PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Aggie The Pew message board (Message 135655)
Posted 2 days ago by mackerelProject donor
Just set my fastest two GPUs to 22... 2070 estimating over a day to do it.
2) Message boards : Aggie The Pew message board (Message 135597)
Posted 3 days ago by mackerelProject donor
I just put on a Vega 56 and already regret it. It drowned out the noise of everything else in the room until I put power save mode on, dropping it from 150W to 135W. Debating if I should NNW and discontinue it... especially as it is quite warm in my house already...
3) Message boards : Aggie The Pew message board (Message 135580)
Posted 3 days ago by mackerelProject donor
Last update: 2019-12-12 18:00:06 10 SkyHighWeFly Aggie The Pew 464 654.45 3 11 mackerel Aggie The Pew 309 983.14 2 12 Van Zimmerman Aggie The Pew 309 878.32 2 26 Kouhki Aggie The Pew 154 937.57 1 32 mfl0p Aggie The Pew 154 885.93 1

I'm in :)
4) Message boards : Aggie The Pew message board (Message 135566)
Posted 3 days ago by mackerelProject donor
If people are doing longer units to maximise credit/time it could be a while before many appear on the score board.

I expect to return my 1st unit around 5pm UTC. Would be sooner if the start time wasn't in the middle of my sleep...
5) Message boards : Number crunching : AMD Navi 5500/5700? (Message 135565)
Posted 3 days ago by mackerelProject donor
Just wondering if anyone is using Navi cards successfully on PrimeGrid? I don't have one, but with the just launched 5500 I'm having another look.

As far as I'm aware, the compute support on Windows sounds like it is broken, with reports that folding@home and seti@home either don't work at all, or produce bad work. In searching the forums here, I only saw a single post with a user reporting problems running AP on it.

I don't need one really, so it is more just to have another different card. If it can't be used safely for compute, there isn't much point in me getting one...
6) Message boards : Aggie The Pew message board (Message 135539)
Posted 4 days ago by mackerelProject donor
Getting GPUs ready now, although they'll be on 17low until the morning when I'll switch over to 21 thanks to where the start time is.

Edit: 1060, 1070, 1080Ti, 1650, 2070, 2x 580 prepared. Possibility for a Vega 56 and another 580 tomorrow evening...
7) Message boards : Aggie The Pew message board (Message 135500)
Posted 5 days ago by mackerelProject donor
I haven't done GFN work in a long time, and don't recall, does it use AVX-512? If so, the X CPUs can be much faster than mainstream models.

Also just tried a GFN22...
1080 Ti ~32h
2070 ~29h

I've never been a fan of long units... might stick with GFN21 and take the credit hit. DYFL is not an option :)
8) Message boards : Aggie The Pew message board (Message 135475)
Posted 5 days ago by mackerelProject donor
FFT sizes I dug up from an old thread:
GFN21: 2048k
GFN22: 4096k
GFN23: 8192k

I assume multiply by 8 to get ram footprint still works, doesn't include non-FFT data. Without doing the testing, this looks like running multi-thread is the best way to go for any of these.

Same GFN21 task as tested on GPU yesterday, but on 8086k (stock). Estimates rounded to nearest hour.
1 task, 1 thread: 74h (37.4W, 2.8kWh/unit)
1 task, 2 threads: 40h
1 task, 3 threads: 29h (63.8W, 1.85kWh/unit)
1 task, 4 threads: 26h
1 task, 6 threads: 25h (81.4W, 2.0kWh/unit)

Note that at stock the CPU boosts to a higher clock with fewer threads, so the scaling to more threads might look worse than it is.

Power used is the reported value for CPU only, does not include the rest of system. I'm not able to easily measure wall power for this system.

I didn't try running more than one task at the same time, I'd expect times to be a lot worse.
9) Message boards : Aggie The Pew message board (Message 135464)
Posted 6 days ago by mackerelProject donor
Time to dust off my GPUs then?

To get a feel for how long things take for a GFN21 unit:
1080 Ti ~9.5h 275W
2070 ~8h 175W
1650 ~23h 67W

At times like this I wonder if I should bin everything before Turing and upgrade...

Anyone done testing on how these tasks vary by core vs mem clocks? Wonder if a core underclock of the Turing card will help power efficiency much.
10) Message boards : Aggie The Pew message board (Message 135369)
Posted 8 days ago by mackerelProject donor
So many Canon users here... I consider my main the 7D2 but I don't use it much in recent years. I also got the 10-22, but find I prefer a fisheye if I'm going wide.

On pixel counts, this reminds me of the endless discussions on photography forums, that usually end up focusing more about the gear than what you do with it. So with that in mind, I'd really love to see pixel densities increase significantly, but it requires a different mindset to cope with it. The biggest one is not to expect every captured pixel to be "perfect". The concept with more pixel density is to allow more flexibility in processing afterwards. Oversampling was mentioned, and is key to making use of this. It should also help push noise into higher spatial frequencies. Where it isn't practical to be eliminated, it should be less visually noticeable. It has been some years since I worked it out, scaling compact camera sensors of the day to APS-C size gave a number around the hundreds of MP and I think that's roughly where it starts to get interesting. The bigger problem preventing this from general usage is more on the storage and processing cost than the sensor technology itself.

Canon just came out with their Ra model and which has 30x optical zoom ... again, will atmospheric turbulence allow one to really use 30x for focusing purposes?

If it's just a software setting, having it and not use it is better than not having it at all. I do wonder how many of those they sell. Anyone more serious into astrophotography would more likely buy a dedicated sensor for the job. People like me doing it cheap would just modify a standard camera to remove the unwanted filtering. In the past I've mixed 450D with filters removed to capture Ha/SII, and blend it with green/blue channels from a standard 600D. Can get some nice looking output even just with Ha+OIII.

If my mount hadn't died I think I would like to get back into deep sky photography this winter.


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.16, 1.06, 1.06
Generated 16 Dec 2019 | 7:33:23 UTC