A few thoughts from enduser

[rapra]

Hello!
Than you for a good work, I would like to share my thougths:

1. displaying hasrate

• display is made to often, each second is to often - it should be 2-3-5 seconds period
• there should be added additional timer for display enabled after generating dag. And then disabled on share found, and enabled again after receiving new job/set work - now it is reading data with no pauses, what causes visible drops on reported hasrate (problem is that hashrate calculator getting time when hashes are not made when new work appears). It will make it more smooth (as claymore miner is with about 5-10 second hasrate report period), and will make better presence for enduser.
  P.S. Now I am testing newest RC3 with patch and it seems that it was inserted in code. If yes, thank you
• there should be added -lowintensity mode, like claymore miner has. I work on pc with a card on which I mining, and with that miner I have 'overloaded pc effect', claymore with lower -ethi option work good (hashrate drop only ocasionally when I make something more intensive) without any discomfort
  sincerelly,
• on 0.12 rc2 I had one crash - ethminer stopped working, but I didnt make a screen. Will testing this version now
  Radek

[rapra]

Just noticed that newest version has high cpu usage, about 50% cpu power usage.
Maybe it is additional problem of that new feature 'smoth timer'?

[rapra]

I just tested 0.12 RC3, RC2 and RC1 - all utilises ~50% CPU and overloading windows

[rapra]

In comparison, claymore ethminer uses 0-1% cpu

[mnik247]

'smoth timer' not present before version 0.12rc3.
I used (in linux) latest git version=0.12rc3. and not fixed more CPU load.
What you set farm-check?

[rapra]

hmm, I used default 500 (not set). Now I changed to 3000 and it changed period between new line with hashrate is apearing - thanks, it is much better. Ethminer ig getting 40% when starting, and 50% after less than 1 second of existing on processes list (even on presentation of version etc. uses 50% cpu). I think that there are some timers which making that problem. Claymore miner dont do that way at all.
I would like very much to use this miner, because it is getting similar hashrates with no fee but I must have not overloaded windows to work on my pc.
I use windows 7 64 bit with Core 2 duo 3.16GHz and 8GB RAM.

[mnik247]

I used 'farm-check' 10 sec.
More details I posted here:
#303 (comment)

[rapra]

Thank you, I ended with 5 seconds for now. It is have no effect with cpu overload, there must be something working a lot in background (timer or something), what is disturbing cpu so much.
How did you get those nice coloring in your miner?

[mnik247]

You can try 10 sec.
I used my custom build with more "colored" console.

[rapra]

Is it possible to you compile that version for me for windows? Maybe it will not freezing?

[transistorgit]

After upgrading to 0.12.0rc2 from 0.11 I also recognized a huge CPU load! It's currently 389% on a seasoned quad core i7.
Output of "top":
`
top - 22:55:01 up 3 days, 8:17, 1 user, load average: 17.13, 17.12, 17.04
Tasks: 188 total, 1 running, 187 sleeping, 0 stopped, 0 zombie
%Cpu(s): 26.7 us, 73.3 sy, 0.0 ni, 0.0 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
KiB Mem : 8170436 total, 5709832 free, 1154980 used, 1305624 buff/cache
KiB Swap: 2097148 total, 2097148 free, 0 used. 6686740 avail Mem

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1507 root 20 0 74.078g 677536 436748 S 394.4 8.3 17593:07 ethminer
1476 root -51 0 0 0 0 S 1.7 0.0 36:35.22 irq/41-nvidia
9755 root 20 0 0 0 0 S 1.7 0.0 0:18.44 kworker/1:1
1235 root -51 0 0 0 0 S 0.3 0.0 30:35.60 irq/37-nvidia
1258 root -51 0 0 0 0 S 0.3 0.0 13:51.25 irq/38-nvidia
1274 root -51 0 0 0 0 S 0.3 0.0 53:48.36 irq/39-nvidia
1279 root -51 0 0 0 0 S 0.3 0.0 36:08.70 irq/40-nvidia
9763 root 20 0 0 0 0 S 0.3 0.0 0:04.64 kworker/2:3
9818 user 20 0 44208 3956 3256 R 0.3 0.0 0:01.06 top
`

[mnik247]

to @rapra
Sorry, but I not use Windows now, I switched to Linux and compile Linux version.

[mnik247]

to @transistorgit
Try increase 'farm-check' to 5-10 sec.
I haven't problem with CPU load.

[rapra]

Found. I set --cuda-schedule auto - this is causing that cpu overloading. I used that because it stabilizing hashrates on higher values, what is better to observe with less --farm-recheck value (or default).
Could someone look at code and try repair high cpu usage with --cuda-schedule auto set?

[transistorgit]

farm recheck is already 2s. Should be slow enough...

[rapra]

OK, tested those options: auto is causing freezing, probably it is setting yield, what set is causing the same freezing effect. When I set to sync, I see higher values appearing sometimes than with spin. Both spin and sync dont freezing windows