Posts by GT85

1) Message boards : Number crunching : Are these runtimes okay? (Message 1942)
Posted 13 Sep 2019 by GT85
Post:
Just an FYI - if using multiple GPU’s the newest nvidia game ready drivers seem to have optimised my system for gaming. This meant only one of my gpu’s was performing at full power. The other two became significantly slower - by nearly a factor of 10. I swapped my driver for the creator ready option and all went back to normal.
2) Message boards : Number crunching : Your current settings do not allow tasks from this project. To fix this, you can change Project Preferences on the project's web site. (Message 1774)
Posted 23 May 2019 by GT85
Post:
I’ve had the same issue when resuming projects after a while - despite all the right boxes being ticked I cannot get work units. But this is using GRCpool account manager to select projects. The fix for me was to remove the project from grcpool, save and sync host with grcpool and then re-add the project. After that, all worked well.

App config is important to help smooth out the GPU usage curve. Prior to running 2 units, you usage was not a straight line. It jumped all over the place. Running 2 tasks per gpu for me steadies the curve, utilising 98-100% of my gpus all the time. At the end of the day, the result is probably an extra %10.

The biggest difference is to add custom optimisation to the relevant gpu config file in the collatz project folder. Your 1070ti on its own should easily be as fast as my heat-choked 1080’s per WU. I’m getting around 5 mins each, which is made fair with lower WU reward. Still adds extra points at the end of the day though.
3) Message boards : Number crunching : Optimizing the apps (Message 1465)
Posted 24 Jan 2019 by GT85
Post:
On my vega 64’s I run 64 kernels per reduction. The computer isn’t doing anything else but collatz so the system lag I see going from 48 to 64 doesn’t bother moe. WU’s finish faster too. I get about 5 mins 10 seconds on a Mac with vega 64 running via usb c. On a windows machine, with some overclocking I can get it as low as 5 mins 25 ish. Mac is quicker without any overclocking, which is nice as overclocking a gpu in macOS isn’t yet possible as far as I’m aware.
4) Message boards : Number crunching : cpu work (Message 1401)
Posted 10 Jan 2019 by GT85
Post:
Just abort them and re-update. Are you using a project manager or managing settings via your account on Collatz website? I find occasionally with grcpool account manager I get cpu WU’s with a new project when I clearly have them disabled. Once aborted they don’t come back
5) Message boards : Number crunching : Big jobs & short deadlines (Message 1396)
Posted 9 Jan 2019 by GT85
Post:
What gpu are you using?
6) Message boards : Cafe : No longer included in Gridcoin's Project list. (Message 1335)
Posted 24 Dec 2018 by GT85
Post:
Looks like a few hundred GRC landed in Gridcoin Switzerland's faucet just in time for Xmas.
7) Message boards : Number crunching : Running and optimizing on Intel GPU? (Message 1232)
Posted 9 Nov 2018 by GT85
Post:
You should be able to cut the time nearly in half
The settings for Intel GPU that worked for me we’re the same as a 1050ti. If you dig a bit you will find them - can’t remember exactly what they are.

I think it was this - can’t hurt to try!

verbose=1 kernels_per_reduction=48 threads=8
lut_size=17
sieve_size=30 cache_sieve=1
sleep=1
reduce_cpu=0
8) Message boards : News : Use at your own risk (Message 1085)
Posted 26 Sep 2018 by GT85
Post:
What seems like a modest amount today could be much larger in 10 years time. I’ll be happy if grc goes to 10 cents after I hodl for 10 years, though a dollar or two would mean early retirement for me!
9) Message boards : News : Use at your own risk (Message 952)
Posted 30 Aug 2018 by GT85
Post:
Bummer. Was just about to plug a 4th AMD gpu into my MacBook Pro for extra points... the nvidia card I’ve been testing with works, but the remaining time estimate per task is way off, like by >12hrs. Actual time is like 260 seconds, so I run out of work real quick and boinc doesn’t download more tasks for the nvidia gpu. With all amd gpu’s this wont be an issue.
Hope collatz gets back in the grc network soon!!
10) Message boards : News : Use at your own risk (Message 934)
Posted 22 Aug 2018 by GT85
Post:
8 magnitude for 2.5 million is weird. Should be more like 22 at the current level.
What pool do you belong to?
11) Message boards : News : Use at your own risk (Message 929)
Posted 21 Aug 2018 by GT85
Post:
I’ve been crunching Collatz non stop for grcpool 3 and I’ve seen no interruption to my earnings. What I have noticed is I don’t get as much magnitude as when the project first came back online, but that’s surely just down to increasing user numbers competing for the Grc.
Could Gridcoinstats be in error?
12) Message boards : News : Use at your own risk (Message 918)
Posted 17 Aug 2018 by GT85
Post:
Where are you seeing this? In Projects>Pool Status on GRCPOOL website? For me Collatz has always appeared green across all pools and GRC column. Some heavy equipment has joined the crunch for Collatz - my Magnitude is dropping quickly while RAC has not. I've since moved a few big scoring machines elsewhere to help reduce the RAC in the pool I crunch in the hope it makes it more profitable for the computer I've left attached.
13) Message boards : Number crunching : This requires a celebration! (Message 902)
Posted 10 Aug 2018 by GT85
Post:
With the use of an Apple thunderbolt 3 to 2 adaptor you should be able to get it working. I had a sonnet echo express 2 (TB2) and connected it with the Apple adaptor to a pc with tb3 and it worked. In boot camp on a MacBook Pro I installed NVIDIA drivers and it worked. Never tried in macOS. What OS are you running on the Mac Pro’s? They are getting on a bit and if you can’t run high Sierra you are a bit stuffed. But that’s the easy way. With some jiggery pokery after poking around the egpu’ forums you’ll find a script which should enable your system to use an Nvidia card on an Apple OS. With AMD and High Sierra it’s as simple as connect gpu, power on, crunch with all connected GPU’s.
14) Message boards : Number crunching : This requires a celebration! (Message 899)
Posted 10 Aug 2018 by GT85
Post:
Another reason to celebrate: Apple have re-entered the high performance game by supporting multiple thunderbolt 3 egpu’s on their new MacBook pro’s! And AMD gpu’s supported right out the box run faster than they could on max overclock on a windows system!

My most new most powerful machine on Collatz looks like a cyborg octopus with limbs of thunderbolt cables and sonnet egfx enclosures. Should eventually hit an average of 24 million a day once I change out the lone rx 580 for a 3rd Vega 64 :)
15) Message boards : Number crunching : Credit calculation (Message 897)
Posted 9 Aug 2018 by GT85
Post:
I run 3 work units per gpu on primegrid - with the default of 1 my gpu's were only hitting 55-65% ulitizisation. Created app_config.xml and set the amount of gpu per work unit to 0.33, and the same with how much cpu to dedicate alongside the gpu. There are examples online of app_config.xml for primegrid which you can then tweak as you like. I find 3 WU's per GPU is best on anything in the region of gtx 1080, 1080ti or Vega 64 gpu's. With my 1050ti it already takes long enough per task so I leave the default setting of 1 WU per gpu on that computer.
16) Message boards : Number crunching : Credit calculation (Message 896)
Posted 9 Aug 2018 by GT85
Post:
Give ODLK Latin squares a try - short tasks, less than an hour if i recall
17) Message boards : News : Use at your own risk (Message 893)
Posted 8 Aug 2018 by GT85
Post:
For anyone interested - it seems to have been the latest intel THunderbolt 3 firmware that was causing 2 of my machines to lock up. Rolled back to a previous version and computers have remained online since *fingers crossed*.

Another curious thing I’ve noticed - egpu’s don’t seem to be getting recognised by boinc. If I have a 1080ti hooked up to a machine that contains a 1080, that machine is reported as having a gtx 1080’s.

Same on another machine running an internal gtx 1080 and an external 1050ti, it is reported as being 2 gtx 1080’s.


I’m experiencing an issue where 2 of my computers, both equipped with tb3 NVIDIA egpu’s and windows, crash and lock up in the night. I run other projects and all is fine. I have the latest NVIDIA drivers. Does not happen on machines only running gpus internally. Also does not happen on macOS with egpu’s.
Will continue to work on a solution and will let you all know - in case some of you use egpus. I suspect its an NVIDIA software update issue as it’s a new problem. Will roll back tomorrow and test!
18) Message boards : News : Use at your own risk (Message 865)
Posted 2 Aug 2018 by GT85
Post:
I’m experiencing an issue where 2 of my computers, both equipped with tb3 NVIDIA egpu’s and windows, crash and lock up in the night. I run other projects and all is fine. I have the latest NVIDIA drivers. Does not happen on machines only running gpus internally. Also does not happen on macOS with egpu’s.
Will continue to work on a solution and will let you all know - in case some of you use egpus. I suspect its an NVIDIA software update issue as it’s a new problem. Will roll back tomorrow and test!
19) Message boards : News : Use at your own risk (Message 857)
Posted 1 Aug 2018 by GT85
Post:
Why do something that reduces the total power available to Collatz? To handle all the requests from additional users maybe puts some strain on the odd bit of server equipment, but surely the goal all along for this project has been to get as much computing power as possible to have the best crack at solving the puzzle?
20) Message boards : News : Use at your own risk (Message 852)
Posted 1 Aug 2018 by GT85
Post:
Looks like Collatz is back in the GRC network again. Not sure what the blips are all about. Other projects are far from perfect I've noticed since moving some resources elsewhere. Computation errors and the like which seem to have been a problem for some time. At least we have an active admin here :)


Next 20


©2020 Jon Sonntag; All rights reserved