Could any one direct me to the Optimized version' downloads for SETI.
This is a triple gtx 2080 ti computer and 64 bits.

Miklos

---

Not sure about optimized seti builds. been building an optimized linux system for mining/seti and gaming.

Idea's for optimizations

Building the kernel with 'make localyesconfig'
Adding (in arch for my cpu) "-march=x86-64 -mtune=haswell -Ofast -faggressive-loop-optimizations(s.c)" to my makepkg.conf file
Looking into using different compilers ( IBM, intel, gcc )
Using powertop to determine optimized power settings
I hardcoded avx256 into my strcmp.hhttps://pastebin.com/raw/et4GvNa6 function in glibc, but not libc yet ( still testing it though, and a bit taboo as it breaks cross-architecture support )

http://mikesworld.eu/download.html

You need to make sure your driver is 431.60 or earlier or you will have errors on all Arecibo VHAR WUs.

---

Grant
Darwin NT

Not sure about optimized seti builds. been building an optimized linux system for mining/seti and gaming.

Then you'd be better off making use of the Special Application for your GTX 1060s, or at the very least some command line values to get considerably more work out of the SoG application.

---

Grant
Darwin NT

Could any one direct me to the Optimized version' downloads for SETI.
This is a triple gtx 2080 ti computer and 64 bits.

Miklos

Looks like you reverted to Windows. So the Lunatic's distro that you were pointed to is probably your best choice. It is also possible that you can run multiple gpu tasks per gpu which will increase your total production while slowing the individual task processing. 2 or 3 is common. You MIGHT be able to do 4.

Last but not least take a look at the command parameters for your gpu tasks. You can usually "tune" and pick up some processing speed while not making the system unusable for your "daily driver".

HTH,

Tom

---

A proud member of the OFA (Old Farts Association).

The best Windows application is the "SoG", but to run this successfully requires a lot of CPU support. At first, do not be tempted to run more than 1 instance per CPU so you can iron out any wrinkles and ensure everything is working correctly. Also this will give yo a chance to "calibrate" your system in terms of tasks completed per hour.
Go to Mike's download site (http://mikesworld.eu/download.html) and get a copy of the Lunatics Installer - make sure you get the 64-bit version. When you run i makes sure you select the nVdia SoG application and not the CUDA one.
You need to make sure that you "free-up" one CPU core for each instance of the GPU application you are running - I see you have a 36 core CPU and 3 GPUs,, now there are several ways to achieve this, perhaps the simplest is: in the BOINC GUI go to the "advanced" view, the to the "options" button, the select the "computing" tab - under "usage limits" set "use at most xx % of the CPUs", the default is 100, in your case the easy figure is to go for 90% of the CPU cores, which makes 4 "free'd up", and 32 available directly to BOINC.
That should get you started. Next we can worry about running multiple tasks, command lines an so on, once you know everything is working as it should.....

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Currently only using ~300mb of ram on each, and not working them very hard. Going through the config xml, I could adjust the prority to realtime, and increase the number of tasks per gpu. Going to test that now and post back in a few minutes.

Currently only using ~300mb of ram on each, and not working them very hard. Going through the config xml, I could adjust the prority to realtime, and increase the number of tasks per gpu. Going to test that now and post back in a few minutes.

Kevin, instead of trying to work something out for yourself why not just complete your current work units, remove BOINC, download TBar's All-in-One, unpack it into your home folder, then just double click on the boincmgr add setiathome and away you go.

Just check out my 2 rigs for the results. ;-)

Cheers.

No UI on the rig, like figuring out the technical stuff anyway. Is lunatic's version opensource.

Slightly unrelated but, heard a rumor nvidia might be open sourcing the nvidia drivers in 2020.

Checking them out though, has the app_config.xml on their ftp. :) Thanks

Is lunatic's version opensource.

The Lunatics Installer is a Windows thing, it does not work with Linux Kevin, and I've already pointed out the best alternative for you in my previous post.

Cheers.

@Keven,

Are you running a 4 thread cpu?

Intel(R) Pentium(R) CPU G4560 @ 3.50GHz [Family 6 Model 158 Stepping 9]
(4 processors) 

Is that I think Seti@Home is showing for you. Even with linux you get better performance when the cpu is not running flat out. If it is running flat out you need to take a core offline using the Boinc Manager or your Seti@Home website where you can set it to run 90% or 75%.

I am told the CUDA60 is relatively slow. Under a Windows rig running standard SOG gpu tasks, it takes 7-8 minutes to run 1 task on 1 gpu. My experiments with the gtx 1060 seem to indicate that under Windows it can only run 1 task without slowing down more than running 2 tasks in parallel gain you.

Wiggo has already pointed you to the way to get Gtx 1060's to run down in the 2-3 minute range using Tbar's "All-In-One". To do that you will need to COMPLETELY un-install the version of BOINC you have and install Tbar's AIO someplace handy like your home directory or the desktop.
If you don't succeed in a complete uninstall of the previous BOINC you will have a mess on your hands that is mostly easily fixed by a complete re-install of your OS.

Tom

---

A proud member of the OFA (Old Farts Association).

Funny sig @TOM.

would prefer stick to the open source version of seti, than used precompiled bin's. ( Almost the whole os is compiled from source for the specific cpu ).

Also just reinstalled stock glibc as my mod'ed version still has some kinks.

This os is not solely for seti ( probably run it for a week, before I get back to the crypto stuff ( My brothers hobby rig ) )

Have lowered cpu usage and changed the config.

Stat's look good though

Number of tasks completed 132
Max tasks per day 178
Number of tasks today 0
Consecutive valid tasks 145
Average processing rate 294.03 GFLOPS
Average turnaround time 0.40 days

Number of tasks completed 241
Max tasks per day 293
Number of tasks today 0
Consecutive valid tasks 260
Average processing rate 323.34 GFLOPS
Average turnaround time 0.35 days

No UI on the rig, like figuring out the technical stuff anyway. Is lunatic's version opensource.

Slightly unrelated but, heard a rumor nvidia might be open sourcing the nvidia drivers in 2020.

The Lunatics website has the Linux apps available, not just the Windows stuff too.
http://lunatics.kwsn.info/index.php?action=downloads;cat=1
The Seti applications are open source if you want to compile your own.
https://setisvn.ssl.berkeley.edu/svn/branches/sah_v7_opt/

Heard the same rumors about Nvidia too.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Stat's look good though

Number of tasks completed 132
Max tasks per day 178
Number of tasks today 0
Consecutive valid tasks 145
Average processing rate 294.03 GFLOPS
Average turnaround time 0.40 days

Number of tasks completed 241
Max tasks per day 293
Number of tasks today 0
Consecutive valid tasks 260
Average processing rate 323.34 GFLOPS
Average turnaround time 0.35 days

I take it that those figures are your GPU stats from this rig with the top being the "opencl_nvidia_SoG" app and the bottom the "opencl_nvidia_sah" app while this is your Cuda 60 stats.

Number of tasks completed 103
Max tasks per day 137
Number of tasks today 0
Consecutive valid tasks 104
Average processing rate 138.78 GFLOPS
Average turnaround time 0.79 days

Here's what the 3GB GTX 1060's do on my 3570K rig using the Cuda 0.98 app.

Number of tasks completed 188571
Max tasks per day 18506
Number of tasks today 867
Consecutive valid tasks 10476
Average processing rate 1,246.27 GFLOPS
Average turnaround time 0.52 days

And here's what the 3GB GTX 1060's do on my 2500K rig using the "-nobs" command on the same app.

Number of tasks completed 205467
Max tasks per day 18029
Number of tasks today 362
Consecutive valid tasks 1833
Average processing rate 1,316.74 GFLOPS
Average turnaround time 0.53 days

So not only do you have a long way to go, but you're going to be wasting time trying to compile something better.

Cheers.

Got an ssl security warning there, but is almost defiantly a false positive. Have 5E:16:01:BC:17:68:42:BF:55:8D:3C:A4:8C:8B:EF:41:D4:AB:47:C1:87:EE:C0:64:90:1A:E6:11:55:EE:18:F0 for the sha512sum on https://setisvn.ssl.berkeley.edu, however other sites report it as a different ssl cert.