Problem with BlagoMiner for Qbundle on Fast Blocks with Pools [SOLVED] Unoptimized POC1 Plots Keep Scanning Prior Round
-
Watch your miners on Fast Blocks they will stall out and then miss a block. I tested this on multiple pools and a new block will hit and the scan stops and won't start a new scan. So I restart the miner and it picks up the new block. It skips blocks too. Normally both my computers start a block in sync together. When the fast blocks hit they go off on their own and decide when they want to scan the next block.
I think this may be why the Net diff is down. People are losing shares and blocks and may actually benefit some miners too.
-
it did it to me when I forgot to set my miner to another pool I reassigned at
-
@zapbuzz I am assigned and even reassigned and tested on other pools. There may be a cache setting causing this. I am not sure what settings to use for Cache and Cache2 since I just left them as is from Qbundle. I don't run BlagoMiner from the Bundle since it is such a pain to load the plots. I copied and pasted the directory and set my plots and mine from there.
-
I just added bleach to this post
-
@zapbuzz Test the settings for cache & cache2 - one size does not fit all, and will change as your proportion of PoC1 v PoC2 capcity changes.
-
Interesting I readjusted according to that link now Blago miner runs with 43% cpu thanks @haitch
-
@haitch Thanks! Scans POC2 much faster now and POC1 sped up a bit after fiddling around with the cache settings. Still hesitates on fast blocks though and my drives pause to wait for the block. I will keep watching to see if it mitigates over time.
-
Nope still misses blocks all the time during fast blocks and doesn't even attempt to scan them or report that they happen. I restart the miners and it picks up fine. Saved a few good deadlines.
Most miners probably don't even see this happening at all. Good way to keep the diff down though and get a few extra bucks over time as an exploit.
-
The modified version of blago miner seems to have a memory allocation problem. If someone wanna let restart the blago after time, wrote little alg. save it as .bat and save it in Blago Miner Folder:
Parameter:
version: BlagoMiner_see.exe | BlagoMiner_avx.exe
t: seconds when it should restart :: 3600 sec --> 1hour :: max. value 9999 sec@echo off :loop start BlagoMiner_sse.exe timeout /t 1800 >null taskkill /f /im BlagoMiner_sse.exe >nul goto looplovely regards.
-
@mrwho I think you and @CryptoNick are mistaken. I'm running a very large miner, and I see no memory leak/allocation issue. Depending on your cache/cache2 settings it may appear that blocks are skipped, but during the cache allocation time nonces are being processed, but output is suppressed.
To verify, try setting your cache settings to very low values.
-
I agree with @haitch because the first thing that comes to mind when I see block issues was a pc with little memory (all used)
-
my three miners all using the modified Blago running steady for weeks with minimal memory being used.
-
No one has displayed any specs, or configs lol is this fan envy??? c'mon guys!
-
@zapbuzz What specs do you want to see ?
-
@CryptoNick can I please see your system cpu type and speed, memory, storage capacity and blago miner version number
-
@zapbuzz I'm not @CryptoNick , but specs:
Dual Epyc 7301 @ 2.2GHz
64 GB DDR4 ECC
Capacity: ~ 1.1PB, but have a bunch of drives converting.
Blago 1.700911 AVX2Complete the PoC2 (and some PoC1) plots (900TB) in ~ 45 seconds @ 4,500MB/s
-
I just added a bit more bleach oh it is bright
-
@zapbuzz I'm not familiar with the Asmedia card - USB 3.x I presume ?
My drives are all on SAS2 - lots of drive denity without having to manage all the power bricks and cables.
Windows Server 2016.
Try messing with the Cache/Cache2 sizes, you may get better performance lowering them - I'm using 262144 & 65536.
I'm building a miner server that will be for sale, but shipping to Oz will be expensive, over and above the price of the server.
-
i'd love to be admin instead of bleaching my posts I'd delete this thread
-
@zapbuzz Use the eval version in the meantime ....
