I have an older AMD FX 8350 system with 24GB of RAM, In an MSI 990FXA Gaming motherboard at 4.2 GHz on all cores, I've been running it with a Zotac GTX1060 AMP 6GB GPU and an R7 250X 2 GB GPU, My GTX 1060 died after running almost 19 months of 24x7 SETI@home. My GTX card had a 2 yr warranty so I got a replacement. Zotac Sent me an RTX2060 6GB. I've been using Lunatics _Win64 to utilize both GPU cards. but Lunatics _Win64 is not been updated for the new RTX GPU Cards to take advantage of the Processing power of these cards. Does anyone know what's going on updates? also, I have to edit the App_info.xml file to change the seti@home_v7 to read seti@home_v8 because when I run the CPU benchmarked it gives an error.

I also had a GTX 1060 3GB card in the system and took out the RX7250X out but BONIC does not see both the RTX2060 and the GTX1060 GPU's only the GTX1060. so I put pack the R7 250X back in with the RTX2060 as my primary GPU.

Any help will be great,

Kristyanna Virgona - N6KV

---

1st you'll need a cc_config.xml inserted into your BOINC data folder to use both with this content.

<cc_config>
 <options>
   <use_all_gpus>1</use_all_gpus>
 </options> 
</cc_config>

Then go to http://mikesworld.eu/download.html and download the latest Lunatics Win64 0.45 Beta 6 version and run it (be sure that you scroll down and select the SoG app in the Nvidia section) and you should be good to go.

After you get that settled down then we can move on to fine tuning the SoG app, but remember that you must reserve a CPU core for each GPU task run.

Cheers.

I would like to fine tune my system as well.. I have lunatics installed, all my gpu jobs for the last few days have been open_cl_SoG.
I am running a 2700x with dual 2080 ti's, I was hoping to hit 100k daily with this machine, but so far it has not.
I am running 1cpu-.33gpu in my config.xml, with 12 concurrent.,

Any help to make this machine more efficient would be great.

The SoG app works best if it has a full cpu core to support the gpu task. You can also apply the most severe tuning for the app with dual RTX 2080 Ti's.

---

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

A proud member of the OFA (Old Farts Association)

I am running 1cpu-.33gpu in my config.xml, with 12 concurrent

My RTX 2060 is doing around 13.6 WUs per hour. Your RTX 2080Tis are doing around 22.5 WUs per hour.
In the past, 2 at a time has been the preferred value for high end cards- I've tried 2 at a time on my GTX 1070s & my RTX 2060 and it was no better or worse than 1 at a time (unless they were running 1 Arecibo & 1 GBT WU and then output dropped off noticeably), so I've just stuck with one at a time.

It would have been easier to do over the last few weeks, but i'd suggest running 1 WU at a time for a few hours to see just what the baseline performance actually is, then try 2 at a time for several hours & see how things go there.

---

Grant
Darwin NT

I don't understand where you get Work Units from. as some of the jobs vary in size...??
I have dropped to .5GPU:1CPU, 12 concurrent (12 of 16 threads) 85% CPU load
will monitor over the next coupl days

Also on a side note.. I just loaded an astropulse, but its running cuda 100, is that norm?
My app_config.xml for astropulse_v7 is "opencl_nvidia_100"

Astopulsev7 - 1.8M Gflops - completed in 7:17.
only ran one, and now back to (2) blc26
https://www.dropbox.com/s/fxkvltr854x0wm1/1.JPG?dl=0

I don't understand where you get Work Units from. as some of the jobs vary in size...??
I have dropped to .5GPU:1CPU, 12 concurrent (12 of 16 threads) 85% CPU load
will monitor over the next coupl days

Also on a side note.. I just loaded an astropulse, but its running cuda 100, is that norm?
My app_config.xml for astropulse_v7 is "opencl_nvidia_100"

Astopulsev7 - 1.8M Gflops - completed in 7:17.
only ran one, and now back to (2) blc26
https://www.dropbox.com/s/fxkvltr854x0wm1/1.JPG?dl=0

He's probably averaging time for a work units then dividing into an hour.

You are actually running OpenCL, you can see by the executable at the bottom of your photo.

When he talks about 1070, you have to remember that has a different RAM than a 1080Ti so it makes sense that 2 would be the preferrable compared to 3 on the 1080Ti. On top of that, the 1080Ti has a large amount of RAM compared to the 1070 and higher data transfer. When talking about the amount of RAM on a card, only 27% or so is available to OpenCL applications. On ATI cards, that value is higher. Closer to 50%. So you can run 2 at a time if you want but really 3 at a time is the best setting for High End GPUs running SoG.

Z

---

ok, thanks.. i will monitor the 2:1 for today, and if its lower than my prev 3:1 I will revert back to my 3:1 setting.
Also how are credits calculated?.. I am trying to find a way to figure out which is more efficient on my machine astropulse or MB..
I only get a AP once in a while, and I'm looking at 1,8M Gflops in 7mins compared to 17k Gflops on 4min (example).. it appears that AP is much more efficient... so just trying to figure out how the credit calcs are made...
thanks again.

Hahaha.....

I'm sure someone will point you to the actual formula somewhere but I think of it as a random number generator combined with a blender.... If you haven't notice. The faster you crunch a work unit, the lower your score will be... This is because the original score was based on a work unit crunched on a CPU. Since GPU crunch much faster, they are rewarded with a lower score. Some have joke that eventually the credit will go NEGATIVE. Why we call it CreditScrew. But that's my personal opinion.

Z

---

You can read about CreditNew algorithm here:
https://boinc.berkeley.edu/trac/wiki/CreditNew
Yes, AP are more efficient in awarding credit both for time and power consumed. But AP tasks only get generated from Arecibo data and Arecibo has a very spotty observing schedule.

---

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

A proud member of the OFA (Old Farts Association)

How do you all get BSRAC in the millions, I'm only at 50k... I see one of you has 1million Seti Bsrac.
There's gotta be something I'm doing wrong here.. or you all running entire server rows?

I just figured with $3,000 in new gpus I'd hit a little higher.

You'll need to hit this thread if you'd like to improve your performance by about 2.5x, but it will require reimaging (or at least dual-boot.) All the top machines/contributors are using this custom client, and it's only on Linux currently.

(For example, you have to go down to #43 in the Top Hosts listing before a Windows machine appears.)
Edit: Also I don't have 64 video cards in one machine... just 4. :^) Any number above 10 is similarly inflated.

---

How do you all get BSRAC in the millions,

Run lots of machines.

---

Grant
Darwin NT

How do you all get BSRAC in the millions,

Run lots of machines.

Or run a few machines with lots of good GPU's.

Cheers.