy was getting
i have them too
seems normal to me
i haven't timed them but i have my save interval set to 3 minutes
they seem to happen fairly consistently though throughout work units and on all GPUs
aight
bloom what's your hardware that you're only getting 7k for a WU?
did you sign up for an account to get quick return bonus?
or is it just really old/slow
someone tell them to feed my rx580 a WU. my room is gonna get cold xD
im gaming now on the 5700xt for a bit but it's getting cool out and i'm to lazy to get up and reach behind the case to turn off the box fan that blows through it xD
or rather, through the window at the back of it
anyone know if this would help boost F@H
http://squirrelsresearch.com/acorn-cle-101/
it's a fpga that connects via nvme that takes processing load off of the gpu
nvme uses x4 pcie lanes
and you can add multiples if you have additional nvme slots
didnt noticed that in gaming while folding
welp fotgot to pause it
but frames are still good
oof strooke
wish someone could test this out and see if it helps folding
nope not really no one has tested it with folding
its a fpga that was used for crypto mining it basically takes processing load off the gpu
but im sure it can be done as well for folding
it boosts gpu performance
it also says it doesn't work in combination with a high end gpu
yea the idea i think was to use lower end GPUs in a small mining rig and offload some intensive things to this, it's probably just doing a bunch of FP calculations or something, but a high end GPU will be faster is my guess
the real question is, can it run in tandem with a GPU to power a single slot, or does it require it's own input if it were mapped to F@H?
depends on the code i'd guess more than anything
sounds like you have to program for it
i tried to get a bitstream from hashaltcoin but they said it wouldnt be worth it
=/
can I purge the work queue?
its against the TOS running F@H to delete work that's already been sent with the pure intent of just deleting it
so if you have work already downloaded, you gotta just pause it lol
I cant get it to download is the problem
its on its 16th attempt at finding a WU
@brittle coral
I wonder if the well is dry
ooh got it
and yea the well has been having supply line issues as it were xD
not enough clouds = no rain baby
sorry that was a really foggy pun
i'll show myself out
I've had to delete 2 WU in 3 days
I keep getting ones that cause my PC to hard lock
I got rid of the advanced flag and updated my gfx drivers
ive had to delete a few for various reasons but not nearly as many as when i was dealing with those nvidia driver issues that caused all non primary GPUs to spit WUs right back to the server as bad lol
good idea, the advanced flag is only for people who can monitor the rig, or those who don't have stability issues running them so they can see what configs worked well
if you try with the new drivers it may also be more stable
I went from 441 to 445
the only one I intentionally deleted ever was one that was 15 days completion time, when i knew i was reinstalling windows the next day anyway so i reinstalled F@H that night so i could at least contribute some completed WUs
idk if the 20 day expiry window is even up on that WU yet lmao
It was perfectly fine during the folding event
word
yea 445 is what fixed my issues as well
and AMD i haven't had any issues with the last 3 months worth of drivers
since they update more often
I still can't get the Kubuntu F@H client to use a GPU
@storm talon is probably asleep now but is the one to ask
I already did
for sure
He only does CPU folding
thats right
i am very tired i saw that earlier, sry for @ing you larry lol
is it the same in ubuntu?
or debian or some other distro?
Kubuntu is Ubuntu with a different skin
idk these things so well xD
but for sure good to know
i installed it on a rig with bad hardware and never got very far setting it up so idk m8
im thinking on my next rig I may dual partition the SSD and fill it up with OS (using a 120gb 840 evo lol) and see how it runs, but it's also the newest rig and i may just want to stick windows on it again anyway
It could be the card is too old
what card is it
GT430
yea probably
The latest driver is 390 lol
i know they don't support anything older than the 7970 generation on AMD side
i'd assume nvidia isn't QUITE the same as long as that old of a card supports the most recent version of CUDA, but idk how the CUDA client works or if it's even used still since I'm pretty sure everything i've run so far was openCL even on the nvidia cards
I don't think it has Cuda
and older cards may not support 1.2 CL
oof
yea
if it doesn't support OpenCL 1.2 or any version of CUDA thats probably why
According to Wikipedia it supports 1.2
But
Doesn't have double precision floating point
I think anything older than Kepler is a no go
I can pull a GT 730 out of one of my windows folding rigs
yea it'd be worth it to test at least
i ment for the GT730 to finish its current WU
oh christ xD thats still pretty slow
wow this is a 46 hour WU
the GT730 can do about 8k PPD
and this is a 15396 base credit WU
what WU
11741
lol word
i dont have any of those rn
ac tually of my GPUs
only one is running =/.
=/
hah that isnt a real emote here
=\
xD
okay
uh lets see
im at 71% on this WU on my GTX1070
1hr 26m left
it's worth 28800 base, getting estimated 151000 after QRB
lol my p106-100 is at 17c idle
980ti is idling at 19c
take the pc inside or take the outside out of the room lol
its in the garage
im monitoring it with VNC
xD
there's a 1200cfm blower aimed at it and a 250cfm 120mm fan also rigged into the case between the hard drive cage and it to direct airflow further
the blower is basically aimed right down the airflow channels on the fins of the 980ti and the miner dgaf its soaking it all up for days
that miner folds at like 45c load
oh sick the 980ti got a wu
such oxytocin receptor wow
both 1070 and 980ti on 13878
I really need to figure out why I only get 7k- 12k / wu
I just had two GPU WU of 120k+, only just noticed it whilst waiting for a CPU WU.
Really hit and miss with the WUs though.
waiting for the credit to drop on these 2. lol
generally only 1.5 - 2hrs roughly.
Only 1080tis?
yeah, just testing at the moment. showing 4.2mil PPD and still have 1 GPU idle.
I should still be able to get 60k+ for a Wu with my 2070 :/
I know the 1080tis are faster
But 7k is too low haha
yeah, like 80 - 100k at least I'd have thought.
You have a really decent system tho
The best of 10 series eventho 4 way sli isn't the best
It's still the highest raw power you can get
PPD dropped to 3.7mil. still 1 GPU idle.
running all on an ASUS X99 WS.
was hoping to experiment and throw another 2x 1080Ti's in there, but I need to find my dual PSU cable and run them that way.
Board supports 7x GPUs at 16x/8x/8x/8x/8x/8x/8x bandwidth. Awesome board, much like the X299 SAGE I have here too. Tonnes of bandwidth
woohoo. got 4th GPU WU. lol. just crunching these for a stability test then it's bedtime.
Pulling a little over 1000w from the wall at 90% power level.
this is the highest i've read out so far between both clients lol. stoked to get a 3rd up i might be able to hit that 3.7 you're getting xD
the one on top could be higher but i never end up with those maximum output WUs on every slot at once usually
that's pretty good though, every bit helps.
I'm tempted to turn on the X299 and the other 2x 1080Ti's there, but I get abused for chewing up tonnes of electricity for too long (no discounts for fighting Coronavirus with electricity around $0.16 USD per kWh unfortunately).
yea and i have a 3rd rig coming
i have an idle 290x for sure
and a p106-100 to test
there's another rx570 local but i might try risers on that board and build a frame around it and order a bunch of used p106-100 if they fold well in a 1x lane
@silver wing That is definitely intriguing to me, I'm always fascinated in seeing how manufacturers put more onto less
my FAHClient seems to have some issues lol
was literally running at 100% on CPU and 0% GPu despite CPU being removed from the slot list
weird. Did you save changes and restart the client?
no I removed all the slots and added GPU again
@odd barn @brittle coral this is why I only had base credit :P
buggar.. get that passkey happening. I got mine again really quick this time (last time I tried it didn't even email out)
I haven't hit 10 wu yet, just trying to work out how many I've tested on this evening.
@onyx canopy what CPU and what GPU(s)
nvidia GPUs use a whole core to themselves to run it seems (or at least a whole thread)
good shit
was a FAH bug
Nvidia GPUs eat a thread each.
just before I @ you
was 8 threads at 100% for a single gpu 
thats my nvidia rig lol
idk why OCCt is using so much CPU rn xD it usually uses less
1060 1070 980ti ?
why do you own these
i bought the 1070 and was loaned a pair of p106, friend owns a 980ti so its on loan too
If cpu slot is set to use -1 threads FAH should reserve on thread per Nvidia GPU.
to be clear, same guy sold me the 1070 and the loan on the 2 1060
just to run F@H so he didnt have to set up a rig himself
cpu slot is disabled tho
yea that CPU doesnt have time for it
less than 6kppd after the 3 GPUs
it can do 34kppd on its own
not much
isnt vega 64 rly good for folding ?
@smoky rose any idea if that is whenever the program is running (ie, including GPU idle), or whenever the GPU is folding?
one of my friends is getting pretty decent performance on his vega64 yea
let me ask him again
raw Tflops is the best
he may be afk/asleep but i asked
I haven't tried running cpu slot only.
weird, my old 5930K is still at estimated 26k PPD (CPU only)
and that's running all 4 GPUs too
bloom is it all double that matters?
if so my 5700xt should be a lot higher than it is lol
The cpu threads are used to feed the gpus work. It doesn't get there automagically.
lol, yeah, I know that.. just surprised my dinosaur is pulling 26k PPD whilst feeding 4x nvidia GPUs when @brittle coral mentioned down to 6k PPD.
its a 6/12
that sounds about right if you're feeding 4 GPUs and have 2 cores left over plus whatever time you can get out of the other threads on each core
even at only 3.6ghz vs 4ghz on mine those 2 cores and 8 threads extra makes a huge difference
but aren't you running an 8350K?
I was thinking AVX had something do to with it as your clock speed would be smashing mine, 2 extra cores or not.
https://ark.intel.com/content/www/us/en/ark/compare.html?productIds=126689,82931
Might be the extra memory channels though..
Intelยฎ product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.
mine is pegged at 3.5GHz.. might be the low profile HSF is holding the CPU back from boosting. lol. Ooops. I had 3.9GHz on my old 4930K on air (Noctua D14-2011) but need to revisit the bios settings for the 5930K and 9900X to see what I've done.
Nooooo... my rig just died... had 30min to go.
PSU not lighting up the mobo (it's an EVGA 1600 P2).
NOOOOO
FFFFFFFFFFFF
i was about to say
i ordered a water block for my 3800x to get the turbo to run higher
its only boosting to 4.1
mid 80s temps
stock cooler
did you see the pic of the thiccboi rad i have a pair of from when i had one per 290x along with an MCR320 and a dangerden 2x120 radiator as well?
the cpu was on the loop too but it didn't really get cooled much by it lol
temps were good but hotter than i wanted
some days i legit felt like it was heating the cpu with the GPUs because at idle it would have load temps
and load temps were only a few degrees hotter than idle
lol
tried unplugging the PSU, no deal.
tried turning it off and on (at the back), no deal.
tried a different power point, no deal.
Not getting any mobo lights whatsoever. Might have to swap the P2 with the T2 to see if it works...
yea test that
that will determine instantly if it's mobo or PSU if you know the other PSU is good
before you do tho
got any wire?
Think the PSU has just up and died though
or a paperclip
boost the PSU without it plugged into anything else, pin 4 (green wire) to any ground
some bad PSUs will spin up and then the second they have load put on them they will fail to POST internally and shut off but it'll help determine point of failure at least if the other PSU also spins up, could also be it died and took stuff with it tho who knows
Just got told off by missus. lol. looks like the PSU is a tomorrow job now ๐ฆ
there goes that stress test.
thanks. Busy day tomorrow so won't be for another 18hrs till I can get to it.
Looks like it'll be an RMA job.
that sucks... probably gonna lose all those points too. buggar.
anyway. catch you guys later. Happy folding.
yesterday just added my NH-D14
can't fold at turbo else it thermal throttles with 1 fan but works very cool like the 120mm rad at base speed folding
dangit Evo X kit for being too tall
the gap widens. but thats because we all do less points than a few days ago
I really hope some people at the folding team can get this sorted and back up to bigger assignment numbers
A few days to get things back on track after easter should do it.
oh man, my new water loop is going to look amazing
i looking at automotive parts for hose clamps
Or did we lose that many after the folding event ended?
i found red anodized strain relief clamps size AN-10 (5/8" OD) for my primochill red 5/8" OD tubing ๐
gonna get that cpu going so fast
the red should add like 100mhz right?
@smoky rose nah the assign rate is down and people still complain about not getting wu's. LTT has prob gone down after the event but still enough
@smoky rose Outer Diameter
@brittle coral add some flame stickers... they make everything run faster its a scientific fact!
Just look at these guys.
Flames makes everything better.
As long as the rig it self is not on fire that is.
More GPU work units are coming for @foldingathome. We've had to shift some of our effort from setting up projects to moving data off servers to make room for more data. Its amazing how quickly the data is coming in! Something like 6TB/hr.
i decided to go digging
im gonna just use my old parts
i found ap ile of like $300 worth of fittings i forgot i had
and 2 tubes of shin-etsu TIM
and some other things
Don't forget the flames.
fah1.eastus.cloudapp.azure.com is sitting there with 102' A7 jobs. ...and no disk space.
yeah, for at least a couple of days
I think the bottleneck in this case is FAHs storage.
We've had to shift some of our effort from setting up projects to moving data off servers to make room for more data.
Their internal storage.
come on LTT, petabyte server
@drGregBowman @foldingathome Hey @LinusTech. It looks like FAH have a need. A need for, not necessarily speed but, a bunch of Storinators. I propose this:
Industry leaders provide the hardware.
You build 'em.
@drGregBowman run 'em.
The rest of us fold 'em.
guys what say you
should I install BOINC and start running rosetta as well for when the F@H client isnt getting work?
been a long time lol
this is the main issue tho
More GPU work units are coming for @foldingathome. We've had to shift some of our effort from setting up projects to moving data off servers to make room for more data. Its amazing how quickly the data is coming in! Something like 6TB/hr.
they are getting 6tb an hour of data xD
bah wrong link
that was supposed to go to another discord
"a bunch of storinators"
I didn't specify how many. ๐
hahaha
yea i saw that
with 6tb an hour of data that's no surprise they ran into issues
i wanna know something
why does russia suck at contributing to this despite having so many gamers
i get china cause of the great firewall
russia you slacking
facepalm
People seriously believe 5G is spreading SARS-CoV-2.
They are carrying out sabotage operations against the infrastructure.
yup
I'm going to spend the next hours reporting every single post in a couple of Swedish groups to get Facebook to flag them as false.
i saw a video on an anarchist site i frequent because they're local and my friends are on there sometimes doing activist stuff, and someone posted a video of some jarhead from another group tossing a molotov on one Y_Y
they were putting the dude on blast, to be clear
the anarchist community has been really helpful in san francisco's china town though protecting people and businesses, as well as a group of veterans that's been organizing there
there's some crossover in both groups lol
i wonder what cool stuff they're learning from all the work we've done that's gonna help them speed things along
Facebook should simply do a "DELETE FROM posts WHERE '5g' IN text"
why it generates clicks, clicks generates ad impressions, ad impressions pay their server bills ...
even with the server issues it seems like im still averaging 2.4-2.6k per day right now lol
based on stat sites
i wish my birthday were closer, facebook always asks if i want to donate my birthday to a cause and this year imma send it to the F@H team me thinks xD
man they really need to optimize some of these units better
the 14549 unit runs so slow on my AMD gear xD
and im looking in metrics and im seeing 50-71% gpu utilization
@queen aurora The users of tinfoil are already abandoning the platform. They cant stand the "censorship". ๐
600kppd on the 5700xt
when ive seen it peak out at 930k =/
ive had my turbo clocks wide open all night because the card is running super super cool cause of it xD
65c core 92c junction peak xD
@smoky rose yeah but they wont stop existing so they just shift from platform to platform ๐
anarchist community
something about those two words together doesn't seem right
@queen aurora True. But as long as I don't have to see them I'm happy.
I talked to one of them a while ago. He claimed that "any and all wireless traffic on 10+ GHz band is lethal even in the smallest of doses". That is kind of funny as the globe is well covered by thousand of satellites providing satellite tv on the Ku band.
What was the Wi-Fi Linus tested that could be stopped by the cardboard box the router came in? 50GHz right?
Pretty sure everyone that did the testing is still alive
Unless... Linus is a zombie....
That would explain all the dropping
Combining cherry picked facts do not pass peer revie...oh, never mind. They don't care about that sort of thing.
@hushed bobcat it would explain a lot #lienus
On April 9th, we hosted a Fireside chat with some members of the Folding@home team, including lead developer Joseph Coffland (actual chat start at 6:30), F@h head of communications Anton Thynell, and F@h team member Sukrit Singh.
We discussed the architecture of Folding@ho...
Soon there will be fpga powered folding ๐
?
To bad the LTT staff doesn't pay attention to their own discord
lol it's against the rules to tag staff
but it would be nice to share this info over to a member of LMG staff like Luke or Linus
How big of a deal is fpga folding and why should the lmg staff know about this?
Numbers?
its in the works but 1 fpga would take the place of over 50 high end gpus
im getting it together as we speak but thats info for LMG staff
Why should they know though?
content ๐ and this is very valuable information
I'm calling that they will ignore the vague bs
@unique sigil you there?
Rule 13 bro
he is active ๐
Just make sure you are folding for teamLTT
@verbal kraken What cores do you have support for?
i already talked to Jake about it ๐
plan on getting a bitstream that will work with all hashaltcoin fpga miners
How is that going to help with folding/boinc?
Bitstream -> fpga executable binary
if the works was available a single fpga would produce about as much work as 30 high end gaming rigs
which miners? I've got a bunch of antminers sitting around doing nothing
antminers are asic not fpga
yea
this would be like F1+ F1 Mini etc
And it depends on the fpga as well
but as Jake said it wouldnt help to much right now as the work just isn't there
if the works was available a single fpga would produce about as much work as 30 high end gaming rigs
@verbal kraken
It was 50 gpu's not that long ago. like 30 mins
Stanford needs to pick up there game to handing out work
@rich depot if you knew about how fpgas worked you would understand
I know. I just think you are way to vague and change details way to fast to be credible
@verbal kraken are you developing this fpga bitstream?
i'm working with hashaltcoin to get one created
Hi everyone, I am lookong to upgrade my rig for folding. I am thinking of switching to a Ryzen 9 cpu for its extra cores, so i can work while folding. What do you guys think is the best cheap am4 board. I am looking for something with good performance, power delivery (for OC) and RGB (I usually try to stay with asus ROG). What do you guys think of the rog b450-f gaming? I worry that it is too much of a budget board, as it is only 120$
@verbal kraken gotchya
and do you guy shave any experience with the asus crosshair motherboards? I've never used them
Thanks!
@oblique rivet I have a 3900 and I can have it use all cores and still do basic things I need. LIke searching the internet, email, video etc.
I am using a Asus Rog Strix X-470-F Gaming mobo
I have used Asus mobo's for most of my computer needs.
If you have any questions let me know. I wouldn't have any issues answering them if I can.
@brittle coral finally getting full credits ! I needed to complete 10 wus first :P
@forest sapphire Thank you! Does that mobo need a bios update before using the matisse cpu?
is there a big difference between x470 and b450?
@oblique rivet it didnโt actually. And I am glad lol cause I didnโt think about that till after I had all the components. And I didnโt have an CPU I could use lol.
I am not quite sure the difference between the mobos. I know I didnโt want the 570 as I didnโt need the PCI 4.0.
Ah. is there genearlly a difference between x-series and b-series? I've always used lga115x so i don't know too much about am4
I might actually just keep my current motherboard and go with an i9 9900k. Do you think that will work as well for folding as the 3900?
b = cheaper chipset than x iirc
worse audio - ethernet - oc
fewer pcie lanes sometimes
Less overclocking headroom
9900K should be find as well.
But I do feel that AMD will be better spot if you wan to be able to multitask a bit more
That extra 4 cores, and AMD is known to be a more well rounded CPU for all the things, compared to Intel and the single threaded advantage they have
but even then it isn't that much
The 3900 is great for gaming, folding, and everything in between.
Anybody know why I am getting Next Attempt: Unknown??
Well nothing like a pause, unpause to get the CPU to kick in
Ah ok. I will probably go with the intel, just bc i already have an intel chipset
understandable.
don't really want to replace my motherboard tbh
@oblique rivet I did Intel for a good 5-6 years. Before that I was AMD. I went to Intel cause of the gaming, but now that I don't game nearly as much and I run VM's. It was nice to have that extra core count
ah
Which is why I went AMD
Thanks for all of the help!!!
Hey anytime!
YEah like I am running my CPU at full right now
total CPU usage is about 90%, which gives me the 10% overhead to do other things
like I don't notice any delay or slowness when moving and opening apps and what not
Has codeman gone... is it safe to come out?
Id look at individual core utilization imho.
For gaming imfind that i can spin off 4 cores for background folding.
Need to verify but i think i lost 11 frames when i was discord streaming though
I can game with no issues with it using full CPU and GPU I have found. I get about half the frames I normally would.
anyone know if discord streaming is primarily cpu?
I'm glad F@H has started with these issues after the event... would prefer them not to have the issues at all but I understand why
I have a 3900x and 2080TI. And I am basing my frames on BF5 at Ultra settings
@stark bough I think it is CPU
But I don't quite know
If i had a 3900x i would hace spun ifff an entire ccx for it :p
But I think you can change that to do GPU more
Yeah I have thought about spinning my CPU to use like only 8 for folding
but I say USE it all at once for those big ones lol
That's what it is using right now on average for using all cores to fold
And that is at about 4.1Ghz across all cores
oh my
what?
that is a lot of processing... i'm using the "light" option ;c
oh haha. Yeah I let my CPU just crank out
yeah, like 80 - 100k at least I'd have thought.
@golden fulcrum 78k for 1h20
I think I had 95k for my previous wu but it was longer
Man my CPU is getting like back to back WU
but my GPU is showing Next Attempt: Unknown
I can't get it to check in
Wondering if this is normal or if someone could help me out...?
My GPUs were folding but have stopped since yesterday as shown in this screenshot:
CPU is still folding but the GPUs have been sitting like this and not folding. Tried searching for some info but didn't find too much. Is this normal, or what's happening? I'm a complete noob with f@h.
Even after restarting the application and the PC, same thing...
Running the latest drivers.
Both cards running stock.
go to logs
also can you select one of your gpus to see what it says about the status ?
for one of your gpus
then it was downloading
How long does it download?
it depends on the WU, the servers
Been showing "download" for both since yesterday.
in the log window, you can see the %age of the download
if it doesn't move, then there a problem between you and the server
Showing a lot of "failed to get assignment" above what you see there.
I guess you can't control it
btw folding at home is going to evolve
they're hiring
so we should see software improvement, website improvement, visualization of the protein etc
but only in the 4 next months
also they're getting new servers
Ah so it's likely an issue on their end?
yeah
They're probably getting swamped right now, lol
those messages are perfectly normal
also I wouldn't recommend cpu unless you have a really good one because it takes way longer to fold / you get lower points
they've been getting swamped for the last 3 weeks
bloom is correct however there are only jobs that CPU's can do
The CPU seems to be going most of the time, so I'll let it run.
so if you can afford it and its not running too hot let the CPU work away
Thanks for the help. ๐
with my 4790k
your 4790k is better than an 8350
i run mine at 4.4
actually on older CPU's i find GPU's run a little faster with 2 threads per
I haven't tried because I don't fold on cpu anyway
my 4790k machine has 2 GPU's in it and when I set the thread count to 4 in F@H my GPU's had a small boost in PPD
have you tried 5 ?
cant do 5
F@H wont do prime number thread counts
it wont spit out an error per se
but if you look at the logs it will say its reducing the thread count
You can enter any number from 1 to the number of threads your CPU supports.
If you have GPUs, F@H reserves one CPU per GPU to feed it data across the PCIE bus.
F@H has difficulty with large primes and their multiples number of CPUs.
7 is always large, 5 is sometimes large, and 3 is never large. Try to choose a number that is a multiple of 2 and/or 3.
2, 3, 4, 6, 8, 9, 12, 16, 18, 24, 27, etc. are good numbers of CPUs to choose.
5. 10. 15, 20 etc may work most of the time. Other numbers will bite you.```
Is that high enough thread count for it to throw errors instead of the warning about reducing number of threads?
it said 29 was a prime number so it reduced to 28
i did have an interesting bit in the logs
saw it was on a WU but the PPD was super low (like 1/3 normal) according to the logs it reduced the thread count from 30 to 12 for that specific WU
didnt do that for the next one
wish i had a screenshot of that
Don't you rotate your logs? ๐
F@H blue screened the PC that monitors all my clients 3 times in 2 days
all my logs rolled away
Darnit
But what about heating?
at $0.13 USD / kWh, that's about $33 USD isn't it?
A municipal politician in Sweden
"Corona can be a biological weapon. Just search for biological weapons on google and you will get a lot of results."
That's it. I'm going to bed.
@golden fulcrum which means +8$ in the bill
25$ + 30.7% = 33$
does the estimated credits go lower and lower as time passes by ?
@onyx canopy +$33 USD to the bill.
Wrt estimated credits, it can fluctuate depending on how the WU is being processed.
On another note, busy day today.. hopefully troubleshooting the PSU later this arvo. Fingers crossed.
why +33, 251 is total
which means 191 last month
oh you don't pay forfaits ? like same amount each month then at the end of the year you pay extra
@smoky rose probably 20-40% of people think this was engineered and weaponized.
I prefer the other explanation:
โThis wasnโt part of a plan at all. Like a volcano or earthquake or CME, statistics and chance just rolled a big one. Itโs comforting to have a specific group to blame. Itโs not so comforting when you canโt blame anyoneโ
@onyx canopy good point, sorry I misread the screenshot. That's $33 total, or roughly $10 more than usual.
Had 3 power outages in 3 days. Inverted detected a fault to grid tie and shut down production during several days.
RIP folding
With any luck, I'll have my cluster booting for the first time in the next hour. First run is only 4 boards, if everything works, I'll start to build the infrastructure to add the other 24 of them lol.
Eli5 please
I have 28 Mini ITX motherboards with Atom n2600 CPUs and 2/4 gb of ram. I'm trying to make them in to a 'supercomputer' with clustered OS (Pelican Linux)
2 core, 4 thread @ 1.6ghz with no AVX lol
It's not really worth the power, but it's more an exercise in boredom and science lol.
All 28 running at full load will draw about 500 watts
its very low for this amount of cpus
Nope, all were free.. saved from the scrap bin at work
totaly worth it then
The TDP is 11-14 watts each lol
Scale indicators. They're last generation, and obsolete, so they were getting scrapped out for the newer boards
They ran Windows 8 IOT with custom software
Work perfect in that application lol
are you going to fold ?
If I can get it running, yeah.. I want to see if the entire cluster can break 20k ppd lol
I'm guessing that's about what it will do
I think frequency x thread is what matters the most
compared to "single core performance"
Step 1 is to get it working on 4 boards, beta test and debug
Yeah, but without AVX, it falls back to MMX instructions lol
Which is why the estimate is so low
it will even struggle to run the vm haha
56 Cores, 112 threads total if I can get it all running lol..
key word is if
2x 28c56t xeons 1.6ghz
The way Pelican distributes the workload should be interesting. Imagine a threadripper with a 1gbps infinity fabric LOL
It's gonna be SLOOOOOW
but fun
well distributing the load across that many cpus will be REALLY hard
can you specicy the atom model ?
I've got a nice 3com gigabit managed switch with 48 ports... Each board has a realtek (lol) gigabit nic
holy shit
the CPU's are soldered on...
poor switch
Can't swap them lol
whats their model ?
N2600
when you said previous gen
but if you hook them all together getting them to work as "one" on a single task seems like more work than its worth
Pelican handles all the "heavy lifting' as far as splitting the workload. It just 'works'
Even has XFCE GUI
The GUI presents like a single CPU with X amount of cores
make sure this chipset is even supported by the os lol
It is
ohhhhhhhh im not familiar with pelican, i figured you had to do it manually
sounds fun
Exactly lol
Not even sure I can get F@H to play nice on it.
We'll find out as soon as my ISOs are done burning and I configure a persistent storage drive on the main host.
well
if it makes the program think its a single CPU it shouldnt be an issue
but if F@H realizes its not a single cpu it might throw a fit
shouldnt' Lol
This is uncharted territory for me... So we'll see
I've done some HPC with raspberry pi's before for shits and giggles, first time trying on x86/64
pretty sure it will try to assign each cpu a wu lol
assuming it isnt tricked into thinking its a single large cpu
the purpose is this trick tho

otherwise its useless haha
Yeah, I'm not sure one of these boards by themselves could finish a WU in the allotted time lol.
you would have more WUs to do but they would take longer
arent you allowed like a week per WU?
1.6ghz dual core on MMX... Might take longer than that lol.
shhhh small details
They recommend a P4 as a minimum, those were clocked a lot higher, and can barely squeak it out in a week.
Maybe the 2c/4t might make it equal out, but who knows lol
yikes
Anyway, back to the dungeon.. I'll update in a while when I get closer or fail... Lol
good luck
you will have the equivalent of a 2011 xeon 9282 lol (released in q2'19)
400w tdp for 14nm
avx on
oh yeah . 3.5mil PPD
does it output display on all of them or only one of them?
What motherboard is that?
Technically all will have some video output, but the nodes just have a warning not to touch anything lol
Jetway NF9c-2600
Failsafe mode seems to be working
weird
Most likely a driver issue
does it only use one of them as the "host" or are they all a collective "host"
One is host, rest are slave nodes
so the hosts drivers are the only ones that matter then
Yes
that makes it infinitely easier
If I have any more problems, I'll spin up a "modern" laptop as host
i love how you put "modern" in quotes
I have too many computers. Over 120 at last count
My newest laptop is only a i7-3000 series
good god
Yeah same tbh, I got a 4th gen i7 precision laptop just for having desktop power on the mobile
i have a laptop with a dead battery and broken wifi card
then i got a i5-5300u ultrabook for free, all i had to do was buy and replace the motherboard for $40
Yeah, my 3rd gen was free. Works perfectly except for the battery
Closer shot of these boards
Pci slot, but pcie m.2 lol
the i7 laptop has 32gb of ram a 256gb ssd and a 1tb hdd
the other is i5-5300u 8gb of ram and a 500gb sshd
i use it for Microsoft office and internet usage and some light programming its a ultrabook so its just a laptop i use for when i dont need alot of PC horsepower
So while failsafe boots, it goes straight to GUI and skips the setup script LOL
Guess I'll login and see if it runs a GUI script
That's something...
SUPERCOMPUTER
4 * 2c
Lol
looks like
you're loosing speed
its not scaling correctly
is it ?
the heck
you dont have the same speeds per node
you should check the bios
of each board
Probably because of throttling
thermal
still, check the bios
if the settings arent the same
it's not good for the project
Three boards have subpar chunks of aluminum instead of heatsinks. Need more thermal compound
try to get sensor info
Vs
No idea, not easy to see lol...
Not viewable in bios
And Linux needs packages
I'll get there
My main rig averages 1.2-1.3m
So, got the client and control installed. The client won't start. Gonna troubleshoot tomorrow. Bed time lol
Just made it to over 1.1m points ๐
@smoky rose Didn't know they made the fireside chat video public, nice. Some good info in there, nothing too detailed but enough to explain where the pain points actually are
yea im watching it now myself
also jo is probably sleeping as its 4:24am there
@silver wing any idea who I'd ask about why AMD's performance is so much lower than expected based on floating point precision numbers?
i assume it's just software optimization level stuff, curious mostly
wikipedia says my 5700xt should in theory be faster than an RTX titan in double unboosted, and another 20% faster with boost above that
instead im seeing performance similar to on average but recently worse on average to a gtx1070
14549 seems very poorly optimized
almost a 40% drop from previous peak and 20-30% from expected at least
1454* in general and some of the 1455* units that came next
this is right now =/
From what I've observed F@H actually cares most about Single precision not Double Precision
peak is over 900k, avg should be at least 700-800k
for sure
well single is still supposed to be really high
5700xt single is around a gtx 1080
well
in that wu i just showed above
the 980ti gets double performance, same with the 1070, vs whats shown
or like 40% more
thats the kind of performance the GTX1060 is getting
=/
that says 5600
not a 5700
according to this the 5700 XT should be around 1.5-1.6 mil
which would put it around a GTX 1080
its not got enough space
and yea that sheet only has one person though
that sheet does not have enough data
this is true
also my RX580 is having the same problem
its not just the 5700xt
according to wikipedia the RTX2080 has the same single performance as an unboosted 5700xt
=/
i assume you checked and each openCL is actually referring to the correct card?
the 2080super the same as the boost
it is
both cards are active
heating up ect
under load
on both of my Dual GPU folding rigs it got the openCL backwards
the card isnt hitting full load either
50-80% typical
the units that were getting up to 900kppd were hitting 90% utilization
and ive never seen higher
its underperforming the 980ti and 1070 for days now and im wondering if there's something wrong but the only other person i know with a 5700xt hasnt had time to do a lot of testing
they shut F@H down the night before and then when they got hteir code to get QRB they had to keep working on the same unit from 24hrs before =/
rx580 =/
they have been exclusively getting these WUs for like 4 or 5 days now
while the nvidia cards get plenty of other ones that i know are better optimized for AMD and run well on nvidia as well
i did lol it was fine
you can see here GPU 0 is actually CUDA Device 1/OpenCL Device 1
yea def looks right
yea we both went over this before xD
when i was having trouble getting nvidia cards to work
the only other thing i can think of is a BUS limitation
but a modern Ryzen platform should have no issues with 2 cards
so its WAY more bandwidth than it needs
it has it afaik it just doesn't need it at all
xD
lets people run multi-gpu configs more easily basically
which im happy with them doing even if it raises the cost per card a few bucks since i use that feature
lol
rx580 dgaf either
if you run and NVME SSD it can cut one of your slots to 4x
but thats usually the 3rd slot
but im pretty sure it has a dedicated 4x
one is from the chipset one is from the cpu
this is the gigabyte AORUS x570 board, the one at about $200
i also have an X570 UD coming which does have 3 slots
but only one M2
they seem to trade you a 2nd m2 slot for another PCI-e due to board real estate
but i dont need another for that rig lol
i think MSI afterburner can show BUS usage
its probably not your issue
but it wont hurt to check
yea i want to look at that just for fun now
let me install afterburner on this PC, its only on the nvidia one
was using it to monitor the p106-100 temps because it isnt showing them in OCCT since it isn't a GPU
afterburner didnt at first either but then i set it to kernel mode
and now it sees all
@_@
but yea, the guy who has the 5700xt besides me that i know is a programmer
he got it because of the high paper stats
he wants to learn to code for it better xD
is upset that the entire coding environment is optimized for a company that is monopolistic and has just as many of it's own hardware side issues as the other guy, they just get a lot more money because of long term market share and marketing (historically)
i've dealt with their internal culture a few times tho and it's kinda toxic lol
used to personally know one of the lead marketing ppl at nvidia, they were our neighbors till they sued us frivolously and lost, so we moved and they one by one turned on everyone else on the cul de sac xD
had to move themselves, out of a home that had been in the family since they were kids
and at the time this was all going on she was busy trying to astroturf over AMD's massive success in compute with the 5870
with huge nVidia marketing campaigns
@hushed bobcat have you watched the fireside?
they are talking about want to move away from proprietary environments as well ๐
(cuda)
they didnt say it outright but in general they want the whole thing open source except the point system
point system needs to not be cheatable lol
fireside?
the one you were talking about i thought
no idea what you're talking about
nvbm
bad memory
it weas someone else earlier, let me link you it was posted by jo while they were still awake
On April 9th, we hosted a Fireside chat with some members of the Folding@home team, including lead developer Joseph Coffland (actual chat start at 6:30), F@h head of communications Anton Thynell, and F@h team member Sukrit Singh, and Jรถrgen from a communications firm BE that i...
the bit about open sourcing is around 1:03-1:04
@brittle coral I want to try out an OG Titan
you were talking about double floating point
practically nothing beats the OG Titan in Double Precision floating point
supposedly the RTX titan is slower than the 5700xt in double
let me look at the og
1300-1500
so about 3x without boost
either way its a ton
but that chart i linked earlier said the OG titan was only good for about 200k ppd
BUT only 1 report
no way its slower than a 780 Ti
actually
the 780ti in the wiki right above what i pasted above
is faster
not by much but it is
ah yea
but you were talking about in PPD
and single is what matters there
apparently
well thats just a guess on my part based on my observation
thats why i want to buy an OG Titan to test myself
yea but if a GTX titan is slower than a 780ti in single and you test it then we'll know for sure XD
thats the most obvious way
im betting you it is though
even with just 1 result
that result is REALLY low
it would be the fastest card on the network if it were double
im really glad i got a pile of nvidia cards tho for the rig out there instead of buying AMD stuff after seeing the optimization issues that seem to affect some WUs on AMD tho xD
especially since that rig is dedicated
and this one i still game on the 5700xt
i leave my CPU and the RX580 folding still of course
60-80fps in witcher 3 ๐
same as without folding
AMD systems are def better optimized for cpu useage at least, i wonder why nvidia needs a whole thread to itself
or, could be that the whole thread to itself is what's keeping the nvidia card up
its got to be the way the CPU streams data to the CUDA cores
the only thing i can think of
you look at the BUS usage?
oh no i forgot
one sec
let me close witcher 3 it was launching, bout to finish a WU on the 5700xt
man i remember when MSI wasn't even in the performance/OC space and it was just rivatuner and you didn't get most of this cool shit
@silver wing I didn't expect them too. It was a nice surpise.
@brittle coral Triss or Yennefer?
Back to bed. Ta ta.
You're dead to me. ๐
xD
i mean
i tried to do both in the first playthrough
but apparently they dont like that
You got that right. Yen is not a fan of you getting on with her friend.
lmao
also to be fair
the main deciding factor in picking triss was that she wasn't shacked up with the tankie authoritarians xD
@hushed bobcat where do i find the bus stream
we got 9 minutes before i run out of work on this slot and cant check the bus useage xD
i think MSI afterburner can show BUS usage
@hushed bobcat
GPU-z does too.
lol, probably 'cause you've either gotta select it in the options, or scroll down through all the little graphs to find it.
I just use GPU-z as it's heaps easier to find
yea im still not seeing it in gpuz either lol i dont see it in options either
look under sensors tab
yea im there
its idle now
just finished the wu
will have to test the bus useage later but at least i can figure out where the setting is now lol
just change one of the drop-downs. lol.
those drop downs don't change what the sensor is monitoring just the graphing
I know for Nvidia it's about 1/2 to 2/3rd way down on sensors page. Team red perhaps check options
maybe AMD doens't have a bus readout
yea
its nvidia vs amd again
no bus monitoring on AMD side
my afterburner on the nvidia rig has it
found it immediately
hold on
interesting, the iGPU on my 7600T doesn't have that readout either
Hmm
in afterburner
there's a value called "GPU FB usage, %"
thats only showing up for the 5700xt
oh haha
okay FB is frame buffer
so nvm thats not it
buggar. ok, I'm outta ideas. lol. just assumed team red had a similar reading for PCIe bus usage as what team green has in GPU-z
yea its weird lol
but apparently i can look at the frame buffer useage
idk if i can on nvidia, didnt look
using it a ton now cause witcher 3 is running again
yea they do as well
Pretty sure my 1080 Ti has FB usage in Afterburner
yea the entire set of them do
the rx580 isnt showing it but that may be because it's not primary anyway, but in afterburner on the other rig a least 2 of the GPUs had FB useage listed, i thin the 106-100 was the one that didnt
wish i had this running earlier
