Collatz

bcavnaugh over at EVGA said his 1800X was on par with a 3930k if that gives any insight of the new chips so far.
 
Clock for clock they seem to be keeping up with my intel chips on the few projects I've run so far. The exception to this was R@H where they were dog slow in comparison.
 
Try Prime Grid.... They will actually utilize optimizations like AVX.
 
How so? They did 16M yesterday and 64M the last 7 days. We've done 44M yesterday and 309M last 7 days.

They're gonna need to try a lot harder.
 
Indeed. They've increased from 5 mil to 16 mil a day but still won't escape at that rate.

It was probably lost in transatlantic translation but that's why I said bless (aww, how sweet).
 
When I first started going after them they were around the 15 - 16m daily mark as well. Guessing they just switched things back to Collatz recently.

teamdaily1.php


I've been keeping an eye on them. Since I don't want to receive another $600+ power bill 3 months in a row I've got most of my rigs offline right now. I've seen you've been handling them just fine though. So good job man.
 
Thanks but I'm missing you and your GPU army already Skillz. No AC and around 30 degrees C in the UK over the last few days .... my poor bloody gpus :eek:
 
Yeah it's been that temp plus this whole month. I'm working on getting all my boxen moved to my garage so it doesn't put a strain on my HVAC system, but it's not going to happen anytime soon.
 
Collatz.jpg

Looks like the Gridcoin poolers are starting to do more trickery to gain the system. Unfortunately, there is no word yet on how it impacted the "science".

Work has been temporarily shut down until they can patch some holes and assess the damage.
 
Last edited:
In collatz, it appears that the admin deducted 1.8B points on yesterday result. Ouch. They should do this for the all the days when the trickery occured.

upload_2018-4-5_12-40-35.png
 
Doozer take a look at this thread when you get some time so we can get your GPUs tweaked for maximum points.
 
In my "C:\ProgramData\BOINC\projects\boinc.thesonntags.com_collatz\" folder, I just used the below settings in the nvidia .config file for my 1080ti in my secondary rig. The task completion time dropped from 13 minutes to around 4.5 minutes.
verbose=1
kernels_per_reduction=64
sleep=1
threads=9
lut_size=18
reduce_CPU=0
sieve_size=30
cache_sieve=1
 
Neither. You use them both for different configurations and tweaks.
 
Pros / cons of going the .conf route vs. the app_config route?

For Collatz there really is just the need for the one file that is named the same as the appropriate ATI or NVIDIA exe file, but with the .config file ending. If that config file pushes the gpu to the max, then there is no reason for the app_config.xml file where normally a gpu project might specify to run two instances of the program.
 
Does anyone have the config settings for more modern cards they can post here?
 
this worked well on 10 series nvidia cards and my 2060ko,

verbose=1
kernels_per_reduction=46
threads=8
lut_size=17
sieve_size=30
cache_sieve=1
sleep=1
reduce_cpu=0

my teammate posted this for a 3060ti on linux,

File: collatz_sieve_1.40_x86_64-pc-linux-gnu__opencl_nvidia.config
verbose=1
kernels_per_reduction=64
threads=7
lut_size=18
sieve_size=30
cache_sieve=1
sleep=0
reduce_cpu=0
 
Back
Top