Cost Benefits to Environment on Old Hardware

[FastScience]

Hi

New user here. I was setting FAH on an Lenono t420, Intel 3920xm, with Xubuntu.

I was wondering if this is to old to be beneficial, because of how much power it draws for processing and the effect of using that much electricity?

Do you think it is ok? The CPU still gets in the 75th percentile on benchmarks.


Someone on YouTube had an opinion it's not worth starting up old computers because of the wasted energy vs a newer PC.

Thanks!!!

[Joe_H]

People on Youtube often do simplistic analysis of the situation. In this case they are probably ignoring the environmental costs of creating that new PC and the costs connected with recycling the old. Both involve energy usage as part of the equation.

So is it worth it for you? That will come down to how much electricity costs you and whether you want to spend the money getting something new. The CPU is capable of processing WUs at a reasonable rate.

On my own choice I still fold on an iMac that is a couple years older and a MacBook Pro that while not as old was designed in the same period as your Lenovo. The MacBook Pro processor is a 3520m.

[Paragon]

Excellent question! On the one hand, by using this hardware, you are producing useful science while giving the old parts a longer life (keeping them out of a landfill, potentially). On the other hand, if the goal is to do as much science as possible for the least amount of power (environmental impact), you might look at recycling these parts and buying newer, more energy efficient hardware.

One other thought...is it cold where you are in the winter? I specifically bought old hardware to make a space heater with (turning electricity into science + heat) for the winter months. This provides useful science while heating my house, offsetting me burning fuel oil for heat (I live in New England)

https://greenfoldingathome.com/2020/05/ ... u-want-to/

[markdotgooley]

This condo in south-ish Florida usually has the air conditioning at 76F (about 24.5C) when I’m at home, and off or at 84F when I’m not. In winter it often needs heating, and that’s electric anyway. If I upgrade my folding rig, maybe the older hardware can be cobbled together to run intermittently on cold nights...

[bruce]

I run on a mixture of old and semi-new hardware. All the logical considerations mentioned above apply to me except for a couple of additional refinements.

1) Heating this house in the winter is a good thing but using an A/C to cool the house in the summer is environmentally costly. That means I MUST shut down thermally inefficient hardware on hot days.

2) I happen to have an excess / of solar panels on my house so the environmental cost of burning [oil/coal?/plutonium/etc.] really doesn't apply to me except for the original costs of the panels. Ask yourself what percentage of "green" energy does your electric utility produce?

[JimF]

The real question is the ratio of the benefit of the science to the "cost", however you calculate it.
No one can answer that, since we don't know what discoveries will be made, or how (or even whether) they will be put into use.
And how much is getting a COVID (or any other treatment) earlier worth? Don't know.

But I assume that it is worthwhile, or I wouldn't do it at all. For that, I keep reasonably up to date hardware (mostly Ryzen 3000 now, or a generation back), for energy efficiency.
I would not suggest using the older hardware on this project, since it slows down all the other work units following it. That is why they give a "quick return bonus".
If they have calculated that reasonably well, then speed matters.

[NormalDiffusion]

I would not suggest using the older hardware on this project, since it slows down all the other work units following it. That is why they give a "quick return bonus".
If they have calculated that reasonably well, then speed matters.

That's not completely true. For example, my old 960's get around 400k ppd on 13425 (each one, one a little bit more, the other one less), where my rvii is around 350k on this project.
So sometimes running old hardware has advantages (the 960's are around 80w for 13425...).

[UofM.MartinK]

Speed matters - but not too much, there is enough parallel work out there that anything that completes the work within the timeout should definitely be useful.

The FAH team can and will adjust these times if it holds back science.

In fact, they are currently actively working on ARM clients, which are even slower.

I think of FAH as a massive-parallel pipeline, were throughput matters more than latency (within limits)

My personal interest is Energy Efficiency in Folding - so I often run some of my rigs with turbo disabled, saving 20-30% energy at 3-5% PPD loss, for example.

But as other posters here, I also heat with the rigs when it's cold - so I will start folding with some less efficient computers and enable the turbos when I need the heat.

The Linux/ARM guys on rasperry pi, for example, will never be close to the energy efficiency of modern GPUs or even CPUs - but they might be many. But Linux/ARM is the road to Android/ARM - and some of those mobile phone chips, while not super fast, are among the _most_ energy efficient systems out there!

Long story short, the answer about what's worth folding on and what not isn't simple, and a very personal one.


But if you are new to folding, and can afford burning the extra 60-90 Watts on your Lenovo (and are not worried about fan noise or wearing it out), I'd say: Go For It!

You will get involved, learn about the project... spread the word... help your workplace using idle compute resources for folding... etc. etc...

But be warned: You might get hooked, and spend $$$ soon on faster and more efficient hardware

[JimF]

I think of FAH as a massive-parallel pipeline, were throughput matters more than latency (within limits)

The rules of the game have probably changed a bit with COVID, since there are a lot of work units out there, and additional devices may be useful, if slow.
But I am wondering whether it will be true as things get back to "normal" (whatever that is). The additional slow ones might just slow the projects down overall.

And remember, the costs of computing are not born by FAH, they are born by the crunchers and the environment. But there are many trade-offs, and not a simple right answer.

[UofM.MartinK]

Post-Covid FAH is definitely a different animal than pre-Covid, I re-joined because of it... and more researchers are now aware.

The Assign Server logic, if not already able to, could prioritize latency-sensitive projects to faster machines/users, on top of the deadline which the project owners can adjust - or the deadline defines the latency-sensitivity then taken into account be the AS.

In general, the multitude of different types of projects should be allow to make good use of many tiers of hardware at very different speed & capabilities.

[JimF]

The Assign Server logic, if not already able to, could prioritize latency-sensitive projects to faster machines/users, on top of the deadline which the project owners can adjust - or the deadline defines the latency-sensitivity then taken into account be the AS.

Yes, I was thinking the same thing. Otherwise, with all of the (very) slow machines mixed in with some very fast ones, I don't know why they bother with QRB at all, at least for the CPUs.
The GPUs may possibly be more consistent. The really old ones probably aren't compatible anyway. I hope they can make progress with their AS logic too.

[Paragon]

What I ended up doing is high-end CPU folding + low power GPU folding in the summer here in CT (Ryzen 9 3950x CPU, Nvidia GTX 1650 GPU), which doesn't fight the air conditioning too bad and isn't enough to give me an electric bill (have 26 solar panels on the roof). In the winter, I switch to GPU folding (last winter was 3 x RX580s, 1 x 980 Ti, 1 x 1070 Ti, 1 x 1080 Ti) to reduce the amount of home heating oil I use (lets me shut off the entire basement heating zone and just heat one room). When its cold upstairs, I fire up the 1000 watt space heater rig (2 x GTX 480s, 1 x GTX 580, 1 x GTX 460). Those old Fermi cards put out a lot of heat, but they all still meet the deadlines. For this coming winter, I replaced the 3 x AMD RX580s with three new GTX cards(1660 Ti, a vanilla 1660, and a 1650) for more efficiency, and I'm considering building another Fermi rig for more heat in a different room. Electric bill in the winter (after solar) was still close to $500, but I used almost no heating oil all winter (was doing a 250 gallon tank a month before I switched to Folding for heat)

[Foxbat]

I recently had to do a cost benefit analysis when the power supply on my 2006 Mac Pro finally packed it in after 13-1/2 years of continuous Folding (down for power outages, updates, new GPU updates, etc., but running the rest of the time). I looked at “new” hardware, but like others pointed out, paying 200-350 USD for a refurbished/used box that isn’t as energy efficient as a modern P/S, processor, and MB for double that price may make more sense since a new machine doesn’t need to be made, offset the resources that will be used to operate the older hardware.

Fortunately for me, I found a refurbished Mac Pro P/S for a good price that should get me back online with my two dual-core 2.66 GHz Xeon Mac Pro for the foreseeable future.

[MeeLee]

Considering that most people throw away PCs due to lack of performance, you can certainly get a few PPDs extra out of them.
However, if you're concerned about the environmental effect, I think it would be good to invest in solar, and get one of the RTX GPUs to fold on instead.

The CPU may be too old to make a significant contribution (even an RTX 3080, the fastest GPU currently available to fold on, in a year it hardly makes a dent in terms of performance), it can still be good enough to feed a powerful GPU.

[FastScience]

Hi

Here is an almost two months update, running 24/7.

Score 1,746,084
WUs 330

And this is a newer MSI laptop, GL63 i7-8750h, Linux also.

Total score 5,439,823
Total WUs 678

Thanks and be Safe
Sam