...
Kano or Ckolivas - any thoughts? Getting a bit desperate here.
Well ... firstly, as mentioned, I'd prefer to see the overall CPU usage not a screen snapshot at one particular time ... as I've explained already how to.
Run it for 15 minutes and get the top elapsed CPU and elapsed from the API (as I asked before)
Secondly, what is the CPU on that computer and what does
cat /proc/cpuinfo | grep MHz say at the same time.
Thirdly, cgminer CPU usage includes the USB CPU usage ... that other miners will not have counted against them when they mine
They all do USB I/O but other miners CPU usage is not correct since the OS doesn't count the USB they all do, USB accounts for quite a lot of the CPU usage ...
SOLVED! (sort of, eventually)
After days of downtime trying to troubleshoot this high cgminer cpu usage problem I have finally found the culprit, and am happy to say that it is most definitely NOT a cgminer problem, but is - as I originally suspected, a usb controller problem, in my case the controller is nvidia. It appears that some usb controllers just don't handle blocks of usb3 powered hubs filled with eruptors very well, but the problem seems to manifest itself as a cgminer issue showing extremely high cpu usage together with a high zombie/error rate. I'd like to post my findings here as, from scrolling through the thread, I can see that I am not the only user who has had this particular problem, and wrongfully assumed it to be a cgminer issue, so hopefully this will help other users in the future. I'd also like to thank Kano for his suggestions about cpu monitoring previously, as if it weren't for this I might never have found the problem. I'll try to keep it as brief as possible:
I connected a monitor & keyboard to my headless rig to carry out Kano's suggestions and noticed that even though the cpu was at a constant 33% with Xubuntu 12.04 (compared with 37% with 13.04!) it was still only using half it's cores. So, I switched off AMD cool n' quiet in the bios & rebooted. It was then that I noticed how long the rig took to boot up, minutes instead of seconds. I then disconnected all 4 usb bricks of eruptors & rebooted again - it took seconds to boot up as it did when I originally installed the OS. This behaviour only occurred on this one machine, so I knew it was a local problem. I ordered a usb addin pci card (VIA chipset) which arrived this morning, disabled all onboard usb ports in the bios, installed the usb card, plugged the 4 bricks into a powered 4 port hub & the 4 port hub into the new usb card on the rig & fired it up. It was up & running in seconds as per normal. It has now been mining for 4 hours, the cpu usage is ~12%, usb HW error rate is <1%, both of which are the lowest it has ever been.
So there you have it. I probably could have found the problem a lot sooner if a) it had not been a headless rig, b) I had paid more attention & c) I was more knowledgeable with the workings of Linux/Ubuntu. Again, big thanks to kano who put me on the right track, & ckolivas for the best mining software out there - bar none.
Conclusion:
Nvidia should stick to GFX chipsets.
Xubuntu 13.04 performed poorly compared to 12.04 (I reinstalled 12.04)
I need to learn more