ive noticed some odd behaviour distributing between 2 identical machines (one is master workstation)
on the master machine, when the lightcache starts, it sits on almost no cpu usage while the ram fills up. this has always been the case on my monster scenes, and ive often wondered what its actually doing at this point.. the ram usage goes up pretty slow ( sometimes only 100meg a sec or slower increase, and if it needs 25 gb ram to render, thats a long wait!) and considering i have some very fast ssds feeding the data in, why am i waiting so long?
anyway thats a side-question.. the main thing ive noticed is my *identical* slave machine, when distributing, fills up the ram to that magic point where the cpu kicks up to 100% usage -way- faster.. i.e. in task manager the ram usage curve goes up much faster, and its pegging at 100% cpu while my main machine is still slowly doing whatever its doing, at 8-10% cpu usage, filling the ram through a drinking straw it seems.
once they have both kicked in they render at the same speed..
point is, if my slave can do it faster, maybe my main machine should too?
on the master machine, when the lightcache starts, it sits on almost no cpu usage while the ram fills up. this has always been the case on my monster scenes, and ive often wondered what its actually doing at this point.. the ram usage goes up pretty slow ( sometimes only 100meg a sec or slower increase, and if it needs 25 gb ram to render, thats a long wait!) and considering i have some very fast ssds feeding the data in, why am i waiting so long?
anyway thats a side-question.. the main thing ive noticed is my *identical* slave machine, when distributing, fills up the ram to that magic point where the cpu kicks up to 100% usage -way- faster.. i.e. in task manager the ram usage curve goes up much faster, and its pegging at 100% cpu while my main machine is still slowly doing whatever its doing, at 8-10% cpu usage, filling the ram through a drinking straw it seems.
once they have both kicked in they render at the same speed..
point is, if my slave can do it faster, maybe my main machine should too?
Comment