LTT Official Folding Month V (2022)

[NetoriusNick]

16 hours ago, GOTSpectrum said:

Best alt connection to virgin where I am is 6/3 lmfao 

I would recommend in the UK, Zen & Starlink. both have worked very well for me. currently in the process of getting fttp

[justpoet]

16 hours ago, BondiBlue said:

I fold on at least three other Macs, and none of those have any issues with FAHControl. Two of them are running macOS Catalina and one is still running macOS High Sierra. The only big difference is that those Macs are Intel based (2011 iMac, 2012 Mac minis) while the problematic machines are Apple silicon based. 

My macs are all intel, and across several versions of Mac OS (I think ElCapitan through Big Sur).  It happened to all of them all at once, and was reported by others as well, so there's definitely some config/update combination that caused it at the same time as the last F@H version increase back then.  I'm just not sure what.  I've fully uninstalled (including configurations) and reinstalled F@H of various versions and that didn't solve the issue.  With my lack of time these days, I can't be bothered to figure it out, but if you do, please let us all know.  I'd definitely spin up a few of the Macs in addition to my daily driver laptop to try out something you found to fix it for you.

[Alex Atkin UK]

7 hours ago, NetoriusNick said:

I would recommend in the UK, Zen & Starlink. both have worked very well for me. currently in the process of getting fttp

Unfortunately Zen have some random issues with their GEA backhaul they seem to be struggling to sort out, fortunately they don't have GEA at my exchange so I'm using the standard BT Wholesale backhaul which works great.  At some point Cityfibre is installing here too so I will be able to upgrade Zen to symmetrical on their network.

 

I'd only go Starlink if there are zero other options as I strongly believe its not an economically viable service and will go bankrupt, then you're left with an expensive paperweight.

[Gegger]

20 hours ago, BondiBlue said:

Anyone here folding on an M1 Mac? If so, how do you use FAHControl? I've got it running on both of my M1 MacBooks right now (a 13" M1 MacBook Pro (8GB RAM, 8 CPU cores, 8 GPU cores) and a 16" M1 Max MacBook Pro (32GB RAM, 10 CPU cores, 32 GPU cores)), and FAHControl is basically unusable on both of them. They're both folding just fine in the background, and I can control them through the web control page, but FAHControl just keeps updating over and over. 

 

Also, these laptops are getting much different results. The 13" is only reporting ~21K PPD while the 16" is reporting ~127K PPD. Both are folding on 8 CPU cores (the M1 Max has 8 performance and two efficiency cores, and F@h can only use the performance cores, at least by default), so they shouldn't be this far apart. Any thoughts? 

I'm on Monterrey on a M1 MBP 2020 13" and it seems to say updating forever until I open the web client at the same time then it magically decides to work?

I'm getting like 50-60k PPD but it often drops to like 20k with cancer WUs - that's also an iffy number because I set it to only fold when it's plugged in and I'm usually working at the same time

[Alex Atkin UK]

4 minutes ago, Gegger said:

I'm on Monterrey on a M1 MBP 2020 13" and it seems to say updating forever until I open the web client at the same time then it magically decides to work?

I'm getting like 50-60k PPD but it often drops to like 20k with cancer WUs - that's also an iffy number because I set it to only fold when it's plugged in and I'm usually working at the same time

Did you manually change the number of threads so its exactly the same as the number of performance cores (assuming this is Apple Silicon that's 4 for the base M1)?

[Gegger]

4 minutes ago, Alex Atkin UK said:

Did you manually change the number of threads so its exactly the same as the number of performance cores (assuming this is Apple Silicon that's 4 for the base M1)?

Assuming that I can read and am using FAHControl properly, I have it set to letting the client choose (-1) because I was lazy when setting it up

[dogwitch]

1 hour ago, Gegger said:

Assuming that I can read and am using FAHControl properly, I have it set to letting the client choose (-1) because I was lazy when setting it up

i do wise setting up core count .would be easier thru.

anyhow both TR machines will be doing wu.

am trying to get a third machine up that has 3 gpus.

 

[cbigfoot]

18 hours ago, Gorgon said:

50/10 here on 50 year old buried copper. Lucky if I get 32Mbps down. 10 Up is never a problem and I've never had upload issues with F@H that weren't attributable to the servers at WUSL.

 

Just to taunt me the ILEC here installed Fibre. Only problem is they delivered it at the front of the house and the front of the basement is the finished portion so I'd have to do some major surgery to get it where I need it.

 

Doen't really matter though as I'd never go with them as an ISP again so I'll have to wait until their forced to open up Fibre to competition.

Same here 1972 Copper to Cat 6EShileded when it enters the BLDG LOL…they wanted to run cat 5…i was like i want over kill i have cat 6EShileded from a job i did…they gladly pulled it with the Cat 5 next to it Just in case…

[cbigfoot]

2 hours ago, Alex Atkin UK said:

I'd only go Starlink if there are zero other options as I strongly believe its not an economically viable service and will go bankrupt, then you're left with an expensive paperweight.

I think the opposite that they will put the other satellite vendors out of Business.  Once they get the Super heavy running it will put the Other Vendors to shame…oh wait it already does in most places…My parent have it for their RV and Love it.  Being spacex is vertically integrated they make their own sats and receivers and Routers it’s Highly efficient compared to what any other company does.

[Alex Atkin UK]

17 minutes ago, cbigfoot said:

I think the opposite that they will put the other satellite vendors out of Business.  Once they get the Super heavy running it will put the Other Vendors to shame…oh wait it already does in most places…My parent have it for their RV and Love it.  Being spacex is vertically integrated they make their own sats and receivers and Routers it’s Highly efficient compared to what any other company does.

But by Musks own admission, SpaceX is on the verge of collapse.  The question is if they can survive long enough to get the costs down.

They are also relying on constantly increasing subscriber numbers, which surely will peak and fall as fibre is rolled out across the world, offering a vastly superior solution.

Sure there will still be use cases and rural users where its their only option, but is there enough of those to cover the costs?

[cbigfoot]

11 minutes ago, Alex Atkin UK said:

But by Musks own admission, SpaceX is on the verge of collapse.  The question is if they can survive long enough to get the costs down.

They are also relying on constantly increasing subscriber numbers, which surely will peak and fall as fibre is rolled out across the world, offering a vastly superior solution.

Sure there will still be use cases and rural users where its their only option, but is there enough of those to cover the costs?

I think he’s exaggerating slightly they wouldn’t be cutting launch costs as much as they have been if there was an issue esp being that Russia is out of the picture…but i think once the 1.5(currrently being launched) and ver 2(super heavy launched) get out there with the laser links and can be used for back haul as well and the cruise indistury as well as shipping industry and RVers and there are areas in the US so remote that it would be astronomical to hook them to a fiber feed for a small town.  (Alaska for instance) there are towns there that once the polar sats all get in place will be on the high speed internet for the first time…No i don’t consider Hughes net and other GSO sats as high speed.

[Gorgon]

13 minutes ago, cbigfoot said:

I think he’s exaggerating slightly they wouldn’t be cutting launch costs as much as they have been if there was an issue esp being that Russia is out of the picture…but i think once the 1.5(currrently being launched) and ver 2(super heavy launched) get out there with the laser links and can be used for back haul as well and the cruise indistury as well as shipping industry and RVers and there are areas in the US so remote that it would be astronomical to hook them to a fiber feed for a small town.  (Alaska for instance) there are towns there that once the polar sats all get in place will be on the high speed internet for the first time…No i don’t consider Hughes net and other GSO sats as high speed.

I spent a week at Iqaluit on Frobisher Bay, Baffin Island a few years ago. I was volunteering taking pictures for an organization and had to upload low res JPEGs every evening. The single 100Mbps TeleGlobe Sat Link serving the whole community was so over subscribed I had to schedule the uploads starting at 02:00 to get the 200MB or so of the day's images through.

 

At work we have sites in the far north with Sat Links and I can tell you that there are a lot of communities in Canada that can benefit from Starlink and similar services. Most of Canada is only densely populated in a narrow ribbon close to the Southern Border and the Telcos are unwilling to run broadband into rural areas leaving dial-up as the only economical option.

 

We tend to take broadband access for granted in urban areas and moan about only having 50/10.

[Alex Atkin UK]

54 minutes ago, Gorgon said:

At work we have sites in the far north with Sat Links and I can tell you that there are a lot of communities in Canada that can benefit from Starlink and similar services. Most of Canada is only densely populated in a narrow ribbon close to the Southern Border and the Telcos are unwilling to run broadband into rural areas leaving dial-up as the only economical option.

 

We tend to take broadband access for granted in urban areas and moan about only having 50/10.

Thing is, none of that means low earth orbit satellites are a more economic option to land based alternatives.

I mean I get it, they went this route to avoid the anti-competitive BS governments are allowing that prevent alt-nets, but just because its a good idea in principle, doesn't make it economically (or environmentally) viable.

There's a genuine concern that if a few do not de-orbit successfully, they could cause a chain reaction that takes out every satellite and prevents launching any new ones.  We really need a viable backup plan to deal with space junk before putting orders of magnitude more up there.

 

Also if Starlink lasts just long enough to build new businesses in rural places relying on it, it could be a much bigger disaster if it were to suddenly disappear.

[izz]

On 10/3/2022 at 7:03 PM, YoungBlade said:

I was planning on doing folding to heat my room this winter, so this should work out well for me. I'm in.

Sorry, I'm confused. How would folding heat your room?

[BondiBlue]

22 minutes ago, iIzzTheWiz said:

Sorry, I'm confused. How would folding heat your room?

Folding@home uses your GPU or CPU to do work, and it can put them at full load. Doing do creates heat, and on the kinds of components most of us use for Folding there's a lot of heat being generated. 

[Takumidesh]

I'm sure this is normal but i wanted to make sure.

 

So I threw my old 970 into my server for folding. I want to just fold with the gpu so I removed the CPU slot in the config. but now I have noticed that there is still a  core running at 100% all of the time, the core running at 100% bounces around. but there is always one at 100.

 

[Alex Atkin UK]

4 minutes ago, Takumidesh said:

I'm sure this is normal but i wanted to make sure.

 

So I threw my old 970 into my server for folding. I want to just fold with the gpu so I removed the CPU slot in the config. but now I have noticed that there is still a  core running at 100% all of the time, the core running at 100% bounces around. but there is always one at 100.

 

Yes its normal. It often takes 100% of a core to feed the GPU, so much so that upgrading the CPU will sometimes increase the GPU PPD.

[Takumidesh]

Just now, Alex Atkin UK said:

Yup, it often takes 100% of a core to feed the GPU, so much so upgrading the CPU will sometimes increase the GPU PPD.

I see, that makes sense. Do you think it is worth it to pin that core then? or just let it bounce around. I am assuming that feeding the GPU work is a single threaded operation, or is there a way to split that up among the cpu.

[Alex Atkin UK]

10 minutes ago, Takumidesh said:

I see, that makes sense. Do you think it is worth it to pin that core then? or just let it bounce around. I am assuming that feeding the GPU work is a single threaded operation, or is there a way to split that up among the cpu.

I don't think we've ever looked into that specifically.  I wonder if switching between cores actually lets each core boost higher?

[Takumidesh]

7 minutes ago, Alex Atkin UK said:

I don't think we've ever looked into that specifically.  I wonder if switching between cores actually lets each core boost higher?

I imagine if anything, switching between cores helps the IHS spread the heat around more effectively and probably runs a little cooler, but on the flip side, constant context switches and cache busting from switching cores may be detrimental. But that stuff is out of my league haha.

[Gorgon]

31 minutes ago, Takumidesh said:

I see, that makes sense. Do you think it is worth it to pin that core then? or just let it bounce around. I am assuming that feeding the GPU work is a single threaded operation, or is there a way to split that up among the cpu.

Likely not. The locked core is a function of the NVIDIA driver. It forces a thread to be locked for some reason and that thread spends much of it's time idle. It is often referred to as "spin-wait."

 

AMD, on the other hand, does not do this so their driver just requests a thread when needed.

[Alex Atkin UK]

4 hours ago, Gorgon said:

Likely not. The locked core is a function of the NVIDIA driver. It forces a thread to be locked for some reason and that thread spends much of it's time idle. It is often referred to as "spin-wait."

That explain why when I tried to lock CPU folding to a single CCD on Ryzen it didn't move the GPU thread off that CCD automatically, I had to make sure I used the other CCD so CPU and GPU WUs didn't clash.

 

It makes CPU folding really inefficient on Ryzen given every WU starts a whole new process and needs affinity pinning again.  Just leaving it bouncing between CCDs lost about 1/3 of the PPD.

[GOTSpectrum]

8 hours ago, Alex Atkin UK said:

There's a genuine concern that if a few do not de-orbit successfully, they could cause a chain reaction that takes out every satellite and prevents launching any new ones.  We really need a viable backup plan to deal with space junk before putting orders of magnitude more up there.

The fact they are low orbit and light weight with high surface area means they will de-orbit themselves pretty quickly. They have engines on board that are needed to keep them in orbit for any length of time. 

[Alex Atkin UK]

1 hour ago, GOTSpectrum said:

The fact they are low orbit and light weight with high surface area means they will de-orbit themselves pretty quickly. They have engines on board that are needed to keep them in orbit for any length of time. 

But when you have the density that they are planning, "pretty quick" is not necessarily quick enough to avoid issues.

[NetoriusNick]

1 hour ago, GOTSpectrum said:

The fact they are low orbit and light weight with high surface area means they will de-orbit themselves pretty quickly. They have engines on board that are needed to keep them in orbit for any length of time. 

BINGO, passive deorbit is <8years supposed to be <4 years according to musk but with current orbits, (unless they change the sat design to create more drag) its not doable.

I had no other option really for FAST internet, (we have two). my home one the cost to upgrade to a commensurate plan was the same cost as starlink. so I said F IT and have two independent networks now. two years later fibre is now being installed but so far is not actually in the area. (pipes installed but not actual fibre).

I can see your point about relying on one supplier but amazon and one web will get operation within the next 5 years (one hopes) and knowing amazon they will run it as cheap as thy can to price everyone out. 

all of the old sat companies charge 400+ for sub 10mb down connections so I see a lot of people who had old sat internet will take one or more LEOsat internet. This removes the one provider concern.

as for Spacex going bankrupt according to musk's letter a year ago this as he even stated in a follow up letter was a prediction for a 5 year (i think) if they dont get starship working within that 5 year