So, I've been running E@H for a while, but I have a question I can't seem to locate an answer to.
I run an MSI Ventus RTX 3090 and have E@H set to run GPU only. I only have the one GPU, but I've seen people with also only one 3090 with RAC in the millions, whereas mine is barely keeping up with 200k.
Am I doing something wrong?
Edit: I am running two GPU tasks concurrently.
Copyright © 2024 Einstein@Home. All rights reserved.
things you're doing
)
things you're doing wrong:
1. Running Windows
2. crunching low reward tasks
#2 solution. if you switch to running Gamma Ray tasks exclusively (FGRPB1G), you get more credit per unit time.
#1 solution. if you switch to Linux, you can run a "special" custom FGRPB1G application which is about 2x faster than the stock application. 2x faster = 2x the points.
_________________________________________________________________________
I see. I have basically zero
)
I see. I have basically zero experience with Linux so I have no idea how to even set that up, outside of a separate boot partition. After that I'm lost :)
David Comeau wrote: I see. I
)
Separate boot partitions is the "old" way of installing Linux. With the advent of cheap SSD's you just install Linux into a separate SSD from Windows and simply boot the Linux installation instead of Windows.
Linux installs are a simple click through via the Live USB installer and you are up and running in less than 15 minutes on a fast SSD and good internet connection.
Good point. I have three
)
Good point.
I have three NVME's in my system, Windows is on two of them in RAID 0 (pointless but I wanted to).
Do you know of any guides to setting up BOINC through Linux?
Please don't stop running the
)
Please don't stop running the Gravitational Wave work units because they have a lower credit reward. They are just as important and there is currently a huge backlog because most people don't have enough memory in their GPU to run them. Thank you.
David Comeau wrote: Good
)
Start here: https://boinc.berkeley.edu/wiki/Installing_on_Linux
If you need help, we may be able to help you... but read up on it first.
Proud member of the Old Farts Association
One of the BOINC devs is
)
One of the BOINC devs is trying to get BOINC back to automatically creating Linux packages for the major Linux distros like Debian, Ubuntu, Fedora etc. so that they are in sync with the Windows packages.
Looks like he is making good progress as I see build artifacts for each distro at the main Boinc Github repo. Should see the current Linux packages available again at the main BOINC website again in the future. Right now the main Boinc website has very old and deprecated versions which should not be used.
Your best way to get a current or almost current Boinc installation is via the Flathub repository.
https://flathub.org/setup
https://flathub.org/apps/edu.berkeley.BOINC
Believe the Boinc Flatpak is at the 7.22.1 level. Looks like the Boinc master release is going to be 7.24 shortly.
Don't know how often Flathub updates their version.
Ian&Steve C. wrote: 2.
)
David,
It looks like you have switched to the "high return" tasks. Which is the "low hanging" fruit.
Congratulations!
This system is running a single rtx 3080 ti under Windows. You should be able to beat his/hers RAC without changing OS's.
You would join the top 50 of e@h.
For most people it has been difficult to run e@h CPU tasks without having them fill up your system cache/buffer and crowd out the GPU tasks.
Assuming you stay with Windows, you can explore other projects that offer CPU tasks and join the higher performing ranks.
Universe at Home is very noticeable in not being a Windows friendly CPU project. You can get something like a 4x improvement in production by switching to Linux.
There are a number of other projects like World Community Grid etc that have interesting CPU projects you may find appealing.
Hope you enjoy your stay in the BOINC world of Citizen-supported Science.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:Ian&Steve C.
)
Ian&Steve C. wrote: #1
)
Ian&Steve C. wrote:
#1 solution. if you switch to Linux, you can run a "special" custom FGRPB1G application which is about 2x faster than the stock application. 2x faster = 2x the points.
You've got my attention. How do I get the "special" custom FGRPB1G app or does it just run under nVidia cards?
Allen