SETI@Home Wow!-Event 2019

[Modular]

@OrangeKid what nvidia card are you running?

 

[emptykoolaidcup]

All signed up! Missed this last year!

<thinks>... Nope, I participated last year apparently! #OldMan

 

[Orange Kid]

@OrangeKid what nvidia card are you running?

RTX 2070

 

[mindless1]

Heh, the title of the topic made me wonder if you'd found aliens.

 

[StefanR5R]

For those who haven't registered themselves yet for the Wow!-Event:

In order to double-check your "SETI@home name" and your "SETI@home User ID" which you need to enter at the registration page, you would normally visit your user page at setiathome.berkeley.edu and find them at the top left. (Name is topmost, but you probably recall that anyway. The ID is a few rows below and called "User ID/ Used in community functions".)

But today is Tuesday. Hence setiathome.berkeley.edu says: "SETI@home is temporarily shut down for maintenance."

But luckily, you don't need to wait for setiathome to come back. Instead, you can look up the info locally. Navigate to the BOINC data directory. With some luck, you may have a file called "sched_reply_setiathome.berkeley.edu.xml" in it. Open it in any text viewer. Fairly at the top, you will find <userid> and <user_name>, whose values you enter at the Wow!-Event registration site.

 

[biodoc]

Thanks. Makes no sense to me as when making the changes in the 'app' and writing to the xconfig you have to 'authenticate'.

Another thing to try is to run 'sudo nvidia-xconfig' and then reboot.

How did you install the nvidia driver?

 

[Orange Kid]

Another thing to try is to run 'sudo nvidia-xconfig' and then reboot.

How did you install the nvidia driver?

Installed the driver through the driver manager. It was listed there, so I checked it and whaala.
The cuda drivers worked right away, but found a post here to install generic opencl headers?
sudo apt install ocl-icd-libopencl1
Anyway those work now. Running some Milkyway as Seti was down. Only problem is the resolution change not staying after a reboot. If I ever get it all sorted I will hopefully not have to reboot much sooooo not a big deal.
Now that Seti is back going to try some. Hopefully I have lunitics apps configured correctly.

 

[Orange Kid]

Well, all that and GPU tasks now take 7 minutes instead of 5 on Windows. Disappointing to say the least.
CPU tasks are a bit longer too.

Edit: These are the SoG variety on GPU. Lunitics apps all around

 

[Modular]

@Orange Kid - seems odd that things are running more slowly. Do you have the same issue on Linux? You should be processing WU's much more quickly on Linux.

 

[VietOZ]

OK, do a fresh install of Mint
Update after install
For Turing cards like 1660ti, RTX 2xxx. You'll need to have driver 418 or later

to install nvidia driver:

sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt update

Run update on the update manager, the little icon next to the clock.
Then open Drivers Manager and install the latest driver from there.
Install OpenCL

sudo apt install nvidia-cuda-toolkit

Reboot after it's done.

Don't install Boinc. Download
http://www.arkayn.us/lunatics/BOINC.7z

Save it to wherever you want. I just left it at Downloads.
Right click on the file and select "Extract here"
You'll see a Boinc folder. Open it ... find the file boincmgr
Double click on it. And you should be in business
If you see the project box empty, then cancel it. Double click boincmgr again and it should have the projects appear.

Good luck with the event!

Edit: forgot to mention that don't run more than 1x per card. And leave a cpu thread open for each task

 

[biodoc]

Edit: These are the SoG variety on GPU

The SoG GPU app is a standard project app. See VietOZ's post above.

 

[StefanR5R]

I performed a subset of @VietOZ's steps myself. I had a well running OS + driver already. Hence I updated nothing of that. Also, I continue to use the boinc-client which comes packaged with the OS.

I only copied projects/setiathome/ from the archive into the client data directory and restarted the client. (Beforehand, I completed or cancelled all tasks. It's better to have no tasks in progress before such changes.) This is possible because this works out-of-the-box for the hardware and driver which I have (Pascal, 384).

OK, I did actually a bit more, like experimenting with another driver version. But the above is the gist of it.

 

[VietOZ]

I agreed with @StefanR5R There are definitely more things we can experiment with. But for users that are not familiar with linux, then the easiest/fastest way to get up and running is to use the All in one Boinc. For people who are more familiar with Boinc configurations and linux then you can try -nobs setting/drivers/even % of cpu load to get you a perfect 100% GPU output. Advance linux users can intergrate the AIO Boinc into their existing Boinc, but for regular normal Joe like I was, it's very intimidating. So I always put myself in their shoes to try to help. XS does not participate in this event. And while we're competing against each other on other events, I'd like y'all to do well in this event.

And I'll see y'all on the battlefield at the end of the month
Cheers!!

 

[ao_ika_red]

In less than 24 hours, wow! event will be started.
Check link below whether you're already registered for the TeAm or not.
https://www.seti-germany.de/Wow/stats/teams/2019/30191/TeAm AnandTech

 

[lane42]

Looks like they reset the stats

 

[ao_ika_red]

Looks like they reset the stats

It's like yellow light just before green light appears.

 

[Orange Kid]

@VietOZ Thanks for the setup.
All is going and working fine.
Just have to limit the CPU's and it'll be great.

 

[Assimilator1]

And leave a cpu thread open for each task

Does it have to be a real thread or will a virtual thread do? (running win7 here btw).

 

[burninatortech4]

Does it have to be a real thread or will a virtual thread do? (running win7 here btw).

I've tried both 90% and 95% CPU max utilization (a real core and only a thread, respectively for my 2700x) and leaving a full core for each GPU to use makes the most difference. A single thread at the very least. Not doing so adds a significant amount of time to each GPU task in my experience.

This seems to be project specific. I've seen little noticeable difference in Milkway@Home for example. Even with 100% CPU utilization - though I usually leave at least one thread free.

 

[phoenicis]

XS does not participate in this event. And while we're competing against each other on other events, I'd like y'all to do well in this event.

And I'll see y'all on the battlefield at the end of the month
Cheers!!

Thanks for the well wishes! Y'know you could always help us do well. I wouldn't want you to get bored over the next two weeks

 

[lane42]

I think I see a force coming alive, welcome back Pokey old Friend.

 

[Orange Kid]

I've tried both 90% and 95% CPU max utilization (a real core and only a thread, respectively for my 2700x) and leaving a full core for each GPU to use makes the most difference. A single thread at the very least. Not doing so adds a significant amount of time to each GPU task in my experience.

This seems to be project specific. I've seen little noticeable difference in Milkway@Home for example. Even with 100% CPU utilization - though I usually leave at least one thread free.

On my 2700x what I have done for this event as I am running both cpu and gpu tasks is create an app_config.xml with the following in it.

<app_config>
<project_max_concurrent>16</project_max_concurrent>
</app_config>

What this does is only allow 16 tasks to run a one time. Either 16 cpu tasks or 15 cpu tasks and one gpu task. Without it Seti will use whatever cpu's are available to run both cpu and gpu. IE: without it, Seti will run 16 cpu tasks and one gpu task, 17 tasks total, not giving a whole cpu to the gpu.
An added benefit is if you are using the computer and the gpu task shuts down a cpu task runs.

@Assimilator1 You only need one thread free.

Hope some of this makes sense.

EDIT: GPU tasks are around 60 seconds.
CPU tasks are around 60 minutes.

Much, much better and what I was hoping for..........Thanks all !!!!!!!!!!!!!

 

[crashtech]

"Max concurrent" seems to allow pretty random behavior from what I have seen, it can exclude a GPU in favor of another or in favor of CPU tasks, it's not a great way to control the behavior of multiple devices in a single instance, imo.

 

[Modular]

Good luck guys! I'll probably get my computers up and running before I go to bed tonight. It'll be one less thing to think about in the morning.

 

[Assimilator1]

I think I see a force coming alive, welcome back Pokey old Friend.

He is?

Hmm, I'm having problems with my rig since I've started running GPU tasks, a problem I had before when I ran GPU tasks but I'd forgotten about!
Anyway, the problem is my PC reboots after an unknown time, but always when I'm not using it!
The CPU has recently passed an OCCT stress test for 33+hrs so it's not that, the grx card isn't o/ced. GPU temp atm is 61-65C atm, & it doesn't overheat playing Elite Dangerous either (or reboot/crash), peak CPU temp's are 61-62C. No invalid tasks for SETI, 2 error tasks but that is due to them expiring.
I'm thinking maybe a voltage issue, but nothing is looking dodgy in speedfan atm, although it can't read the 12v line. This is going to be a PITA to pin down........
Any thoughts anyone?