GPU version: kernel size tuning and less UI lags

Message boards : Number crunching : GPU version: kernel size tuning and less UI lags

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Sergei Chernykh
Project administrator
Project developer

Send message
Joined: 5 Jan 17
Posts: 388
Credit: 72,451,573
RAC: 0
   
Message 934 - Posted: 29 Sep 2018, 7:00:25 UTC

From what I heard, RTX series has separate integer and FP units, and this project is integer only. This is probably why RTX gets such boost.
ID: 934 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jozef J

Send message
Joined: 24 Jan 17
Posts: 16
Credit: 471,119,114
RAC: 30,498
    
Message 935 - Posted: 29 Sep 2018, 10:40:04 UTC - in response to Message 933.  

dont expect much increase for RTX cards, nearly half of gpu core take that "tensor"..


Just installed my RTX 2080.....averaging 134 seconds a task. :)


Nice numbers )) Thank you for sharing.
Soo all about ..is like sergey wrote in up ..
Custom 2080Ti can get maybe 80-100 sec ..? have you also 2080Ti ?
Saddly prices on custom or msrp are too high .. on rtx 1080ti have 699 usd ... two used 1080ti is about 1000 usd now.
ID: 935 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
vseven

Send message
Joined: 15 Mar 18
Posts: 12
Credit: 586,210,438
RAC: 10
  
Message 936 - Posted: 29 Sep 2018, 15:08:57 UTC - in response to Message 935.  

No, I only got a 2080, no Ti. I'm actually really surprised the performance was that good. My 1080 only eked out 300 second WU's but this was very surprising.

Have to do more testing with other Boinc projects and see if these results are similar or something that is just benefiting this project.



Soooo.....about that 24 or 25 kernel size. lol. :)
ID: 936 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jozef J

Send message
Joined: 24 Jan 17
Posts: 16
Credit: 471,119,114
RAC: 30,498
    
Message 937 - Posted: 30 Sep 2018, 20:30:23 UTC

133 vs 601 sec on some tasks ... Linux Fedora
Fedora 27 (Twenty Seven) [4.18.7-100.fc27.x86_64|libc 2.26 (GNU libc)]
https://sech.me/boinc/Amicable/results.php?hostid=54096&offset=0&show_names=0&state=4&appid= this host with nvidia RTx do task average 133 sec. but there is also some task 600 sec..

<core_client_version>7.10.2</core_client_version>
<![CDATA[
<stderr_txt>
Initializing prime tables...done
X server found. dri2 connection failed!
DRM_IOCTL_I915_GEM_APERTURE failed: Invalid argument
Assuming 131072kB available aperture size.
May lead to reduced performance or incorrect rendering.
get chip id failed: -1 [22]
https://sech.me/boinc/Amicable/result.php?resultid=17599793 is from 600 sec task
this one is from 133 sec task
<core_client_version>7.10.2</core_client_version>
<![CDATA[
<stderr_txt>
Initializing prime tables...done
X server found. dri2 connection failed!
DRM_IOCTL_I915_GEM_APERTURE failed: Invalid argument
Assuming 131072kB available aperture size.
May lead to reduced performance or incorrect rendering.
get chip id failed: -1 [22]
param: 4, val: 0
ID: 937 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
vseven

Send message
Joined: 15 Mar 18
Posts: 12
Credit: 586,210,438
RAC: 10
  
Message 938 - Posted: 1 Oct 2018, 11:13:52 UTC - in response to Message 937.  

They probably have a 2080 and a second older card. BOINC only reports back how many cards you have installed and only the model of one, usually the fastest one. So the 133 is coming from the 2080 and the 600 is from whatever other card they have.
ID: 938 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Jozef J

Send message
Joined: 24 Jan 17
Posts: 16
Credit: 471,119,114
RAC: 30,498
    
Message 939 - Posted: 2 Oct 2018, 23:02:20 UTC

https://foldingforum.org/viewtopic.php?f=83&t=31051&sid=3692e2e3ee44efee1f6c8544bea63eae&start=15
2080Ti amazing impresive 2,2 mil. PPD in folding home
https://docs.google.com/spreadsheets/d/1v5gXral3BcFOoXs5n1M6l_Uo3pZpQYogn6gVlxRPnz0/edit#gid=0

Sorry bit off topic.. but can maybe help in some way for this project ) to show performance for users if not any more
ID: 939 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MagicEye04

Send message
Joined: 26 Apr 18
Posts: 2
Credit: 1,004,920
RAC: 12
  
Message 1150 - Posted: 24 May 2019, 19:56:49 UTC

Hi,

i am using a Vega VII and with standard kernel size 21 i got a run time of 12754,61s. The GPU-Load was about 0,00...1%.
For me it looks as if the GPU is not used.
Now i try 23 - but maybe there is something completely wrong if others have some minutes run times and i need hours?
ID: 1150 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Kellen

Send message
Joined: 14 Nov 17
Posts: 37
Credit: 868,230,768
RAC: 379,951
    
Message 1151 - Posted: 24 May 2019, 20:22:03 UTC - in response to Message 1150.  

Hi MagicEye04,

There is nothing wrong with your computer, so please do not worry about that :)

Presently; the calculations performed for this project are almost exclusively performed with the CPU and almost nothing is done on the GPU. You can run multi-threaded CPU tasks and, with your CPU, your run-time should be approximately 30 minutes per task if you are using all threads.

Regards,
Kellen
ID: 1151 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
MagicEye04

Send message
Joined: 26 Apr 18
Posts: 2
Credit: 1,004,920
RAC: 12
  
Message 1152 - Posted: 24 May 2019, 21:45:21 UTC

But why are there tasks send out that seem to use a GPU?
Wouldnt it be better to send only the CPU tasks?
ID: 1152 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 · 2

Message boards : Number crunching : GPU version: kernel size tuning and less UI lags


©2019 Sergei Chernykh