- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
https://setiathome.berkeley.edu/hosts_user.php?userid=8728222
It's on, now. Or at least my power bill cries mercy.
It's on, now. Or at least my power bill cries mercy.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Back to bring the PAIN!
- Thread starter Voodoo Rufus
- Start date
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #3
So.....y'all just gonna let me take all your ranks or what?
https://boincstats.com/en/stats/0/user/list/12/0/30199
https://boincstats.com/en/stats/0/user/list/12/0/30199
Alaric
New Member
- Joined
- Dec 4, 2011
- Location
- Satan's Colon, US
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
you can triple your output if you switch to linux 
these are my systems:
https://setiathome.berkeley.edu/hosts_user.php?userid=12474
these are my systems:
https://setiathome.berkeley.edu/hosts_user.php?userid=12474
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #7
What's with your 32, 48, 64 GPUs per system?
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
Those systems have 2, 3, and 9 GPUs respectively. I do this because the systems are so fast that they would normally run out of work during the Tuesday maintenance outages (the server only sends you 100 CPU WUs + 100 WU per GPU)
Making the server think I have more GPUs means the server gives me more work to hold on my cache. Which means I can keep working longer when the project goes down, expectedly or unexpectedly.
You can do this manually by rescheduling GPU jobs over to the CPU but that takes a long time for a large amount of WUs.
I’m achieving this with a customized version of the BOINC client. It’s open source so you can make whatever changes to the code you want and recompile it.
The normal BOINC client has a client side WU limit of 1000, but custom versions have been made to bank more. There is a public version 7.4.44 (Linux) that has a 3000 WU limit
Your real gains will come from moving to Linux and running the Special Cuda application made by petri33. It’s about 3x faster than the SoG apps on Windows. Your 2080ti would be doing units in about 40 seconds.
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #9
I've heard Linux was faster, but not to that level of envious speed.
Nah, not moving to Linux. Still, why is Windows so slow, and why hasn't there been any optimizations past the SoG Lunatics installer?
Nah, not moving to Linux. Still, why is Windows so slow, and why hasn't there been any optimizations past the SoG Lunatics installer?
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
The developer of Linux Special App is Linux only. He does not have the same level of experience to port it to windows. Others have tried, but no one has been able to properly port it to windows and get it working properly. Something about issues syncing the data up due to problems with the drivers or something. I don’t know the details.
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #11
How hard is it to set up the Linux/Seti combo? I haven't played with Linux since a noobtacular Gentoo install 15 years ago.
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
not too hard.
basically these steps:
1. download/install Ubuntu 18.04 LTS
2. install nvidia drivers from the Ubuntu nvidia PPA
3. download the BOINC package and put it in your home folder
4. maybe some small editing of some config files depending on what apps you want to run and what hardware you have
i can write up a more detailed guide if you want, when you want to actually try it. it'll probably take about an hour to get up and running. i would set No New Tasks on whatever system you want to update and exhaust your cache of WUs before the switch, so you don't leave any in limbo.
basically these steps:
1. download/install Ubuntu 18.04 LTS
2. install nvidia drivers from the Ubuntu nvidia PPA
3. download the BOINC package and put it in your home folder
4. maybe some small editing of some config files depending on what apps you want to run and what hardware you have
i can write up a more detailed guide if you want, when you want to actually try it. it'll probably take about an hour to get up and running. i would set No New Tasks on whatever system you want to update and exhaust your cache of WUs before the switch, so you don't leave any in limbo.
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #13
I wouldn't mind a more detailed version if it's not too inconvenient.
Maybe I can set up a cheapo SSD to run it off of.
Maybe I can set up a cheapo SSD to run it off of.
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
a separate SSD would be ideal. disconnect any other drives in the system so that grub doesnt try to install a bootloader to them. i find it's easier to totally segregate windows from linux this way, and to just determine which OS to boot by selecting the boot device from the BIOS.
I'll try to be as detailed as i can:
1. Download the Ubuntu 18.04 LTS (dont get 18.10 for now) .iso file, and burn it to a DVD, or create a bootable USB drive with Rufus.
2. Boot from the disc or USB
3. follow the installation prompts to install Ubuntu to your SSD (make sure it is the only drive connected, as i mentioned before)
4. after it's done, it should prompt for a reboot. remove the installation media.
5. you should now have booted to the desktop. you only need to move some files around, and install a few dependencies, and install the nvidia drivers. you'll have to reboot a couple times
6. get the BOINC "All-in-One" package from here: http://www.arkayn.us/lunatics/BOINC.7z This package includes the special CUDA application by petri33 and the proper config files to use it. The whole package keeps BOINC all in one place and gives you permissions to easily change configurations, but by default it does not auto-start on boot, you have to manually initiate it. you can change it to auto run at boot after you get it up and running properly.
7. extract it to your home folder (you will have a folder called "BOINC" in your home folder after this, extract directly to there, or extract it and manually move the directory).
8. install dependencies for BOINC from the Terminal
9. install nvidia drivers PPA
10. install nvidia drivers. Turing cards require 410 or 415 drivers, but the 410 drivers are slightly faster so i use those.
11. install the OpenCL driver components (these are needed for Astropulse tasks, and lately the nvidia driver install hasnt been including them)
12. reboot
13. check that drivers were installed properly by running nvidia-smi in the terminal
should return something like this:
14. start BOINC by double clicking on the "boincmgr" file in your BOINC directory (which should be in your home folder)
that should be it. it will open BOINC, you attach your project and should start using your GPU and CPU. change how much CPU you want used by adjusting the CPU use percentage in the Compute preferences in BOINC manager.
let me know if you get any hangups. this should get you up and running in the most basic way. after you get it going, I can show you how to make some additional tweaks to fine tune the GPUs, set GPU overclocks and fan speeds via scripts, and switch you over to using a slightly more optimized version of the special app for CUDA 10.0 (the default package uses CUDA 9.1 i believe)
I'll try to be as detailed as i can:
1. Download the Ubuntu 18.04 LTS (dont get 18.10 for now) .iso file, and burn it to a DVD, or create a bootable USB drive with Rufus.
2. Boot from the disc or USB
3. follow the installation prompts to install Ubuntu to your SSD (make sure it is the only drive connected, as i mentioned before)
4. after it's done, it should prompt for a reboot. remove the installation media.
5. you should now have booted to the desktop. you only need to move some files around, and install a few dependencies, and install the nvidia drivers. you'll have to reboot a couple times
6. get the BOINC "All-in-One" package from here: http://www.arkayn.us/lunatics/BOINC.7z This package includes the special CUDA application by petri33 and the proper config files to use it. The whole package keeps BOINC all in one place and gives you permissions to easily change configurations, but by default it does not auto-start on boot, you have to manually initiate it. you can change it to auto run at boot after you get it up and running properly.
7. extract it to your home folder (you will have a folder called "BOINC" in your home folder after this, extract directly to there, or extract it and manually move the directory).
8. install dependencies for BOINC from the Terminal
Code:
sudo apt install libcurl3 libwebkitgtk-1.09. install nvidia drivers PPA
Code:
sudo add-apt-repository ppa:graphics-drivers/ppa10. install nvidia drivers. Turing cards require 410 or 415 drivers, but the 410 drivers are slightly faster so i use those.
Code:
sudo apt install nvidia-driver-41011. install the OpenCL driver components (these are needed for Astropulse tasks, and lately the nvidia driver install hasnt been including them)
Code:
sudo apt-get install ocl-icd-libopencl112. reboot
13. check that drivers were installed properly by running nvidia-smi in the terminal
Code:
nvidia-smishould return something like this:
14. start BOINC by double clicking on the "boincmgr" file in your BOINC directory (which should be in your home folder)
that should be it. it will open BOINC, you attach your project and should start using your GPU and CPU. change how much CPU you want used by adjusting the CPU use percentage in the Compute preferences in BOINC manager.
let me know if you get any hangups. this should get you up and running in the most basic way. after you get it going, I can show you how to make some additional tweaks to fine tune the GPUs, set GPU overclocks and fan speeds via scripts, and switch you over to using a slightly more optimized version of the special app for CUDA 10.0 (the default package uses CUDA 9.1 i believe)
OP
- Joined
- Sep 20, 2001
- Location
- Bakersfield, CA
- Thread Starter
- #15
Thanks. Is it possible to set it up as a dual boot, e.g. giving me a choice which drive to boot off of?
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
you can, but like i said, I prefer to keep them on totally separate drives with separate bootloaders.
if you want to boot one OS vs another, just hit whatever hotkey brings up your boot menu on your motherboard after POST (F12 on your gigabyte board) and select the hard drive you want to boot from. that will allow you to boot whichever OS you want from separate hard drives.
if you want to boot one OS vs another, just hit whatever hotkey brings up your boot menu on your motherboard after POST (F12 on your gigabyte board) and select the hard drive you want to boot from. that will allow you to boot whichever OS you want from separate hard drives.
- Joined
- Mar 18, 2015
Curious what advantage this has over just using GRUB as it was intended. It's literally what grub is intended to do. However you like your system is fine of course, I just don't understand.
I guess the only real world reason I can think of is if you re-install windows often and don't want to reinstall GRUB every time.
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
because it just complicates things.
if you ever remove linux, grub is still there, and you have to jump through some hoops to get the windows MBR back on there. I got tired of dealing with it and just decided to always keep them totally separate.
if you ever remove linux, grub is still there, and you have to jump through some hoops to get the windows MBR back on there. I got tired of dealing with it and just decided to always keep them totally separate.
- Joined
- Apr 3, 2019
- Location
- Manch-Vegas
Newp, imma throw down the gauntlet. In the immortal words of a great band I...am..the...painkiller
- Joined
- Feb 21, 2008
- Location
- Baltimore, MD
for anyone that cares, has an nvidia GPU with CC 5.0 or greater, and willing to run linux.
petri has released a new version of the setiathome v8 nvidia GPU application. version 0.98b1.
this comes in CUDA 9.0 and CUDA 10.1 flavors. (10.1 requires 418+ drivers, FYI)
on high end cards (GTX 1080ti, RTX 2070-2080ti)
expect 10-15% speed increase on Arecibo non-VLAR
expect 25-30% speed increase on Arecibo and BLC VLAR tasks
on lower cards (Maxwell cards, GTX 1050-1080)
expect 10-15% speed increase on all tasks
this is average speedup ON TOP of the already fast v0.97 application, which was about 2-2.5x faster than the windows SoG applications.
My RTX 2070s can do the BLC32 VLARs in about 50 seconds. My 1080tis do them in about 57-60s.
My best system with 10x RTX 2070s, is back to over 1,000,000 RAC
get the whole "All-in-one" package here: http://www.arkayn.us/lunatics/BOINC.7z and run it in your home folder, or cherry pick out the relevant apps and edit your app_info file to use them with your repository BOINC install if you prefer that.
petri has released a new version of the setiathome v8 nvidia GPU application. version 0.98b1.
this comes in CUDA 9.0 and CUDA 10.1 flavors. (10.1 requires 418+ drivers, FYI)
on high end cards (GTX 1080ti, RTX 2070-2080ti)
expect 10-15% speed increase on Arecibo non-VLAR
expect 25-30% speed increase on Arecibo and BLC VLAR tasks
on lower cards (Maxwell cards, GTX 1050-1080)
expect 10-15% speed increase on all tasks
this is average speedup ON TOP of the already fast v0.97 application, which was about 2-2.5x faster than the windows SoG applications.
My RTX 2070s can do the BLC32 VLARs in about 50 seconds. My 1080tis do them in about 57-60s.
My best system with 10x RTX 2070s, is back to over 1,000,000 RAC
get the whole "All-in-one" package here: http://www.arkayn.us/lunatics/BOINC.7z and run it in your home folder, or cherry pick out the relevant apps and edit your app_info file to use them with your repository BOINC install if you prefer that.
Similar threads
