Overclock.net › Forums › Overclockers Care › Overclock.net BOINC Team › Hardware Optimization › Collatz Conjecture project support thread
New Posts  All Forums:Forum Nav:

Collatz Conjecture project support thread - Page 52

post #511 of 521
The mmonnin method will be the only way, you'll wind up with a bunch of small bunkers.....

Well, with it being FB and one days notice, you will have 2.5 bunkers for one days worth of work.....
post #512 of 521

I was browsing through the database at Collatz, to try and find some Windows results to compare to my 290s in Linux.  None of the results I found were close to the times that I'm seeing in Linux.

 

Average runtime on my 290s is 615 seconds/task.  My 290s are running in Arch Linux at 1000 MHZ core 1300 MHz mem | Driver: AMDGPU-PRO 17.10 | Kernel: 4.12.13

 

gpu.config:

verbose=1
kernels_per_reduction=48
threads=8
lut_size=17
sieve_size=30
cache_sieve=1
sleep=1
reduce_cpu=0

 

If anyone has a a 290/390 or 290x/390x running in Windows, it would be awesome if you could post some times for comparison. :thumb:

post #513 of 521
Quote:
Originally Posted by tictoc View Post


If anyone has a a 290/390 or 290x/390x running in Windows, it would be awesome if you could post some times for comparison. thumb.gif

Hard to find as most of the top machines are now running some flavor of 1080 or combination of 7970 (7990's)

Will see what I can find. but 10:15 for a single WU is a great time. I doubt if anyone running windows will match it....

Update:

Ian @ # 39 here is running a pair of R9 290's in windows seven and his times seem to fall between 1,000 -1,500 seconds and he is the highest one on the current list. but we don't know what config he is using. I suspect a fairly tame one.

CLIFF HANGER @ # 46 here is running a pair of 390's in Windows 10 and his times appear to hold a bit under 1,000 seconds. And he is the highest one currently running 390's. but also we do not know his config...

Hope that is what your looking for.

Of course I'm running in the 760-780 range on all mine... thumb.gif
Edited by Egilman - 9/26/17 at 1:00pm
post #514 of 521

Thanks for taking a look.  I dug through the top 200 or so users, but didn't see anyone form the OCN team that was running Collatz on Hawaii.  Hard to do a comparison without known what config was used.  Would be interesting to see if it really is faster in Linux.

I can actually run those tasks significantly faster than the times they are running at right now.  In Windows I had those 290s at 1265 MHZ core, so once I flash back to that BIOS they will really be cranking along. :devil: 

 

*EDIT*  Another interesting thing is the 290s are significanly faster than my 480, which is also running in Linux on the same driver with the same gpu.config.  My 480 averages about 765 seconds/task. 

I haven't had a chance to see why there is such a big difference in runtimes between the 290 and 480.  At the current clocks that I'm running, the 480 should be at least a little quicker than the 290.  :headscrat

post #515 of 521
I have a R9 390 that I could test.

Testing it now with that config you posted ---
My config -
Windows 10 Pro x64
MSI R9 390 @ 1040Mhz core, 1500Mhz mem
Catalyst driver 17.9.2
Edited by spdaimon - 9/27/17 at 7:32am
post #516 of 521
Quote:
Originally Posted by spdaimon View Post

I have a R9 390 that I could test.

Testing it now with that config you posted ---
My config -
Windows 10 Pro x64
MSI R9 390 @ 1040Mhz core, 1500Mhz mem
Catalyst driver 17.9.2

 

Thanks spdaimon. :)

post #517 of 521
Its running run ~1080-1100 seconds. Thats stock factory OC? speeds. Haven't really tried OCing it. Could probably shave it down to less than 1000 I bet.
Had to finish this post on my phone. Friggin Comcast crapped out in the middle of my post.




Upped it to 1100 core. shaved a minute off the time, still 17 min+. rolleyes.gif
Edited by spdaimon - 9/27/17 at 2:22pm
post #518 of 521

That is interesting.   Not sure why there is such a huge difference in runtimes. :headscrat 

 

Once I get the OC back on my 290s, my runtimes will be creeping up on my 1070 that is running at 2100 MHz. eek.gif

 

I might have to pass one of the 290s through to a Windows VM, to see what that looks like.  When I was running the 290s in a Windows 8.1 VM, my MilkyWay times were pretty close to a native install.  I only saw about a 4-6% decrease in performance compared to running them in a native Winodws install. 

 

*Edit*

Quote:
Originally Posted by spdaimon View Post

Its running run ~1080-1100 seconds. Thats stock factory OC? speeds. Haven't really tried OCing it. Could probably shave it down to less than 1000 I bet.
Had to finish this post on my phone. Friggin Comcast crapped out in the middle of my post.




Upped it to 1100 core. shaved a minute off the time, still 17 min+. rolleyes.gif

Those times seem way too slow.  Looking at my times they seem to be pretty close to what they should be when comparing it to my OC'd 1070 and EG's 7970/90s.

Did you change the config after you downloaded work?  If I remember correctly, the gpu.config will only be applied to tasks that are downloladed after the file is changed. 

post #519 of 521
Oh, maybe... I attached to the project, then put the gpu.config file in the folder. So probably there are a bunch of tasks that are running default. I'll abort the work I have and try again.
post #520 of 521
If anything restart the client to read the configs.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware Optimization
Overclock.net › Forums › Overclockers Care › Overclock.net BOINC Team › Hardware Optimization › Collatz Conjecture project support thread