Topic: Can't shake the 100% cpu bug. [Sorta Resolved] (Read 5393 times)

using GUIminer with regular -v -w64 settings besides setting it to -f80 when i'm using my rig.

i just thought i'd report in. I just updated to the newest ati drivers today and the %100 CPU bug is gone for me without any other previously mentioned fixes. However there is a very slight decrease in hash rate for my primary vid card: a 1GB 5870 reference card (it is 5-10 Mhash/s[2-4%]) below normal average than before the update that used %100 CPU..... (to me; a small price to keep my CPU from running 100% 24/7 IMO)- my secondary 5770 24/7 dedicated miner card seems to be operating without any perfomance drops/gains at all.....

drivers 11.11 still used 100% of one CPU  of my phenom 955 cores but 11.12 seems to have fixed this issue for me completely with very minor performance drops that i think are worth it- for me anyway. i'm running both GPU's 100% as i type with no loss on CPU usage related to mining anymore. specs below- hope this helps you all figure out the 100% issue 

CPU: phenom 955 (perma-OC to 3.6 Ghz)
MB: MSI GD-70 790fx
RAM: GB Mushkin Enhanced ddr3 (x2 2GB) 1600 7-7-7-20
GPU/s: 5870 + 5770
PSU: 900w Apeiva
OS: windows 7 64-bit


if there's any more system specs/info i can help with fixing the 100% core usage bug please let me know and i'll report back any other settings i'm using- hope this helps

EDIT: I'll also add that both GPUs operate VERY noticably cooler- an average 5-10 C drop in %100 load temps

11.12 didnt solve cpu bug for me ( win7 32)  AND the Mhash went down by 10%..

I skipped 11.11, but installed 11.12 last night, and it also fixes the CPU bug (at least it did for me).  On the flip side, my Mhash/sec went down a little, but I'm happy to have my CPU back

The new catalyst 11.11 resolves this.

I fear there is no good solution other then use old version drivers like 10.12 with sdk 2.1 or have your luck with 11.4 and 11.6 which did not work for me
The 2.4 seems to be the slowest sdk according to many posts on boinc and as well for use on mining.
It also becomes more of a problem since current fast miners need a recent opencl and driver versions, and we can not use one or the other seperate. For now i do the same as most others just set one cpu core to be used by the miner to prevent the rest of the cores running 100% as well.
So for now i am using cat 11.9 with the included sdk 2.5 and do the cpu affinity solution

11.6 / 2.4

I had an issue where after a few different driver installs, in which I had to use Driver Sweeper to clear out the old files.

I just upgraded to Ubuntu 11.10 and now i've got the 100% CPU bug with one GPU. And worse still I csn't install 11.6 (which used to work for me) on the new ubuntu for some reason. Has anyone got it working properly in Ubuntu Oneiric yet?

Oh, you're using 2 seperate monitors. I was about to say "I followed this to the letter, still doesnt work", but now I have to hook up TWO monitors... I was using 1 monitor that had both vga and dvi on the back and I just switch it on the monitor... BRB

Edit: Hooked up my dell 30" via dvi to my main gpu 5830, hooked up a 22 inch hanns G on VGA on my 5870 that normally is used just for mining and usually has no monitor on it. Booted computer with both monitors on and had extend displays as it was turned on. I went to screen resolution and picked "show on 1" and started the miner thats respective for the monitor that is currently showing stuff (the other one automatically switches to standby at this point). That miner is always 100% cpu bug on 1 core. While it is mining, I do "show desktop on 2", the main monitor goes on standby, other moni9tor turns on, and I turn on the miner respective to that gpu. It's still 100% cpu bug. I already force both instances to 1 core so it's still using up 1 core.

What drivers are you using? what sdk? What miners? You're using the same os as me so it cant be that, and i dont thin kthe card difference should do anything since you're using 2 diffrent cards as am i.. this is so frustrating. It's not an option to use 2.1 sdk either because for some reason guiminer as well as phatk 2.0/2.1/2.2 won't run with sdk 2.1

I'm using two real monitors.

Edit: I switch both displays on to set the clocks.  Then I close off a display and start a miner, then close the other display and start the other miner.

Does it have to be a real monitor and/or the same monitor? I tried with 1 real monitor and the other with a dummy plug, 1 attached to each card, and I still get cpu bug. 11.9 release drivers with the sdk that comes with it, guiminr, phoenix 1.6.2, phatk 2.2, -a 250 -k phatk2 VECTORS WORKSIZE=128 FASTLOOP=true AGGRESSION=4 (dont need to specify bfi_int anymore)

Of course it is.  It is just 100% of the core you limited it to.  Granted affinity is better than nothing but you (and everyone else) still have one core blazing away at max frequency (and current draw) doing nothing.

Maybe in 10 years AMD will finally get around to fixing this.

What I did was assigned all the GPUs to one core of CPU in CPU affinity and it runs fine afterwards without the 100% usage.

Guys, cpu bug is driver related, not windows related.

Just use drivers without it. For a single card 11.9 or 11.10
For multiple cards well, something older, try 11.6 or 11.3

Just switch to linuxcoin. It's not really that difficult. Just loading up a usb, insert, and power on.

It does with cgminer.  Totally cripples it.

If you set your miner to a higher aggression, this does not happen (phoenix 1.6.2). I only get a ~3 mhash drop per 5830 with aggression at 16.

Correct.  Just tried it

BES doesn't really work for this. You can reduce the CPU usage, but it will also slow down the GPU hashing rate significantly.

Hmm.  Have you tried that?