FAH provides a V7 client installer for Debian / Mint / Ubuntu / RedHat / CentOS / Fedora. Installation on other distros may or may not be easy but if you can offer help to others, they would appreciate it.
GPU: nVidia: Zotac GeForce 1070 Ti AMP Extreme (6 of these) - product page here
GPU: AMD: PowerColor Red Devil RX580 8GB Golden Sample (7 of these) - product page here
I am currently testing this out on Windows 10 but would prefer to switch to Linux. I am used to Manjaro Linux so am thinking of switching to Arch Linux for this box, it will be dedicated to F@H.
I understand that:
I will not be able to drive all the GPU's with the 4 core 8 thread CPU that I have here, I'm currently driving 6 1070 ti's with the CPU at 85% utilization on Windows 10, GPU's are each at 80% utilization with temps of 45 - 50 degrees celcius
I wont get max PPD due to the reduction in PCIe lanes to the GPU's. I've read that the 1070 Ti should be getting in the range of 800K PPD, when running on Linux, assuming a 20% reduction, I'm guessing these should be pushing around 680K PPD each
What I would like to know from those that have gone before me is:
Anyone have any pointers on an initial headless setup for a F@H system
Any BIOS tweaks recommended? (I'll try to see if I can get these risers operating at Gen-3 speeds, currently testing at Gen - 2)
With a headless setup, can you still use nvidia-settings to set power-limit, core/memory over-clocks?
Do the Arch users use the AUR package for F@H or do you decompress the deb directly from Standford?
Any tips/tricks/suggestions/criticisms/pointers/gotcha's are greatly appreciated.
EDIT: I'll be documenting my setup and configuration below in the hopes it helps someone else looking to go down this route.
Last edited by Asgaroth on Sat Dec 15, 2018 12:21 pm, edited 1 time in total.
There are two major products that came out of Berkeley: LSD and UNIX. We don't believe this to be a coincidence.
-- Jeremy S. Anderson
Partition the drive.
NOTE: Use the GPT partition scheme for UEFI booting, my drive was already configured with the GPT partition scheme so I didn't need to change it.
We will be creating the following partition layout.
# ip link show
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: enp0s31f6: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP mode DEFAULT group default qlen 1000
link/ether b0:6e:bf:5f:a0:7c brd ff:ff:ff:ff:ff:ff
Setup a systemd unit file to enable DHCP on this interface.
NOTE: This wont activate on the initial boot, we will enable the systemd-network unit file as a post-install task.
Well, one thing I've noticed since switching from Windows 10 to Arch Linux is that my estimated PPD went from 1.2M to 4.3M! That is a massive jump with no hardware changes at all.
I know this is just an estimation on the client and is bound to fluctuate from day to day, but its been running for 15 hours now and it is estimating 4.3M PPD, that roughly equates to around 710K PPD per card, not bad going, I'll keep an eye on it and see how it goes.
Now to do more reading around to see if there are any client tweaks I can do.
I'll update these posts if I need to tweak some other settings (eg: overclocking, I suspect I'll need a virtual X server for that, WIP)
Last edited by Asgaroth on Thu Dec 20, 2018 6:13 pm, edited 1 time in total.
There are two major products that came out of Berkeley: LSD and UNIX. We don't believe this to be a coincidence.
-- Jeremy S. Anderson
I've noticed, that my Core i7 4500u processor got very slow results running the client at -1 or 3 or 4 CPU cores under Windows.
I only have GalliumOS as alternate OS, and haven't been able to run the client on it. Only from Google Chrome browser.
But in Windows, I found the best results with the client were 2 CPU cores on my system, even if the CPU has 4 threads.
The difference between -1 (automatic), 3, and 2 cores was between 1400-1700 credits, and 4500 credits.
The Core i7 4500u is a 2 core 4 thread system.
Load at 2 and 3 cores is identical (86% CPU).
I'm not sure if 4 cores brought it up to 100%, but either way, performance is lowered when more than 2 cores are selected.
In my case, 2 cores also allow the fan to run at mid-RPM range, rather than full speed.