PC owners borg into the most powerful computer the world has ever known – all in the search for coronavirus cure A distributed computing project for disease research now has more data-crunching chops than the world's publicly known most-powerful supercomputer – IBM's Summit at Oak Ridge National Laboratory. How so? Well, just like the lunatics stripping toilet paper from shelves the world over, at times of crisis it helps to feel that we …
Actually, I was reading El Reg, and ... what? That's not working? You sure???
Look, they say the average German male gets 25% of his daily nourishment from beer (the good German kind, the stuff that follows Reinheitsgebot). So, whose to say that I can't get 25% of my daily IT information needed to do my job from El Reg?
I'm running on 1 cpu and it takes about 4 hrs to do one while I'm using the laptop. Tried all 16 cores over lunch (PV running well here) but it cant run the fan fast enough to keep it cool - you could say it gives it a fever and breathing problems!
Need some compressed air to try and clean out fan and airways!
I heard about F@H when it first started thanks to the User Friendly (comic strip) team of forum regulars.
Until 2009 my hardware was too limited to be useful. After that, I just didn't care too much nor wanted to spend the electricity.
But lately I have a new comp with cycles to spare, so I joined the CoVi mob and signed up last week.
I think I've done two units; been mostly idle.
If I'm leaving my new big-box running instead of sleeping, please send more work!
> Nvidia are a bit of a problem Open Source wise
Understatement of the year.
Still, my i7 (no grfx card) system gets work units since I started yesterday, with only small breaks in between. SWMBO's system with a Nvidia card get no work unit since many hours.
Fun fact: one folding Vulture has the nick "UrethralAnts".
Hi! No WUs for me either, but as someone said here, this is good news as WU are finished as fast as they can pump them out. Their server(s) seem to be overloaded, not only do you get "Could not get an assignment" but sometimes a "Received short response, expected 512 bytes, got 0".
Icon for you handle.
Done! Folding Vultures are go, team number 250966. Join up, and let's get folding. My Ryzen 5 2600 is just about respectable (although clearly not earth shattering – the 24GB RAM might help though), but my GPU is nothing to write home about... come join in and see what we can do.
I've added my 3700x and GTX1080 this morning and already managed 5 WU (6th and 7th currently being worked on now), tried adding my home server too but kept getting a 'no work order for this configuration error' so I've left it be for now.
With regards to loads, after removing a stupidly noisy corsair fan from my system and replacing it with a spare from my last rig its barely any noiser than a decent gaming session (about 43db from where I'm sat) and it keeps my study nicely warmed. Win all round.
This post has been deleted by its author
Well this is awesome... according to the stats, we already have 44 members and have completed 17 WUs in less than 24 hours! We've reached the heady heights of team #11,947 of 242,342. Mine's sitting idle quite a bit at the moment, and apparently they're working to use their sudden jump in capacity, so I'm going to keep it running for when things get busier. Weirdly, I keep getting GPU units for my feeble card, but it only appears to be using 0.1–0.3% of it despite the slider being at Full, which amounts to almost nothing and means a WU takes a full day. Firefox is using more just sitting there (which probably says more about FF than anything else). It's actually using more of the CPU to keep the GPU fed (~1–2%)... Any suggestions on this front would be welcome.
On the plus side, when I get CPU WUs, they totally max out all 12 logical cores, but balance the priority well enough that I didn't even notice the last one running. This was despite having multiple large Word documents open, an Excel spreadsheet, numerous FF tabs, and Texmaker occasionally compiling stuff. Impressed with that.
Blimey. Only time I ever fill my 24GB is when some shoddy part of Windows 10 has a memory leak and I see silly levels of commit before having to reboot! My general tick-over for day-to-day work is 12GB committed, 3GB standby and 9GB free (with no pagefile – which may spark debate). 56GB has to be enough for some serious 4K video editing :P (And your "only" GTX 1060 is still six times the power of my pathetic Radeon 7750...)
I reckon it all helps. I've disabled GPU folding, because my card wasn't really able to get units done in time (and it didn't even seem to be using its full potential for some reason). I'd say 30min is great, and 2h30 is fine, as the timeout is usually 24h. I'm getting about 1h per unit with 12 logical cores (6 physical). I've found that even with all 12 cores allocated (manually) at the Full setting, I don't even notice the load during general working. Maybe it'd be different if I was photo editing or something, but it seems to have set the priority appropriately low so far.
I'm running a system with two graphics cards, and the processor often sits idle. I think some WU are GPU specific. But the processor does get one occasionally. I expect more will start arriving once they scale up the work allocation servers.
I would just leave it on ready for the WUs when they arrive.
Also you could look at sticking in an older GPU, I just added a GT640 and it's getting WU allocations.
Done, although have a fairly shit PC FWIW (i5 2500K @ 3.3GHz + 1GB RADEON HD 5770 - it was good about 8 years ago, I suppose). All the client web UI is showing is Cancer, Parkinsons, Huntingtons & Alzheimers - any idea how to get COVID-19 added to the list? Or is this to do with the shortage of Work Units?
Not sure if anyone will read this now... but there was a problem with stats for a while. It seems to be fixed now. I can't see you either though (unless your name is different on FAH) – have you created a Passkey? I've made a shortcut to an alternative stats site, anyway:
Over 200 members and counting, and we just made it into the top 2,000 teams. I have a new GPU arriving today, too...
This kind of "hack" says: no problem at all (in fact, even better) IMO: the 250966 team's mentioned in this article as being made up of a bunch of Reg commentards (well, "Reg Readers" - but that's different lingo), so go for it :)
Might be also nice to link to this article, rather than the homepage, too.
I'll shut up and continue crunching - the runner-up is getting perilously close.
Good point. I've changed the link to take people to joeW's post where he suggested setting up the team. I've vaguely thought about setting up a landing page with a description and other links, but I've got actual work to do! All suggestions welcome, though.
PS – Great work on the points front. You must have an absolute beast of a graphics card to get 1.33 million points in one day on a single client yesterday! I'm gonna play the guessing game... I reckon an RTX 2080 or maybe a TITAN?
NVIDIA GeForce GTX 1080 Ti, close enough!
I've not enabled the CPU ones as despite liquid cooling the fans go crazy and even I can hear them...
The graphics card is making some fan noise and it's disturbing my partner... so I might have to take the system down for a thorough cleaning during the week-end.
Glad to give you a challenge ;)
I've got 3 systems I'm throwing at it, one of them only part time since the noise and heat would keep me up at night otherwise (plus it seems to have some overheating issues if it runs at full blast for too long despite having liquid cooling).
Part time i7 6700k with 2x 2080 Ti founders editions -- at least one of the GPUs tend to sit idle more often than not though
Full time VM on a dual xeon e5-2620 v4 with 8 cores allocated (need the rest for my other VMs)
Full time i7 8700b
So... beating all of that with a single 1080 Ti is rather impressive, so kudos :)
16 Cores Xeon E5-2667v2 added to the mix and this is all it will be doing for the time being. This was to be a db / software server for project at work but covid hitting the fan derailed project start just before we were set to get things up and running. Got approval from the boss to dedicate it to this project until we're a go.
Indeed. I've been running World Community Grid (BOINC-based) for years and its very lean GUI allows you for choosing the % of CPU you allow it to use.
The other day I've installed F@H (Win7) and I do think the GUI is not too clear and most of all the settings are a bit rough around the edges.
At first ("Light" setting) it wouldn't use my GPU (GTX1060) at all and very little CPU so I thought something was wrong with my (old) PC, but then I found this page(*) that explains it and switched to "Medium". The CPU went to almost 90% and the GPU to 100% with all fans blowing at full speed.
No, I didn't try "Full".
It's not that I don't want to contribute, but if my main rig gets fried I have a 10-years old Fortnite pro at home who's going to be very pissed.
May give a try with Rosetta@Home which uses BOINC. Still, the guys seem to have plenty of firepower already. Science is cool.
(*) https://foldingathome.org/faqs/fah-v7/v7-introduction/web-control/folding-power-slider/
OK, perhaps I was a bit lazy at first, now I understand how this goes a bit better. Most of you probably know what's below already but just in case...
Here (Win7, ancient Phenom II X6 + GTX1060) when set to "Medium" I have the CPU running at ~85% and the GPU at 100%. "On idle" unchecked.
Both running at ~60°C.
I understand there are options to throttle CPU usage so I'll be tinkering with that, but this is not possible for the GPU, it's all or nothing. GPUs don't have any of the fancy throttling mechanisms modern CPUs have.
My solution is to use the very nice utility from Gigabyte to underclock the GPU.
Still, with this I can compute a WU (GPU) in 3 or 4 hours so I will be generating one per day.
As a comparison while the GPU was on 1 WU + 30% of the next one, the CPU was still chugging at ~25% of its first WU.
Compared to BOINC I don't understand the notion of WU : BOINC uses one core per "task" for any given project, it's simple. Is a F@H WU a "task", meaning that it is balanced across all cores ? Or is it more complex than that ?
Other than that Rosetta@Home works fine, it allows for "fine throttling" of the CPU, but it cannot use the GPU.
Perhaps a good combination if you want to run this 24/7 without too much strain on the system would be Rosetta at, say, 70/80% of the CPU, and F@H with no CPU slot and an underclocked GPU ?
I installed Rosetta yesterday, and found it immediately fired up 12 tasks, each using 1GB RAM... my (pagefileless) system didn't enjoy it... Have squished it to 6GB, which seems to limit it to 6 threads. TBH, I'd rather it did one task across 12 cores than 12 tasks each needing their own chunk of RAM. Seems more friendly to working on my machine at the same time. This is a big advantage of FAH, in my opinion. Though I have 24GB to play with, I'm usually using 12 just ticking over with day-to-day usage. Maybe I should re-enable the pagefile, but I'm reluctant... old habits die hard. (Yes, I understand the reasons why not having a pagefile might be a bad idea, but if I can have 8GB RAM and a 16GB pagefile, surely I can have 24GB RAM and no pagefile... Space on my SSD is limited, and I'm buggered if I'm letting it scribble on my spinning rust. I'm prepared for things to suddenly die for want of RAM in extremis. I'd rather that than sudden disk dumping – it saves me time in the long run.)
With tasks accessing the memory in a more or less random way, disabling paging/swapping makes a lot of sense. Once any non-trivial amount of memory gets paged in and out (all the time), your cores are mostly idle, waiting for memory pages (thrashing). I have seen slowdowns by a factor of more than 1000, the computer basically doing nothing but testing how many operations the disk will survive.
Dedicated "out of core" algorithms exist for certain tasks (e.g, for FFTs) but unless the documentation says they have that feature it is better to assume that it absent.
I'm running it on 3 machines.
One is a Q6600 that we only use to watch netflicks on our dumb tele. I opted for full, but haven't noticed extra noise... but it is a noisy machine at idle anyway.
The other is a i7-2600 for watching netflx on our other dumb tv, and that doesn't seem any noisy than usual.
The third is a i7-6700 built in an old case that isn't particularly quiet, but doesn't seem any noisy than usual either.
They are normally sleeping or hibernating, so having them on is noticeable if you are in the same room. But they aren't in bedrooms, so they can run all night without disturbing us. And keep us warm.
how much higher was your electric bill? I estimated $50-$100/month running things like that... screaming fans, CPU maxed out, etc.. There were a couple o' projects from the mid noughties I tried, Seti@Home being one of them. But modern CPUs have lightweight power usage during idle states, so no screaming fans, no high electric bills from leaving them on all the time... unless you run stuff like this.
and the other thing: It's a fair bet (in my mind) that these work units weren't coded using the MOST EFFICIENT TECHNIQUES. for all I know, they're doing the research scientists' equivalent of a BUBBLE SORT. So before they ask for my computer time to be donated, I'd like to see some peer review on their algorithms...
besides - the BEST research is being done by physicians in hospitals with various experimental (but promising) treatments on real patients right now. "Feel Good"-ism might warm up your house in the vicinity of your computer, but how much good is it REALLY doing?
/me proposes that ONE good programmer and data analyst is worth a MILLION 'screaming fan' machines... given a reasonable amount of time, coming up with efficient algorithms and simulations, rather than throwing data against a wall of computing resources, which is [in my opinion] VERY likely to be the case.
We're not cracking ENIGMA, which sometrimes involved a lot of repeated-trial-and-error until they found the one that worked. But then again, THEY made THEIR process efficient, back then, and were reading Hitler's mail in real time! And so we should do that NOW, make the virus research algorithms as efficient as possible, BEFORE recruiting a bunch of donated CPU resources.
Look, if you don't want to use it then don't. But please don't disparage people who are putting a lot of effort into trying to help. It might be entirely useless, but running this thing actually doesn't impact me (or a lot of people) much at all. And it's pretty likely to have some kind of a positive effect.
As to your proposal that one good programmer and data analyst is worth a million machines... how does one good programmer and data analyst with the power of a million machines match up?
...
I just fed the troll, didn't I?
Protein folding simulations involve stochastic processes, and what FAH essentially does is throw random numbers at the wall to see what sticks. Each folding operation is pretty efficient, but yes, there will be >99% non-world-changing data generated. However, to get that remaining <1%, we have to throw a lot of computing power at it. I can tell you one thing: it's a damn sight more useful than coin mining. If we could somehow throw all the exaFLOPS used for mining at this, the world would be a better place.
I flung my combined room heater and phat old server into it and have been crunching Covid-19 work units that they've allocated 3 days for in less than 6 hours. Only problem is they keep running out of work units, so now it's sat idle for more time than I'd like, but it's great to know that a community like this can (safely) come together to hopefully make a difference!
Check out Rosetta@Home on the BOINC distributed computing client. They're working on Covid-19 too. If there's no Folding@Home work units you can switch over to Rosetta. The projects are complimentary - Rosetta calculates protein shape and Folding, well, folds them.
Just be aware that Folding @home and BOINC are two quite different systems - so both need to be installed separately.
Folding allows you to pick certain tasks although any one can be downloaded to your PC for crunching.
[url]https://foldingathome.org/iamoneinamillion/[/url]
BOINC supports many different projects, of which two are making efforts to tackle Covid-19 issues - Rosetta@home is already up and running and TN-Grid will be doing something similar. Both can be chosen to be installed via the BOINC Manager interface that needs to be installed.
[url]https://boinc.berkeley.edu/[/url]
[url]http://boinc.bakerlab.org/rosetta/[/url]
[url]http://gene.disi.unitn.it/test/[/url]
I signed up yesterday. According to the log files the client completed 99% of its first assigned task, then gave up. It seems to be hung now, so I have killed it off.Sorry, guys, but if I am going to give you access to my computer it will have to be with something I can trust not to fuck up.
They haven't fucked up: they are running out of bandwidth and are working with a number of folk (incl Linus TechTips with their 10Gb fibre link) to move some of the server load off-site. They are also in the process of ramping up the work pipeline, as they are pleasantly overwhelmed by the sudden uptake: at least 3 publications have highlighted what they are doing, all at the same time (some folk have switched from SETi to fold@home). Plus they are a university programme, not a corporation. So maybe give them a break?
Just let it run for a few hours it WILL pick up work as it is available - over the last 2 days mine has been working about 70% of the time, set on medium while I'm browsing.
Medium settings (on a 3950x/vega 64,32gb memory) runs at 30 cores leaving 2 cores for me :-)
Using the creators settings on ryzen master my average temp is around 60 after 2 hours or so.
Also a nice test for your rig :-)
I'm also investigating using my 3d Printer too make facemasks or ventilators,Some guys in Italy have designs undergoing tests.
Would be amazing, if true.
Since most of the compute come from gaming hardware with very dismal fp64 FLOPS capacity, 470 fp64 PFLOPS would be more like 20 fp32 EFLOPS! (and most compute is likely not fp64)
I find that really hard to believe.
I could not find any concrete stats to back this up.
I think you may not understand how scientific and medical improvements are made... It is nerds that do that. Mostly in labs, yes, but there's a lot of computing that goes into it all these days. So, yes, the "nerds of the world" are going to rush in and save everyone. That's their job.
It's not as much the nerds rushing in to save us all that bothers me, it's who gets to profit from this afterwards? It's all very well everyone making a concerted effort to crowdsource the work required for a cure, only for some corporation to step in at the 11th hour, patent it and withhold it to maximise profits for a few shareholders. I can't really remember the last time a cure for anything didn't come with a patent attached. Maybe Lorenzo's Oil perhaps.
Most of the computer nerds you speak of - me included - have no idea how to research a vaccine.
We do, however, know that those who do know, need oodles of CPU power to test their simulations, so we are helping out..
Just like 50,000 people giving money to a cancer charity doesn't mean said 50,000 people are going to personally cure cancer.
Is that really so hard to grasp?
Tasks: 310 total, 2 running, 250 sleeping, 0 stopped, 0 zombie
%Cpu(s): 8.9 us, 9.5 sy, 65.5 ni, 16.1 id, 0.1 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem : 16005364 total, 8189480 free, 3927256 used, 3888628 buff/cache
KiB Swap: 999420 total, 999420 free, 0 used. 10942376 avail Mem
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
5623 fahclie+ 39 19 428184 173880 13548 R 292.4 1.1 11:14.56 FahCore_a7
4679 rkynast+ 20 0 2905864 283376 123000 S 25.7 1.8 11:39.29 Web Content
4504 rkynast+ 20 0 4159288 503024 207480 S 7.3 3.1 37:57.16 firefox
Have used BOINC for years contributing to fighting AIDS and all sorts of cancer projects. Glad to help and adjusting the settings is simple on the gui. Never interfered with my work so would reccomend. Have stealth installed on clients pcs (hide system icon) as well with no repercussions yet....
Agree. What could be done is to ask officially your employer (major IT vendor here) to install whatever useful grid computing software on some big iron in demo rooms, test/PoC labs, currently idle or much less used because of the confinement.
It would be more symbolic than anything, but a nice symbol.
One company is doing its part by using highly advanced supercomputers to simulate/explore the use of lipid molecules to encase Ethyl and Propyl groups. The lipids that have a specific affinity for the surface proteins of the Corona Virion will attach ONLY to that protein bundle and not other body cells/structures.
The lipid that contains Ethyl/Propyl alcohols will encase the entire virion or the spiky projections of the Coronavirus Virion which will then allow the Ethyl/Propyl groups to release onto and then denature the Virion itself, blowing it apart and poisoning/disconnecting the interior RNA/DNA sequences so it cannot replicate. They've already had much success and are seeding parts of the new-found knowledge to teams worldwide.
I'm not the medical/physics guy so that's MY blurb on what they're doing!
P.S. Corona is a word that means CROWN of which MANY have spiky bejeweled projections coming out from the base - ergo, you get Crowned-Virus as the origin of the word!
..
Please foregive my ignorance on the matter, I can't seem to configure my RPi worth a damn.
Is it possible to configure an RPi to work as a F@H client & run 24/7/365 to donate it's admittedly meager computational nudges to the overall effort?
Could a person create their own small farm of RPI's (a server farm of Pi? A PiFarm!) & configure them to run lots of nudges to the effort?
What kind of effort could a (small) army of geeks/nerds all contributing their PiFarms to the totality do to benefit even moreso than just using their gaming/workstation rigs to do their parts?
Is a PiFarm even possible, or do I need to take my dried frog pills again?
I have a couple of nice rigs at work, off-corporate network; normally used for FEA. Before our office closed down (ahead of the gov't) advice; I pointed the lot at F@H.
Apparently F@H now has so much processing capability it can't keep up with recieving the results and dishing new work out.
On Windows most browsers notify the OS when they download a file, and it's automatically scanned by whatever virus checker you have installed before it's saved to disk in the final location.
eg, in Firefox it's controlled by the browser.download.manager.scanWhenDone setting.
OS Arch: AMD64
10:07:50: GPUs: 0
10:07:50:CUDA Device 0: Platform:0 Device:0 Bus:7 Slot:0 Compute:6.1 Driver:10.1
10:07:50: OpenCL: Not detected: Failed to open dynamic library 'libOpenCL.so':
10:07:50: libOpenCL.so: cannot open shared object file: No such file or
10:07:50: directory
Ryzen 2700X 16 cores available. GEFORCE GTX1050
Linux Mint 19.3 How do I get the GPU on the case?
Tah. Used: sudo apt install ocl-icd-opencl-dev
Not sure what is going on but now I show 1 GPU slot and 1 CPU slot running. I had to mess about in config, override -1 on some settings to get the GPU up. Again no idea as to what it did. The only indication that the GPU is doing something is the progress bar is moving and the temperature has gone up from 50C to 67C even though GPU utilisation is stuck down at 3% when I wiggle the mouse.
13:33:14:<config>
13:33:14: <!-- Client Control -->
13:33:14: <fold-anon v='true'/>
13:33:14:
13:33:14: <!-- Folding Slot Configuration -->
13:33:14: <gpu v='false'/>
13:33:14:
13:33:14: <!-- Network -->
13:33:14: <proxy v=':8080'/>
13:33:14:
13:33:14: <!-- Slot Control -->
13:33:14: <power v='full'/>
13:33:14:
13:33:14: <!-- User Information -->
13:33:14: <team v='250966'/>
13:33:14: <user v='Camilla_Smythe'/>
13:33:14:
13:33:14: <!-- Folding Slots -->
13:33:14: <slot id='0' type='GPU'>
13:33:14: <cuda-index v='0'/>
13:33:14: <gpu-index v='0'/>
13:33:14: <opencl-index v='0'/>
13:33:14: </slot>
13:33:14: <slot id='1' type='CPU'/>
13:33:14:</config>
13:33:37:WU00:FS01:0xa7:Completed 10000 out of 500000 steps (2%)
13:34:14:WU01:FS00:0xa7:Completed 240000 out of 500000 steps (48%)
I had no idea this would turn into such a cock-waving contest!
But seriously, well done everyone, and don't forget you can still give blood during the crisis. They need it just as much as ever - and it counts as an essential journey if you want to get out of the house ;)
Just to say, Folding Vultures (team 250966) now has 84 members and has entered the dizzy heights of the top 10,000 teams. Whoever AgentMyth is... wouldn't mind a bit of your system – it's responsible for nearly half the team points so far! Nice work.
>>The F@H Linux software is too old to run on modern Ubuntu and some other distros.
Not my experience - I have it running fine on latest Ubuntu, and a Mint 18.x machine, but when I tried on my old Ubuntu 14.04 box, the Control app wouldn't connect to the Client.
But I would like to know more about platform comparisons - eg is there any difference in performance or usefulness between running it on Windows and linux machines, and would I be better running multiple VMs or a single host on full whack - any knowledgeable folks out there?
It's running fine on my main desktop OC (Linux Mint) and my teeny Lenovo mini-laptop/tablet running a forked version of Ubuntu (cant recall the distro name now, it wasnt one I'd heard of but was the only one that installed with the screen the correct way up rather than rotated 90 degrees). It's also doing SOMEthing (only just installed it and so not sure if it's actually working ok until it gets its first WU) on my backup desktop PC which is very venerable indeed.
Folding Vultures seem to be racing up the board - ranked 1,325 when I looked at the stats a little while ago. Well done all, keep it up!
Set up in 2007; join the old timers!
https://www.theregister.co.uk/2007/04/27/grid_cancer_research_results/
https://stats.foldingathome.org/team/69263
https://folding.extremeoverclocking.com/team_summary.php?s=&t=69263
No matter what team you join, it's for a damn good cause.
Been running BOINC for ages with no problems. However, after 3 days of waiting, I'm still not getting any WU from F@H.
Good news is that there is a BOINC project - Rosetta@home, that is also doing COVID stuff, not sure it's the same stuff as F@H.
So for now, I'm sticking with the BOINC stuff - every bit helps, but my spirit is with you guys!
What surprises me is that no big IT company such as Google has said to vaccine researchers :
You want a ton of computing power ? Here it is. Be our guest.
It is not only a definitive case of Public Service but it would be both great Publicity AND a place in the history books to be able to say :
The company whose computers helped find the covid-19 vaccine
A tad smaller than Google, but definitely doing something about that :
https://www.hpe.com/us/en/newsroom/blog-post/2020/03/challenging-times-call-for-clear-heads-purpose-driven-priorities-and-a-focus-on-communities.html
Mind you, with their "legacy" HPC and now the Cray people and resources, I believe they can do interesting things. Not to mention Aruba wireless deployments in makeshift "hospitals", etc...
Antonio Neri seems to be a rather decent fellow, especially compared to others CEOs. I guess the fact he's Italian only added to the incentive.
Publicity, genuine good will, whatever, anything goes to fight against this little biatch.
Got my 2 main systems running 24/7 and added a third PC that was not in use. Ryzen 2700X and a EVGA 970 an 8350 and EVGA 960 and a 8320 and a Gigabyte Gt450. In less than 24 hours got 65K points. Have a Radeon HD7770 sitting spare but not sure about to different video cards in one machine. May be I'll have to knock up a system out of a case just to put that video card to use. Joined Team 250966 Folding Vultures.
You can run multiple GPUs with FAH – each GPU gets a separate work unit – and modern systems (with an appropriately rated PSU) should handle it. I've just got myself an RX 580 (not super high-end, but fairly cheap on eBay at ~£80), and that runs at around 450,000 points per day (assuming it can get the WUs). It seems that GPU folding is way faster than CPU.
> It seems that GPU folding is way faster than CPU.
Judging by point-count you cite that seems to be true. But there may be situations where CPU only is not that bad. From the top of my head, if the FPU can do transcendental functions (exp etc.) that would be an advantage against GPU when needed, or when tons of communication between threads is required, or when there is a lot of conditional branches in the code.
I'm not saying GPUs are good at everything... just folding! I gather they're better at doing repetitive, relatively (numerically) simple, embarrassingly parallel stuff. Otherwise we'd've all given up on CPU architecture by now. But yes, I've turned off CPU folding since getting a passable GPU. However, I do believe that there are some WUs that require a CPU. So people shouldn't all abandon CPU folding.
But anyway... I'll wait until I get a better CPU cooler before turning it back on... and maybe some better exhaust fans too. Since I've now got what is essentially a 180W heater sitting under it, at full load my CPU was heading for throttling territory at 95°C. I undervolted it to 1.0V (which I was surprised it dealt with stably) and that's improved things a bit, but that's no real substitute for a good cooling system!
> ... wait until I get a better CPU cooler ...
Top tip: disallow the CPU frequency boost partially or (better) completely.
That top end of he freq. range is very energy-inefficient compared to (up to) base freq. For my machine (i7-8700K, 6 cores), base freq. 3.7 GHz, boost 4.7 GHz, the difference at full load is "almost silent" versus "WOOOOSH!". Really, give it a try, you may skip buying a new cooler altogether!
Not science, but I like that icon.
There are two vulture teams, 250966, Folding Vultures (the new team) and 69263, Vulture Central 2.2 (the old team).
As jobs keep coming me and SWMBO keep our systems running day and night.
Celebratory beer -------->
Given that the overal no.1-ranked folder has joined our team (and seems to be folding for us exclusively) - https://stats.foldingathome.org/donor/1437 - I'd say that was pretty likely. I wouldn't want their electricity bill landing on my doormat!
The No1 folder is not a specific person, but anyone who hasn't set up a username, so is folding as the catch-all 'Anonymous'.
However, some of those have chosen to fold as part of team 'Folding Vultures' (team 250966).
Have a virtual beer on me though, because as I write, the team is ranked 379 out of 253,437.
I've had a 10CPU VM running the FaH appliance on some spare blade capacity in our on-prem DC, just added a second VM as we have the bandwidth.
Used the instructions here to create a lightweight VM.
https://www.virtuallyghetto.com/2020/03/how-to-manually-install-folding-home-on-vmware-photon-os.html
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
