Quote:
Originally Posted by LCB001 For sure there is a difference in ppd among WU's within the same project, I've just noticed that 353 and  511 point WU's use a little more cpu than most of the other WU's at least on my rigs, typically the 353's use up to 3-4% while the rest usually hover around 1-2% according to taskmanager. It probably shows up more on my rigs because they are both running dual gpu's.
They seem to impact the 2653's more perhaps because they run faster, Your rigs could be different given different configs and OS's.
It could be within the margin of error but I did notice in the above screenshot that of the two rigs running two 2665 SMP's the one whose ppd's are lower is running a 353 pointer judging by the gpu's ppd, like you say it's super easy to tell... |
I'm not entirely sure, really. But, if it is true, as it has seemingly always been said, that the GPU uses the CPU ONLY for IRQ handling, which I personally believe is the case, then the causes for the differences noted are due to other causes. Making the effect on CPU usage negligible as IRQ's are handled on schedules that don't vary except in rare cases. Rarely because of something software does. Please take my word for it though that if there were any truly significant (in my opinion) effects/interactions in PPD because of different WU combiniations running on the same rig, I believe I would have noticed them. At the same time, I must also admit that I rarely have time to just sit and watch the machines doing their thing. My wife may like to take that on. I hope!!
Nailing them down exactly I think (as a MS-EE) would be very difficult. Certainly not something I would undertake. It's more than a little out of my line of work LOL. I admire you younger guys for spotting and wondering about those little things though. I'm just not of much use or help in cases like this where FAR more is involved than just software.
I normally don't place as much importance on small differences as you do. Just my nature unless I am at work, where I work with electrical/electronic parameters with tolerances measured in terms of fractions of one part per million. I think I'm getting alergic to them in old age.