Folding on Threadripper

[xSotte]

Hi

So I just bought me a AMD 2950x and assembled it yesterday. I installed uBuntu 14.04.1 LTS and been running SMP for about 10 hours now.
The PPD has been showing around 330-370 000 while running all cores at 3,7Ghz. I also have a 1080Ti in this rig, but to get folding working for graphics card feels like a pain when you've no idea what you're doing. I've asked around and got a checklist on what to do, commands to copy+paste and so on. There always seem to be something going wrong though.

So I'm wondering if there's any point to let this try out NACL? I know it's way less PPD, but you get more WUs finished faster.

[toTOW]

With 14.04, you'll have troubles running A7 CPU core because of library compatibility issues. I strongly recommend and upgrade to 16.04 or 18.04.

About GPU folding on Linux, you'll find a lot of similar topics on the forum summarizing the steps needed to install the right drivers for folding.

[xSotte]

Looks like I typed the wrong version.
I'm on 18.04.1. LTS. I downloaded and installed it last night so everything should be up to date.

GPU folding on Linux: I've been talking to a FoldingCoin developer and sent a txt file to follow, but there always seem to be something going wrong.

[SteveWillis]

If you want to direct message me your specific problem(s) I'll take a stab. I've been folding on Linux Mint for a couple of years now and for the most part kept notes on the issues I had to work through. (DarthMouse on Discord)

Personally I'd advise against NACL as I see a lot of problems with it discussed here. Also to me one of the big advantages to folding on linux, at least on mint, is that folding starts on boot even before you log in. If you have it set in your bios to automatically boot after power is restored after an outage your folding is back up and running if your power blinks off for a couple of seconds. I leave my three rigs running even when I go out of town so that's kind of a big deal for me.

[Nathan_P]

There are also guides on a couple of the team forums, EVGA have a good one that i have used a couple of times. I was writing my own for mint until i found that one.

Also thanks for posting the PPD, that would put the 2990WX at about 700kPPD.

[xSotte]

It would be nice to see how much credit a CPU or GPU have earned rather than just a user.
Maybe there's a program for this?

I was told F@H app won't work if the OS detects more than 48 threads. Not sure if that was ment for Windows or in general.
If it's in general then a 2990wx would be useless. I was actually planning on getting the 2990wx, but changed right before I ordered because of this.

I didn't order any DDR4 memory for my Threadripper rig so I took it from my 5820K with a 1080Ti onboard which is turned off for the day.
Probably not a good idea and now I even hate myself for selling some DDR4 (4x4)1 6GB kits that I had before for a 1/4th of what it costs today.

[bruce]

It would be nice to see how much credit a CPU or GPU have earned rather than just a user.
Maybe there's a program for this?

HFM is a Windows program that does that. I don't know of one for Linux.

You only need one Windows machine on your LAN as it can remotely monitor multiple Linux/MacOS-X/Windows machines tracking each slot separately.

It's open-source, so if somebody wants to port it to Linux, I'm sure it would be widely appreciated by those who do not run Windows.

[SteveWillis]

bruce,
There is a utility that sort of runs HFM on linux: mono. I was able to get it to run for a while but certain results were wildly inaccurate and then eventually something changed and it wouldn't run at all. It had something to do with date formatting but I didn't make a note of the exact error. I don't know but suspected the problem was with mono. I couldn't ever get any support so finally just gave up on it but maybe, as you were saying, someone could use that as a starting point.

[chabgood]

I have 4 GPUs, the computer will reboot after a while, but I disable 1 GPU and it will not reboot. I have run nicehash 24 hrs no issue with all 4 GPUS.

[bruce]

I suspect you're having trouble dealing with excess heat / inadequate cooling.

FAH is designed to maximize its throughput on GPUs and CPUs, With that throughput, it's highly likely that FAH will utilize more resources that other applications (possibly including nicehash) and produce more heat .... especially if you overclock or if your case fans don't provide enough airflow to keep the internal temperature down. What does GPU-Z tell you?

The engineer that did the thermal design of your system must not have considered you might have 4 GPUs, all running at maximum power.

All of my GPUs occupy two slots so that their fans have an extra path for hot air to get out of the case. Adding extra fans to provide more circulation can help, too.

[SteveWillis]

Consider upgrading to a bigger/better cpu cooler. When I first started folding and my pc started crashing it was the cpu overheating.

[chabgood]

Ok, for more info, I have an EK water block on the CPU and GPUS, I have a 1600W PSU. 3 GPU's work fine, when I enable a 4th, any card, it will reboot after a while not immediately. the temps are like 55 or so. There is an issue with more than 3 for some reason.

[chabgood]

Btw, there are no errors in the log, causing this to happen.

[bruce]

You're probably talking about (FAH Data directory)/log.txt. That's certainly useful, but there's more information in the science log.

See the signature block in my first post on this page.

Find the science log and either email it to me (I'll PM my email to those who ask) or post it where I can download it.
(FAH Data directory)/work/(WU number)/01/log.txt

The WU number is a single digit found after :WU0 on each line referring to the WU.

F

[chabgood]

Ok, I will give you all 4 logs. What is your email?