|
just installed the newest version of f@h, and it wont let me add my 1080ti gpu in a slot. F@h version 7.6.9. It lets me add it to the list and it shows up under the cpu slot, and then when I hit save it flickers and doesn't keep the gpu in the second slot. I go back in and it only shows the cpu.
|
# ? Apr 21, 2020 03:49 |
|
|
# ? Mar 29, 2024 11:08 |
|
Similar problem here, I just installed f@h on my PC (total newbie here) and it will only recognize my CPU, but not my 1060 GPU. Wondering if this is linked to the latest GeForce drivers?
|
# ? Apr 21, 2020 11:18 |
|
Chikimiki posted:Similar problem here, I just installed f@h on my PC (total newbie here) and it will only recognize my CPU, but not my 1060 GPU. Wondering if this is linked to the latest GeForce drivers? Fixed it but forgot to comment here; I updated to the newest geforce driver and checked "clean install" in the installer. Then I reinstalled f@home and it detected it right away. Apparently this can happen if there are residual config files left by f@h when you reinstall it (I was moving the data directory off my system SSD) jonathan fucked around with this message at 16:22 on Apr 21, 2020 |
# ? Apr 21, 2020 16:16 |
|
OpenPandemics beta WUs are running. Names start with 'BETA_OPN1' Glad it's moving forward so rapidly, but none for me yet
|
# ? Apr 22, 2020 03:32 |
|
jonathan posted:Fixed it but forgot to comment here; I updated to the newest geforce driver and checked "clean install" in the installer. Then I reinstalled f@home and it detected it right away. Thanks, it worked! I uninstalled f@h, did a clean reinstall of the geforce drivers, reinstalled f@h (with the "recommended" parameters) and it works like a charm now!
|
# ? Apr 22, 2020 09:36 |
|
I never removed SETI from my projects list since they said they would just stop sending out work units, but got a huge new dump of stuff in my BOINC queue. I assume its just sending out units that expired from other users not reporting in time, but hey, more ALIENS. edit: it looks like its only GPU workunits, nothing CPU based. Lobsterboy fucked around with this message at 15:16 on Apr 22, 2020 |
# ? Apr 22, 2020 15:14 |
|
WCG OPN update: * Beta is paused while they look into some issues with initial workunits * Autodock has never been compiled for Android before; it may or may not be included initially when beta ends * They didn't size WUs for Android devices, and some took >24h to run * They are trying to support Linux on ARM as well, but no guarantees there yet, either
|
# ? Apr 23, 2020 15:08 |
|
Is anybody else not getting COVID-19 work assigned to them in Rosetta@Home? In Rosetta, how you tell is if a workunit has COVID-19 in its task name. Maybe it's just me, but it seems some projects are either done or are finishing up their current work. TN-Grid seems to be on the final stretch of what they're currently doing for COVID-19. The current relevant batch of workunits there start with a prefix going from 165735_Hs to 166977_Hs, and they seem to have been assigning these in order. I'm currently seeing a 166813_Hs on my own list of tasks. In Folding news, the latest Folding@home client now gives you an option to select COVID-19 as a cause preference. This means you can now tell it to prefer downloading COVID-19 work over work for their other projects, like cancer. This won't actually guarantee Folding will be churning COVID-19 work, however.
|
# ? Apr 24, 2020 02:18 |
|
Completed Rosetta WUs matching 'COVID' in my joblogs: - Last 168h: 108 - Last 48h: 9 - Last 24 h: 1 Each of those includes the WUs from the periods after it (so th 48h number includes the 1 completed in the past day). Definitely a fall-off. I'm sure there's people talking about it on their forums, if you want more direct answers.
|
# ? Apr 24, 2020 05:47 |
|
New project targeting CoViD-19: https://www.csic.es/es/actualidad-del-csic/el-csic-e-ibercivis-lanzan-un-proyecto-de-ciencia-ciudadana-que-busca-farmacos https://boinc.ibercivis.es/ It's in Spanish but they're looking for existing antiviral drugs that could be used to treat CoViD. Not many users yet but I gather that the team (or at least the institution) has run BOINC projects in the past. Looks like there are no work units available at the moment but having literally just launched, that's not really surprising - there are about 25k tasks currently in progress though.
|
# ? Apr 25, 2020 03:12 |
|
Tuxide posted:TN-Grid seems to be on the final stretch of what they're currently doing for COVID-19. The current relevant batch of workunits there start with a prefix going from 165735_Hs to 166977_Hs, and they seem to have been assigning these in order. I'm currently seeing a 166813_Hs on my own list of tasks. That's only about 50% of the way there (~55% based on the highest task in my queue, 867). I do like the transparency so that we can measure progress, though. edit: actually you're right, it's >90%, I am blind and misread the lower number Discussion Quorum fucked around with this message at 04:23 on Apr 25, 2020 |
# ? Apr 25, 2020 03:25 |
|
Discussion Quorum posted:New project targeting CoViD-19: I went and made a goon team for it. Website's kind've broken, but it doesn't look like it'll take much to get on the very top. https://boinc.ibercivis.es/ibercivis/team_display.php?teamid=41 It looks like it's CPU-only, and they only target x86/x86_x64 architectures, meaning no Raspberry Pi (yet).
|
# ? Apr 25, 2020 06:45 |
It wouldn't surprise me if I've had some Rosetta units drop off at some point, those fuckers are huge and I seem to always be getting complaints from BOINC about needing more space. I've allocated 5 gigs and over 3 and a half alone are the 3 Rosetta wu I have running. Fake edit: yeah, I've failed 3 and turned in 3.
|
|
# ? Apr 25, 2020 07:12 |
|
Looks like there's OPN beta WUs flowing on WCG again. I got 7, across 2 machines, in the past 48h. Timings look like this on one nodecode:
code:
mdxi fucked around with this message at 08:07 on Apr 25, 2020 |
# ? Apr 25, 2020 07:59 |
|
vodkat posted:Does anyone have any recommendations or posts elsewhere about energy/price efficient ways to setup a little bionic farm? I really like the idea of having a few machines linked together running bionic, ideally some sort of cheapish single board computer that won't be a massive energy drain. Obviously there seem to be quite a lot of people doing this with raspberry pi's but I was wondering if there are other options out in the flood of single board computers, maybe some with gpu's optimised for machine learning or whatever? As I often turn out to be, I might have been wrong in dismissing SBCs on a price/performance basis. Obviously if your budget is $50 for hardware and ~$0 for power, a Pi is the way to go, but I assumed that by any measure a modern CPU would blow it away. Well, I started noodling around on the idea of "used mid-spec computer vs used lovely cheap corporate box" and made this. Lots of assumptions that could be dumb in there, but based on said assumptions the Pi 3 and especially Pi 4 could win handily on total cost/performance. Based on this and my lack of space for building a dedicated crunchbox at the moment, I may supplement my Pi 3's with a couple more recent SBCs. The new Odroid C-4 looks like it may be good for this if it truly delivers Pi 4 performance without the heat issues. The flip side of course is that a Pi is still low-throughput, and a Pi farm that matches the output of any reasonably powerful PC would be a lot of space/work/etc. But if you want to donate $5-$15/mo in total value on CPU projects and don't mind a bit of computer janitoring, it looks like it's actually not a waste of money. On another note, it looks like a on a per-core basis a Google Cloud Compute n1 vCPU can do gene@home tasks ~4-5x as fast as a Pi 3B+, meaning each vCPU is ~1-1.25 Pis at ~$18-$25/mo per vCPU. Not a good deal if you're looking for another way to do small-dollar donations without a lot of heat or physical computer touching (but there is a $200 free trial credit...).
|
# ? Apr 26, 2020 04:15 |
|
I've got a nice milestone coming up. The Big Round Number will be crossed in tomorrow's processing, but the actual count probably won't be particularly interesting. Total WCG WUs returned as of tonight's update: 499,665!
|
# ? Apr 26, 2020 06:01 |
|
Discussion Quorum posted:Based on this and my lack of space for building a dedicated crunchbox at the moment, I may supplement my Pi 3's with a couple more recent SBCs. The new Odroid C-4 looks like it may be good for this if it truly delivers Pi 4 performance without the heat issues. It would be expensive to set up equipment-wise, but I would be curious to see how a Turing Pi with seven Compute Module 3+'s can do in comparison. EDIT: Seeing 166920_Hs on my TN-Grid task list, so it looks like we're on the home stretch. As a reminder, the end should be 166977_Hs, unless they put out more COVID-19 work. Also, I am getting COVID-19 workunits from Ibercivis, so we can add that to the list of projects BOINC goons fighting COVID-19 should subscribe to. EDIT2: I know the OP hasn't been updated yet, but I've been making modifications to my post a couple pages back with updates that I think should go on the OP. Also, there's a really good effortpost on the GBS COVID-19 thread by forum poster Cacafuego on the process on making drugs and vaccines, and in silico testing's role in it. Tuxide fucked around with this message at 09:58 on Apr 26, 2020 |
# ? Apr 26, 2020 08:53 |
|
I didn't have time to do much more than skim the Ibercivis thing previously, but here is a translation of the first couple paragraphs for some more context:quote:Encontrar un fármaco utilizado en el tratamiento de otras enfermedades virales que actúe contra el coronavirus 2 (SARS-CoV-2) es el propósito del proyecto de ciencia ciudadana COVID-PHYM, impulsado por el Consejo Superior de Investigaciones Científicas (CSIC) y la Fundación Ibercivis. Puesto que algunos medicamentos en uso ya han demostrado ser suficientemente seguros para la salud humana, podrían estar disponibles para tratar a pacientes con COVID-19 mucho antes que un compuesto de nueva creación y, por tanto, acelerar el control de la pandemia. quote:The proposal of the citizen science project COVID-PHYM is to find pharmaceuticals used in the treatment of other illnesses that are effective against SARS-CoV-2, led by the Superior Council of Scientific Investigations (CSIC) and the Ibercivis Foundation. Since some medications in use have already been demonstrated to be sufficiently safe for human health, they could be available to treat patients with COVID-19 much sooner than a newly created compound, therefore accelerating the control of the pandemic. * I'm not a biologist and had to guess at the RNA stuff with the help of Google, could be off there
|
# ? Apr 26, 2020 13:51 |
|
Discussion Quorum posted:On another note, it looks like a on a per-core basis a Google Cloud Compute n1 vCPU can do gene@home tasks ~4-5x as fast as a Pi 3B+, meaning each vCPU is ~1-1.25 Pis at ~$18-$25/mo per vCPU. Not a good deal if you're looking for another way to do small-dollar donations without a lot of heat or physical computer touching (but there is a $200 free trial credit...). I did a little bit of experimentation with running BOINC on a 1 vCPU Linode slice. The results I got were really low. Low-end VPSes and cloud computing instances are massively oversubscribed as a matter of course, because the vast majority of tenants have bursty usage patterns. Your "1" CPU isn't going to deliver anywhere near the performance/runtime you'd get out of 1 physical core. Also, when you're the one running an active process 24/7 you are the noisy neighbor that is making things worse for everyone else on your host. It's just not the model that (cheap) utility computing was designed for. I'd be happy to hear from other people who got better results, but from what I saw, the ROI is very much in favor of physical hardware. Also, I looked at your spreadsheet. I couldn't tell if you were amortizing hardware costs across some span of time, or if you were estimating that a Ryzen 3600 would use over $50 of electricity per month. At least where I live (western-ish US), it's definitely not the case that power and cooling are that expensive. I'm have 4 3900Xs, 2 2700s, a 1650S, a 750 Ti crunching 24/7, plus everything else in my apartment, and my power bill hasn't topped $200/mo even in the hottest part of summer.
|
# ? Apr 26, 2020 20:14 |
|
mdxi posted:lso, I looked at your spreadsheet. I couldn't tell if you were amortizing hardware costs across some span of time, or if you were estimating that a Ryzen 3600 would use over $50 of electricity per month Total cost per month is: depreciation + (electric cost per kWh) * watts * (hours/mo) * (1 + cooling factor)/1000 Depreciation is over 36 months assuming most things have some residual value that can be recaptured (eBay or re-use on an upgraded system, etc.). Electric cost is my incremental cost per kWh. Watts is an estimate from the Outervision PSU calculator. Cooling factor (currently 120%) assumes that for every "watt" of heat generated, the A/C will have to expend 1.2 watts to cool the room back down. That's a SWAG but it seems like it would have to be >1. The cloud instance in this case is a 2-core "high CPU", not one of the shared core instances. Looks like it is hard-throttled to a certain performance threshold since BOINC measured it at exactly 1 billion ops/s across both projects it's running. It's still far more expensive than comparable hardware, though.
|
# ? Apr 26, 2020 21:10 |
|
This week in numbers: In Asteroids@Home, we've gone up to #193 in Total credit, even though we're still focusing on COVID-19 stuff. In Folding@home, we've gone down two ranks and are now #51. Folding's compute power has now risen to 2.62 exaFLOPS. In Rosetta@Home, I've just gotten some new COVID-19 workunits from Rosetta@Home, so COVID-19 stuff is still going on there apparently. Goons are now back in #62 in Recent average credit, and we have risen to #203 in Total credit. In TN-Grid, some people have reported that they've gotten workunits outside the 165735_Hs to 166977_Hs range, so maybe their COVID-19-relevant work will come to an end around the start of May, I don't know! Our Recent average credit ranking is now #15, and our Total credit ranking has risen to #76. Good job! In Ibercivis, the goons are now #19 in both Recent average credit and Total credit. In other words, we're already in the Top 20 on both lists. This is a smaller project that just got started and there's no Gridcoin here yet, so climbing this should be easy (or would be if they supported ARM Linux).
|
# ? Apr 28, 2020 03:27 |
|
TN-Grid staff commented that there are still CoViD-19 related work units coming through, but they are now mixed in with non-CoViD units. They specifically gave ENPEP and TMPRSS2 as examples although I don't have any of those currently. On Rosetta, over 3/4 of my tasks are currently tagged COVID-19 so it certainly seems like the pace has picked up. I was running 100% TN-Grid but am now winding down that queue in favor of Rosetta, so goons may slip a bit. Sorry Apparently Ibercivis is using the default BOINC credit mechanism which is bad and has some people annoyed? IDK, I'd rather they prioritize science (and generating more WUs, I haven't had any in a while) over developing a better credit mechanism but up-going numbers are important to a lot of people.
|
# ? Apr 28, 2020 18:05 |
|
So my folding was going fine, but today my GPU decided to stay in ready-mode indefintely. Advanced control says it's waiting for a download: Any idea what's wrong here? /Edit: nevermind it fixed itself Chikimiki fucked around with this message at 18:29 on Apr 30, 2020 |
# ? Apr 30, 2020 15:53 |
|
Sometimes F@H just doesn't have a work unit to hand out. They've had a big increase in users processing WUs since COVID-19 hit.
|
# ? Apr 30, 2020 18:48 |
|
Ugggh just found out the hard way that the F@H docker container stores the work and settings in it's image storage, not the persistent volumes so updating blew away a few days worth of reserved work and a bunch of half done stuff.
|
# ? May 1, 2020 17:49 |
|
Is anyone else getting hardware fever? Or at least hardware info fever? I want to plan a round of upgrades for later (likely much later) this year, but neither AMD or Nvidia are releasing any info on what's coming next for their product stacks. The other day I found myself thinking about upgrading my low-end nodes to 3950Xs, which is just crazy talk with the 4000s on the (indefinite) horizon. I had beaten my hardware addiction before grid computing pulled me back in. Now it's all "in the service of mankind", so I can justify almost anything. Thanks, science. Since the 4000s are being built on basically the same process as the 3000s, I'm assuming they'll have similar power curves. I'm looking forward to playing with PPT settings and seeing what kind of sweet spot I can find on the compute/watt curve. Especially since WCG has said that OpenPandemics will likely be getting GPU support in the future -- so this time around I'd like to have a low-end GPU in every node.
|
# ? May 3, 2020 17:04 |
|
mdxi posted:Is anyone else getting hardware fever? Or at least hardware info fever? I want to plan a round of upgrades for later (likely much later) this year, but neither AMD or Nvidia are releasing any info on what's coming next for their product stacks. Yep, same. I keep reminding myself that grid computing is intended to be using the idle cycles on hardware I already own. Upgrading for the sake of make number go up faster is just an expensive form of masturbation.
|
# ? May 3, 2020 18:28 |
|
SamDabbers posted:Upgrading for the sake of make number go up faster is just an expensive form of masturbation. I don't view it quite that harshly on the whole -- especially since my wife, who can usually be relied upon to be a more reasonable person than I am, opined that it was a good use of time and money compared to what a lot of guys in my age range (I'm a prime candidate for a mid-life crisis) spend money on. A lot of people spend more on ski gear for a season than I do on compute hardware in a year, and then there's the guys who snap and buy Porsche convertibles and/or boats. I try to remember that and keep the weird impulses (like replacing CPUs with 3950s when I'm already planning to rebuild those machines in 6 months) in check. I admit that line of thinking was almost entirely fuelled by wanting to make the numbers go up, with a little bit of "god I need to do something" on the side. I'm currently trying to channel that feeling into something that doesn't involve spending money.
|
# ? May 3, 2020 19:26 |
|
I talked myself out of building a farm of ODroid C4s on the basis that I didn't really have a way to make a case. Then I got the email that my local makerspace has come up with a way to reopen. A farm of efficient SBCs would be better for my situation than another PC puking out fan noise and hot air, and I'm holding off for Zen3/Ampere before I upgrade. And I have some extra acrylic laying around. I'm bored and want a project but there are probably more productive things I could do right now. We shall see. e: card is fine, I am dumb Discussion Quorum fucked around with this message at 02:51 on May 4, 2020 |
# ? May 3, 2020 19:52 |
|
It's my World Compute Grid quasquicentennial! With last night's stats update my personal team passed 125 years of CPU time Which is nice, but really I just wanted to use the word "quasquicentennial", because how often do you have an excuse to do that? Basically never. Discussion Quorum posted:e: card is fine, I am dumb , kinda, I guess?
|
# ? May 4, 2020 15:56 |
|
More or less. I was going to hold off on upgrading anything until Zen4/Ampere, then get the midrange CPU and whatever the most cost-effective 4k60 GPU is, be it a 2070 Super or 3060 or 3070 or whatever. The result should be a computer that will last me a very long time. So I didn't really need an upgrade but having something nicer to sell would have reduced the net cost mdxi posted:... with a little bit of "god I need to do something" on the side. I'm currently trying to channel that feeling into something that doesn't involve spending money. This is 100% where I was when thinking about building an SBC farm - do something but scaled back a bit. I realized during this GPU fiasco, though, that I could buy a used 1650 (or 1060) and probably make a much bigger dent in computing needs than a fleet of even C4s/Pi4s/etc., and for less money and way less computer touching. So now I feel mentally free to spend my time on something else and if I still feel the itch, throw a $150 GPU at GPUGRID/OpenPandemics(?). Unrelated, anyone else getting wildly inconsistent times on ibercivis tasks (when you get them at all)? I've got some finishing in 3 hours and others estimating almost 15. (e: I think it's just bad at estimating time required, a lot of those tasks are already done) Discussion Quorum fucked around with this message at 23:49 on May 4, 2020 |
# ? May 4, 2020 16:23 |
|
mdxi posted:I don't view it quite that harshly on the whole -- especially since my wife, who can usually be relied upon to be a more reasonable person than I am, opined that it was a good use of time and money compared to what a lot of guys in my age range (I'm a prime candidate for a mid-life crisis) spend money on. A lot of people spend more on ski gear for a season than I do on compute hardware in a year, and then there's the guys who snap and buy Porsche convertibles and/or boats. I definitely spend an extra $150 or whatever to get a fancy i7 for SETI / Boinc / WCG over whatever totally fine i5 I could have bought, just because I know that I can leave it on overnight crunching numbers and such whenever its not in use. Its much easier to rationalize getting the big beefy cpu when its always in use, right?
|
# ? May 5, 2020 16:42 |
|
welp, time to shut down my main f@h setup. I got into the top 300 in the goon team, but now it's starting to heat up the home office. LoL fun idea: add a module to the client that google's the temperature and only runs during a set outside value. Like 13c or less, run F@H.
|
# ? May 5, 2020 20:25 |
|
jonathan posted:welp, time to shut down my main f@h setup. I got into the top 300 in the goon team, but now it's starting to heat up the home office. LoL fun idea: add a module to the client that google's the temperature and only runs during a set outside value. Like 13c or less, run F@H. Unironically want that.
|
# ? May 5, 2020 21:06 |
|
jonathan posted:welp, time to shut down my main f@h setup. I got into the top 300 in the goon team, but now it's starting to heat up the home office. LoL fun idea: add a module to the client that google's the temperature and only runs during a set outside value. Like 13c or less, run F@H. I paused mine yesterday because my condo was getting really hot, but I'm so close to twenty million points. I'll turn it back on tomorrow and crank up the AC.
|
# ? May 5, 2020 21:34 |
|
jonathan posted:welp, time to shut down my main f@h setup. Other, more invasive, possibilities are underclocking (what I presently do), or if you're using Ryzen CPUs lowering the PPT limit (what I plan to do next time I overhaul everything). quote:add a module to the client that google's the temperature and only runs during a set outside value. Edit: a one-liner. for fun: code:
Edit 3: Looks like `boinccmd --read_global_prefs_override` does exactly the thing you'd need there. So an automated, temperature-based, BOINC CPU throttle is very do-able. On Linux, anyway. mdxi fucked around with this message at 00:22 on May 6, 2020 |
# ? May 5, 2020 23:22 |
|
I realized today that my PC will be right by the thermostat in the new apartment we're moving to. Only other place to put it is the bedroom, which is even more problematic. Positioning may help but otherwise RIP my RAC. Gonna cronch hard for the next few weeks while I figure out a more thermally appropriate way to donate a bit each month. SBCs probably. Relevant: http://web.eece.maine.edu/~vweaver/group/green_machines.html
|
# ? May 5, 2020 23:38 |
|
This week in numbers: In Asteroids@Home, we've climbed to #187 in Total credit, despite the fact we're still focusing on COVID-19. In Folding@home, we've briefly gone down to #52 and are now back on #51. Folding's compute power has dropped to 2.38 exaFLOPS. In Rosetta@Home, I'm still getting COVID-19 workunits from them. Goons have dropped to #67 in Recent average credit, and have risen to #197 in Total credit. Of note is that their project URL has changed to https://boinc.bakerlab.org/rosetta/ (they switched from HTTP to HTTPS) so goons who haven't switched yet will need to remove Rosetta@Home from their BOINC Manager and re-add it using the new URL. Note that BOINC Manager still has the old URL listed, in case new goons want to add Rosetta and don't know this. In TN-Grid, all my workunits right now are outside the 165735_Hs to 166977_Hs range, so maybe their COVID-19-relevant work is over. Some people are reporting that they're still getting COVID-19 work, but it's sparingly. Our Recent average credit ranking has dropped to #16, and our Total credit ranking has risen to #70. In Ibercivis, we've dropped to #24 in both Recent average credit and Total credit. Available workunits have been sporadic in the beginning, but they're ramping up now.
|
# ? May 6, 2020 02:11 |
|
mdxi posted:Edit: a one-liner. for fun: FYI just doing "jq .properties.temperature.value" would also be valid. I think the dict notation is only necessary if the keys have punctuation or similar in them that needs to be escaped
|
# ? May 6, 2020 03:49 |
|
|
# ? Mar 29, 2024 11:08 |
I'm not sure why, but although I'm still doing fine getting WUs from Rosetta, and I still have plenty of space, I haven't gotten any more WUs from either World Community Grid or TN-Grid. I made sure to tell Rosetta "no new work units" and finishing up the ones I had before deleting/re-adding it, and I'm showing 6.88 GB of "free, available to BOINC" disk space so I dunno.
|
|
# ? May 6, 2020 04:44 |