Considering F@H

[compcons]

I just saw the "coronoFOLD" project articles and it got me thinking about the hardware I have that sits idle. So I am trying to figure out what is worth turning on for folding that actually has a positive value. I know it's not mining but I would like to understand if anything is as efficient as cheap, modern hardware.

Most anything can be taken apart and reconfigured if it results in an efficient combo.

System 1
Radeon RX 560 (no power connection)
I7-3770
16 GB RAM
500 GB SSD

System 2 (in the midst of turning this into a linux based or unraid virtual host but can do something else with it)
GTX 970
I7-4930k
32 GB RAM
5x1TB HDD
250 GB SSD
480 GB SSD

System 3 (it's in an HTPC case)
2xRadeon 270
Xeon E5-1620
32 GB RAM
500 GB SSD

HP Elitebook 860 G5 laptop
eGPU thunderbolt 3 with GTX 970 (is this supported?)
Windows 10 (don't want to change this)
I7-8xxx
32 GB RAM
1 TB NVME

Recommendations? Comments? Again, is anything efficient enough that wont cost me a crap ton in electricity (Illinois)?

 

[ao_ika_red]

Those GTX 970 will provide best PPD/W than other GPU.
For comparison: https://docs.google.com/spreadsheets/d/1vcVoSVtamcoGj5sFfvKF_XlvuviWWveJIg_iZ8U2bf0/pub?output=html
Look at "PPD" column.

 

[compcons]

Should I disable cores on anything for power savings?
Is there any disadvantage to running both gpus in the i7 system? I could run two VMs and dedicate a gpu to each but I don't know if that is more useful than just running two off the physical cpu.

 

[Markfw]

Should I disable cores on anything for power savings?
Is there any disadvantage to running both gpus in the i7 system? I could run two VMs and dedicate a gpu to each but I don't know if that is more useful than just running two off the physical cpu.

2 GPUs on one system is fine. Forget disabling cores, no real power savings. Linux gives you 30% more ppd for the same horsepower.

Edit: I don't know anything about thunderbolt3, how it would work or if it would work.

 

[StefanR5R]

The Thunderbolt attached GTX 970 should work, albeit a little slower than your other GTX 970 in a regular PCIe slot. That's because Folding@home likes to have some bandwidth between host and GPU, especially on Windows which performs a lot more copies between system RAM and VRAM. Bigger and newer GPUs are more affected by attachments through fewer lanes than smaller and older GPUs.

As long as energy efficiency matters to you, disable or delete the CPU folding slots which F@h creates by default. Folding on CPUs is nowhere near as efficient as on any halfway decent GPU, unless you had modern high-core-count dual- or quad-socket CPU servers.

 

[compcons]

The video card is in an external enclosure (hence eGPU) and connects to the laptop through thunderbolt 3 connector. TB3 gives close to PCIe speeds. I will have to test and see if the program recognizes it. The OS definitely does and allows gaming on an integrated graphics laptop at about 80-90% of the GPUs performance compared to the same card in a PCIe slot.

 

[compcons]

Thanks for the info. I will probably start with the laptop and use the CPU and eGPU. It's usually just sitting around doing nothing.

I'll do some reconfiguring to virtualize the other machines to free up the hardware to do folding. I can then run the VMs when I need them but use that hardware.

I might throw the Threadripper in as well but it's my main pc and is not terribly energy efficient (4x HDDs, 2070, HBA, and a bunch of flash storage).

 

[Markfw]

Thanks for the info. I will probably start with the laptop and use the CPU and eGPU. It's usually just sitting around doing nothing.

I'll do some reconfiguring to virtualize the other machines to free up the hardware to do folding. I can then run the VMs when I need them but use that hardware.

I might throw the Threadripper in as well but it's my main pc and is not terribly energy efficient (4x HDDs, 2070, HBA, and a bunch of flash storage).

Use that 2070 ! It will crank out well over one million ppd. The 970 ? I can;t remember, but like 300k or less.

 

[compcons]

Laptop with the 970 signup and running. I put in team 198 but that's as far as I got. It doing some stuff. I saw Mark is spending like $40 a month for his farm of threadrippers. This is a drop on the bucket that is my electric bill. I'll turn my freshly built water cooled 2950x and its rtx 2070 to the task tonight. I requested a passkey but haven't got the email yet.

Is there a centralized management function in the advanced tab or did I imagine that?

 

[Markfw]

Laptop with the 970 signup and running. I put in team 198 but that's as far as I got. It doing some stuff. I saw Mark is spending like $40 a month for his farm of threadrippers. This is a drop on the bucket that is my electric bill. I'll turn my freshly built water cooled 2950x and its rtx 2070 to the task tonight. I requested a passkey but haven't got the email yet.

Is there a centralized management function in the advanced tab or did I imagine that?

$40 ? Thats an estimate for one. My electric bill is $800 a month for 20 computers.

 

[StefanR5R]

Is there a centralized management function in the advanced tab or did I imagine that?

Do you mean managing more than one computer, from your main PC?

• On the computers which run Folding@home, configure the local desktop firewall (if there is one active) to allow inbound TCP traffic to port 36330. Restrict this permission to your LAN or even just to your main PC.
• On your main PC, in FAHControl's window, click [Add] at the bottom left. Fill in the connection details of a folding computer in the dialogue. Repeat for all your folding computers.

This is easier if you have hostnames in your LAN, or at least static IP addresses.

This remote control can optionally be password protected. First, configure the password locally on the folding computer (using FAHControl locally, but on headless machines it can also be done without FAHControl by editing a config file). Then, add the password to the connection details of that computer in FAHControl on your main PC. However, password protection of this control protocol makes only sense if you work outside your LAN.

I'll turn my freshly built water cooled 2950x and its rtx 2070 to the task tonight.

Occupying the 2950x, or any other of your CPUs, with Folding@home would be energy-inefficient. If you want to donate your electricity to the cause efficiently, just use GPUs, no CPUs. (Well strictly speaking, each GPU also uses one CPU thread to drive it. More so on Nvidia GPUs, less so on AMD GPUs.)

Furthermore, prefer 1 big GPU over 2 small GPUs. Prefer new GPU generations over old GPU generations. Don't overclock. (Rather underclock. Maybe under-volt if you can do it without endangering stability. These long-running computational workloads need stable voltages.)

 

[compcons]

I am in. Laptop with a GTX 970 quietly chugging away on the basement workbench and the 2950x/RTX 2070 working in the office. Even set to full, the threadripper just does whatever is asked of it, even gaming, while folding. I just need to be sure that I break the habit of shutting down at the end of the day.

Is it wrong to justify GPU purchases that I want but don't need (and won't really use) based on the "betterment of humaniry"? I am considering updating and sending some systems back to colleg with my kid since she doesn't pay for electricity.

 

[Markfw]

I am in. Laptop with a GTX 970 quietly chugging away on the basement workbench and the 2950x/RTX 2070 working in the office. Even set to full, the threadripper just does whatever is asked of it, even gaming, while folding. I just need to be sure that I break the habit of shutting down at the end of the day.

Is it wrong to justify GPU purchases that I want but don't need (and won't really use) based on the "betterment of humaniry"? I am considering updating and sending some systems back to colleg with my kid since she doesn't pay for electricity.

Well, you are asking the wrong people, since everyone here says "buy it if you can, yes help cure cancer",

I am the worst of them.... 16 computers crunching. 24/7/365