406
TestNet Discussion Archive / Re: BIBLEPAY -TESTNET TESTING THREAD-Proof-Of-Distributed-Computing(Cancer Research)
« on: February 26, 2018, 04:54:29 PM »
Alright
This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.
Yeah, I was thinking since the zotac is 1.1ghz and 4 core, it would run rosetta at half the speed of the power hungry PC, and maybe someone could stack 10 together and have a low power monster, but in reality so far what Im seeing is a 30 RAC (in one day) from the zotac and a 250 RAC (per day) in the ryzen. So its not quite worth buying these. But let me finish a full 10 day test before trusting any of these numbers or figures. Ill make a wiki page with a table on it at the end.
Here is a pic of the expirimental Rosetta rig Im working on (for RAC comparisons per watt), the Zotac is the little box that looks like a power supply on the top, the Ryzen is the barebones Thermaltake on the bottom:
Update on the next version - 1094c:
So I did a lot of expirimentation with the threading today, and have come to the conlcusion Jaap had a point, the wallet was using too much power even with the new settings in distributedcomputing mode.
So what I have done is in Distributed Computing mode the miner behaves slightly differently (in the next release). It respects the CPU's thread priority meaning that it can run slower, and we now allow you to tweak the setting.
There will be a new setting called minersleep (the default is 0 for Prod and 250ms for Testnet). If you set a numeric value you override it.
Now Im seeing about 10% proc usage by default, and when I raise the value, it does allow the machine to sleep more (down to more like 4% or so).
1094c is now out there for linux if anyone wants to expiriment. This is a pretty minor change, so if you want to wait for the next release its no big deal.
Im compiling Windows now so we will have this to play with later tonight.
PS I did get the unbanked report working for the sanctuary validation system in unit test mode, but its not going to be ready til sometime tomorrow.
Yeah, I think this is potentially resolved now but let me go over the feature now.
I think we can have the Sanctuary pull the unbanked report in the morning before it processes the RAC from its user file.
Please click on the unbanked report in the pool - its under Reports | Rosetta Unbanked report.
This thing pulls the Non-Arm RAC vs the ARM Rac for CPIDs who have ARM Rac > 0.
So what I think we can do is filter this report in the mornings , while its processing its User file, to find CPIDs who have LESS than 15 NonArm Rac, and Arm RAC > 0 into a little list of Ünbanked CPIDs. So in this case, Rastiks is the Only one who would qualify as of today.
Rastiks Rosetta ID is connected to a certain CPID, that CPIDs RAC would be pulled in with no questions - everyone else would not be included.
Now Im going to work on the Sanctuary side to see if I can make that rule. Still looking into the Miner threadings.
You can lower the number of tasks by limiting the amount of RAM used in the settings. Or you can just reduce the CPU time setting.
We dont need a pool, because the boinc infrastructure is already pooling the RAC for us.
Jaap, regarding one thread using too much power, I already made it sleep 200ms between iterations, it could be that you are running this on a low power machine, I suppose I can put in more granular control over this to see if this can be improved. Right now on my windows dev machine, Im running rosetta 6 tasks : taking 50% of the cpu, the plain vanilla biblepay on 1 thread is taking 13% of the cpu, and its hashing at 270hps in testnet. So of course we have to consider the fact that we *do* want it hashing on 1 thread so we have a baseline level of average security. But nevertheless, Ill add a setting to see if we can control the sleep better, and, have it actually attempt to sleep more often - there is also something that I need to look at for the miner process as far as CPU priority - it might be set to high and not normal, meaning that on certain machines its not listening to your wishes, ill check into all that today.
And have same question like jaapgvk.
****
One (minor) thing: the 'setgenerate 1' option makes the wallet hash at the old speed again (not 1%) on both the linux and windows wallets. I think it was also the case with the previous version.
So, to be clear: with the current wallet, when I do 'setgenerate 1 true', on my quadcore 4 thread system, I'm using about 25% of my CPU.
****
I am on linux and last two versions are back on 1 full thread (1.0.9.2 was 1% of all processor).
Thank and excuse my english. I beleve that you understand my question .
He said We were all here for Greed, and alluded that this project is about money and were not really Christians. Your right, I should have replied with more Love.
If you look at the history, I have tried that, and God actually says to forgive your brother not 7 times but 77 times.
I do forgive him....
Alright, windows 1093 is out there, now lets try to increase our Sanc count to ensure all sancs sync to 999.